← Back
Fetching drawings from USPTO…
Embodiments herein provide for an integrated cooling assembly and methods of forming the same. The integrated cooling assembly comprises an interposer, a plurality of semiconductor devices, a first cold plate, and at least one second cold plate. The interposer has a first side and a second side opposite the first side. The plurality of semiconductor devices are attached to the first side of the interposer. A first cold plate is attached to at least one of the plurality of first semiconductor devices. At least one second cold plate is attached to the second side of the interposer.
FIELD
The present disclosure relates to advanced packaging for microelectronic devices, and in particular, cooling systems for device packages and methods of manufacturing the same.
BACKGROUND
Energy consumption poses a critical challenge for the future of large-scale computing as the world's computing energy requirements are rising at a rate that most would consider unsustainable. Some models predict that the information, communication and technology (ICT) ecosystem could exceed 20% of global electricity use by 2030, with direct electrical consumption by large-scale computing centers accounting for more than one-third of that energy usage. A significant portion of the energy used by such large-scale computing centers is devoted to cooling, since even small increases in operating temperatures can negatively impact the performance of microprocessors, memory devices, and other electronic components. While some of this energy is expended to operate the cooling systems that are directly cooling the chips (e.g., heat spreaders, heat pipes, etc.), energy consumption/costs for indirect cooling can also be quite staggering. Indirect cooling energy costs include, for example, cooling or air conditioning of data center buildings. Data center buildings can house thousands, to tens of thousands or more, of high performance chips in server racks, and each of those high performance chips is a heat source. An uncontrolled ambient temperature in a data center will adversely affect the performance of the individual chips, and the data center system performance as a whole.
Thermal dissipation in high-power density chips (semiconductor devices/die) is also a critical challenge as improvements in chip performance, e.g., through increased gate or transistor density due to advanced processing nodes, evolution of multi-core microprocessors, etc., have resulted in increased power density and a corresponding increase in thermal flux that contributes to elevated chip temperatures. Higher density of transistors also increases the length of metal wiring on the chips, which generates its own additional thermal flux due to Joule heating of these wires due to higher currents. These elevated temperatures are undesirable as they can degrade the chip's operating performance, efficiency, reliability, and amount of remaining life. Cooling systems used to maintain the chip at a desired operating temperature typically remove heat using one or more heat dissipation devices, e.g., thermal spreaders, heat pipes, cold plates, liquid cooled heat pipe systems, thermal-electric coolers, heat sinks, etc. One or more thermal interface materials (TIMs), such as, for example, thermal paste, thermal adhesive, or thermal gap filler, may be used to facilitate heat transfer between the surfaces of a chip and heat dissipation device(s). A thermal interface material is any material that is inserted between two components to enhance the thermal coupling therebetween. Unfortunately, the combined thermal resistance of (i) the interfacial boundary regions between one or more TIMs and the chip and/or the heat dissipation device(s), and (ii) the thermal interface material itself can inhibit heat transfer from the chip to the heat dissipation devices, undesirably reducing the cooling efficiency of the cooling system.
Generally speaking, there are multiple components between the heat dissipating sources (i.e., active circuitry) in the chips and the heat dissipation devices, each of which contributes to the system thermal resistance cumulatively along the heat transfer paths and raises chip junction temperatures from the ambient. Such cooling systems can suffer from reduced cooling efficiency due to the design and manufacture of system components.
Accordingly, there exists a need in the art for improved energy-efficient cooling systems, by reducing system thermal resistance, and methods of manufacturing the same.
SUMMARY
Embodiments herein provide for placing semiconductor devices (e.g., one or more GPUs, GPU cores, XPUs, HBMs, DRAM stacks, etc.) on one or both sides of an interposer with liquid cooling on both sides of the interposer. Advantageously, an interconnect distance can be shortened between the semiconductor devices (e.g., GPU cores and HBMs) and thermal issues can be reduced.
A first general aspect includes an integrated cooling assembly comprising an interposer, a plurality of semiconductor devices, a first cold plate, and at least one second cold plate. The interposer has a first side and a second side opposite the first side. At least one semiconductor device of the plurality of semiconductor devices is attached to the first side of the interposer. A first cold plate is attached to at least one of the plurality of semiconductor devices. At least one second cold plate is attached to the second side of the interposer.
In some embodiments, at least one coolant channel of the first cold plate is coupled to at least one coolant channel of the at least one second cold plate by a spacer. A first adhesive layer may be between the spacer and the first cold plate. A second adhesive layer may be between the spacer and the at least one second cold plate.
In some embodiments, at least one coolant channel of the first cold plate is coupled to at least one coolant channel of the at least one second cold plate through the interposer. The interposer may be coupled to the first cold plate and the at least one second cold plate by a spacer.
In some embodiments, at least one semiconductor device of the plurality of semiconductor devices comprises one or more pass-through power TSVs. The first cold plate may comprise one or more electrical connections.
In some embodiments, the at least one second cold plate is directly bonded to the interposer.
In some embodiments, a substrate has at least one opening on a first side of the substrate. The interposer is attached to the first side the substrate. The at least one second cold plate is disposed in the at least one opening of the substrate. The interposer may be attached to the substrate in a central portion of the second side of the interposer. The plurality of semiconductor devices may be disposed on a central portion of a first side of the interposer, and the interposer may be attached to the substrate in a peripheral portion of the second side of the interposer.
In some embodiments, the plurality of semiconductor devices are a plurality of first semiconductor devices. The plurality of first semiconductor devices are attached to the first side of the interposer. The first cold plate is attached to at least one first semiconductor device of the plurality of first semiconductor devices. The integrated cooling assembly further comprises a plurality of second semiconductor devices attached to the second side of the interposer. The at least one second cold plate is attached to at least one second semiconductor device of the plurality of second semiconductor devices.
In some embodiments, at least one second semiconductor device of the plurality of the second semiconductor devices may be an XPU. In some embodiments, at least one second semiconductor device may be a dummy chip.
In some embodiments, the first semiconductor devices are disposed on a central portion of a first side of the interposer. The interposer is attached to the substrate in a peripheral portion of a second side of the interposer. At least one coolant channel of the first cold plate is coupled to at least one coolant channel of the at least one second cold plate through the interposer. At least one coolant channel of the first cold plate is disposed over at least a portion of a backside of each first semiconductor device of the plurality of first semiconductor devices. At least one coolant channel of the at least one second cold plate is disposed over at least a portion of a backside of each second semiconductor device of the plurality of second semiconductor devices.
In some embodiments, at least one second semiconductor device of the plurality of second semiconductor devices is an HBM.
In some embodiments, the first cold plate is directly bonded to the at least one first semiconductor device. In some embodiments, at least one second cold plate is directly bonded to the at least one second semiconductor device.
In some embodiments, at least one coolant channel of the first cold plate or the second cold plate is exposed to a portion of a backside of at least one first semiconductor device or a portion of a backside of the at least one second semiconductor device.
In some embodiments, at least one coolant channel of the first cold plate or the at least one second cold plate is positioned across a first semiconductor device or a second semiconductor device and the at least one coolant channel is not exposed to at least a portion of a backside of the first semiconductor device or at least a portion of a backside of the second semiconductor device.
A second general aspect includes a method of forming the integrated cooling assembly of the first general aspect. The method comprises providing an interposer having a first side and a second side opposite the first side. A plurality of semiconductor devices are attached to the first side of the interposer. The method further comprises directly bonding a first cold plate to at least one semiconductor device of the plurality of semiconductor devices. The method includes attaching at least one second cold plate to the second side of the interposer.
In some embodiments, attaching the at least one second cold plate comprises directly bonding the at least one second cold plate to the interposer.
In some embodiments, the plurality of semiconductor devices are a plurality of first semiconductor devices. The plurality of first semiconductor devices are attached to the first side of the interposer. The first cold plate is directly bonded to at least one first semiconductor device of the plurality of first semiconductor devices. The plurality of second semiconductor devices are attached to the second side of the interposer. Attaching the at least one second cold plate to the second side of the interposer comprises directly bonding the at least one second cold plate to at least one second semiconductor device of the plurality of second semiconductor devices.
In some embodiments, providing the interposer may comprise forming an opening through the interposer. The method may further comprise coupling at least one coolant channel of the first cold plate to at least one coolant channel of a second cold plate using a first spacer between the interposer and the first cold plate and a second spacer between the interposer and the second cold plate.
In some embodiments, the method further comprises providing a spacer material and attaching the first cold plate to the second cold plate using adhesive between the spacer material and portions of the first cold plate and the second cold plate.
The method may further comprise attaching the interposer to a substrate.
BRIEF DESCRIPTION OF THE DRAWINGS
The above and other objects and advantages of the disclosure will be apparent upon consideration of the following detailed description, taken in conjunction with the accompanying drawings, in which:
FIG. 1 illustrates a device package with an external heat sink;
FIG. 2A is a schematic plan view of an example of a system panel, in accordance with embodiments of the present disclosure;
FIG. 2B is a schematic partial side view of a device package mounted on a PCB, in accordance with embodiments of the present disclosure;
FIG. 2C is a schematic exploded isometric view of the device package in FIG. 2B;
FIG. 3 is a schematic sectional view of an example device package, in accordance with embodiments of the present disclosure, that may be used with the system panel;
FIG. 4 is a schematic sectional view of an integrated cooling assembly of a device package, in accordance with embodiments of the present disclosure;
FIG. 5 is a schematic sectional view of another example device package, in accordance with embodiments of the present disclosure, that may be used with the system panel;
FIG. 6A is a schematic sectional view in the X-Z plane of another integrated cooling assembly, in accordance with embodiments of the present disclosure;
FIG. 6B is a schematic sectional view in the Y-Z plane of the integrated cooling assembly of FIG. 6A, in accordance with embodiments of the present disclosure;
FIG. 7 shows a method that can be used to manufacture the device package described herein;
FIG. 8 schematically illustrates an integrated cooling assembly with two-sided cooling, according to some embodiments;
FIGS. 9A-9G schematically illustrates examples of system panels with two-sided cooling, according to some embodiments; and
FIGS. 10A-10C show example methods that can be used to manufacture an integrated cooling assembly with two sided cooling, according to some embodiments.
The figures herein depict various embodiments of the present disclosure for purposes of illustration only. It will be appreciated that additional or alternative structures, assemblies, systems, and methods may be implemented within the principles set out by the present disclosure.
DETAILED DESCRIPTION
As used herein, the term “substrate” means and includes any workpiece, wafer, or article that provides a base material or supporting surface from which or upon which components, elements, devices, assemblies, modules, systems, or features of the heat-generating devices, packaging components, and cooling assembly components described herein may be formed or mounted. The term “substrate” also includes semiconductor substrates that provide a supporting material upon which elements of a semiconductor device are fabricated or attached, and any material layers, features, and/or electronic devices formed thereon, therein, or therethrough. Examples of substrate material that may be used in applications that generate high thermal density include, but are not limited to, Si, GaN, SiC, InP, GaP, InGaN, AlGaInP, AlGaAs, etc.
As used herein, the term “semiconductor device” may refer to a chip, device, processor, Graphics Processing Unit (GPU) core, X-Processing Unit (XPU) (e.g., CPUs, GPUs, any suitable processing unit, etc.), High-Bandwidth Memory (HBM), stacked Dynamic Random-Access Memory (DRAM), etc. As used herein, the term “bump” or “conductive bump” may refer to a microbump, microbump connection, solder bump, solder ball, etc.
As described below, the semiconductor substrates herein generally have a “device side,” e.g., the side on which semiconductor device elements are fabricated, such as transistors, resistors, and capacitors, and a “backside” that is opposite the device side. The term “active side” should be understood to include a surface of the device side of the substrate and may include the device side surface of the semiconductor substrate and/or a surface of any material layer, device element, or feature formed thereon or extending outwardly therefrom, and/or any openings formed therein. Thus, it should be understood that the material(s) that forms the active side may change depending on the stage of device fabrication and assembly. Similarly, the term “non-active side” (opposite the active side) includes the non-active side of the substrate at any stage of device fabrication, including the surfaces of any material layer, any feature formed thereon, or extending outwardly therefrom, and/or any openings formed therein. Thus, the terms “active side” or “non-active side” may include the respective surfaces of the semiconductor substrate at the beginning of device fabrication and any surfaces formed during material removal, e.g., after substrate thinning operations. Depending on the stage of device fabrication or assembly, the terms “active sides” and “non-active sides” are also used to describe surfaces of material layers or features formed on, in, or through the semiconductor substrate, whether or not the material layers or features are ultimately present in the fabricated or assembled device. For example, in some instances, the term “active side” is used to indicate a surface of a substrate that will in the future, but does not yet, include semiconductor device elements.
Spatially relative terms are used herein to describe the relationships between elements, such as the relationships between substrates, heat-generating devices, cooling assembly components, device packaging components, and other features described below. Unless the relationship is otherwise defined, terms such as “above,” “over,” “upper,” “upwardly,” “outwardly,” “on,” “below,” “under,” “beneath,” “lower,” “top,” “bottom” and the like are generally made with reference to the X, Y, and Z directions set forth by X, Y and Z axes in the drawings. Thus, it should be understood that the spatially relative terms used herein are intended to encompass different orientations of the substrate and, unless otherwise noted, are not limited by the direction of gravity. Unless the relationship is otherwise defined, terms describing the relationships between elements such as “disposed on,” “embedded in,” “coupled to,” “connected by,” “attached to,” “bonded to,” and the like, either alone or in combination with a spatially relevant term, include both relationships with intervening elements and direct relationships where there are no intervening elements. Furthermore, the term “horizontal” is generally made with reference to the X-axis direction and the Y-axis direction set forth in the drawings. The term “vertical” is generally made with reference to the Z-axis direction set forth in the drawings.
Various embodiments disclosed herein include bonded structures in which two or more elements are directly bonded to one another without an intervening adhesive (referred to herein as “direct bonding,” or “directly bonded”). The resultant bonds formed by this technique may be described as “direct bonds”. In some embodiments, direct bonding includes the bonding of a single material on the first of the two or more elements and a single material on a second one of the two or more elements, where the single material on the different elements may or may not be the same. For example, bonding a layer of one inorganic dielectric (e.g., silicon oxide) to another layer of the same or different inorganic dielectric. As discussed in more detail below, the process of direct bonding (e.g., direct dielectric bonding) provides a reduction of thermal resistance between a semiconductor device and a cold plate. Examples of dielectric materials used in direct bonding include oxides, nitrides, oxynitrides, carbonitrides, and oxycarbonitrides, etc., such as, for example, silicon oxide, silicon nitride, silicon oxynitride, silicon carbonitride, silicon oxycarbonitride, etc. Direct bonding can also include bonding of multiple materials on one element to multiple materials on the other element (e.g., hybrid bonding). As used herein, the term “hybrid bonding” refers to a species of direct bonding having both i) at least one (first) nonconductive feature directly bonded to another (second) nonconductive feature, and ii) at least one (first) conductive feature directly bonded to another (second) conductive feature, without any intervening adhesive or solder. The resultant bonds formed by this technique may be described as “hybrid bonds” and/or “direct hybrid bonds”. In some hybrid bonding embodiments, there are many first conductive features, each directly bonded to a second conductive feature, without any intervening adhesive or solder. In some embodiments, nonconductive features on the first element are directly bonded to nonconductive features of the second element at room temperature without any intervening adhesive, which is followed by direct bonding of conductive features of the first element to conductive features of the second element via annealing at slightly higher temperatures (e.g., >100° C., >200° C., >250° C., >300° C., etc.), wherein the annealing causes the conductive features to expand faster than the non-conductive features and to bond together.
Unless otherwise noted, the terms “cooling assembly” and “integrated cooling assembly” generally refer to a semiconductor device and a cold plate attached to the semiconductor device. In some embodiments, the semiconductor device may be replaced by a plurality of semiconductor devices on a first side of an interposer. The cold plate may be replaced by a first cold plate and at least a second cold plate. The first cold plate may be directly bonded to at least one of the semiconductor devices on a first side of the interposer, and the second cold plate may be directly bonded to a second side of the interposer opposite the first side of the interposer. In some embodiments, the semiconductor device may be replaced with one or more semiconductor devices on a first side of an interposer and one or more semiconductor devices on a second side of the interposer. A first cold plate may be directly bonded to at least one of the plurality of semiconductor devices on the first side of an interposer, and the at least one second cold plate may be directly bonded to at least one of the plurality of semiconductor devices on the second side of the interposer. Typically, the cold plate is formed with recessed surfaces that define one or more fluid cavities (e.g., coolant chamber volume(s) or coolant channel(s)) between the cold plate and the semiconductor device. In embodiments where the cold plate is formed with plural fluid cavities, each fluid cavity may be defined by cavity dividers and/or sidewalls of the cold plate. For example, cavity dividers may be spaced apart from each other and extend laterally between opposing cold plate sidewalls (e.g., in one direction between a first pair of opposing cold plate sidewalls, or in two directions between orthogonal pairs of opposing cold plate sidewalls). The cavity dividers and the cold plate sidewalls may collectively define adjacent fluid cavities therebetween. While it is preferred that the cold plate is formed of a material whose coefficient of linear thermal expansion (CTE) is the same as or similar to the bulk material of the semiconductor device, in some embodiments the cold plate may comprise one or more materials such as: polymer, copper, aluminum, silicon, glass, or ceramic, for example.
The cold plate may be attached to the semiconductor device by use of an adhesive layer or by direct bonding or hybrid bonding. Direct bonding may include direct dielectric bonding techniques as described herein, and may give rise to direct dielectric bonds. Hybrid bonding may include hybrid bonding techniques as described herein, and may give rise to direct hybrid bonds. For example, the cold plate may include material layers and/or metal features that facilitate direct bonding or hybrid bonding with the semiconductor device. In some embodiments, the backside of the semiconductor device is beneficially directly exposed to coolant fluids flowing through the integrated cooling assembly, thus providing for direct heat transfer therebetween. It will be understood that “coolant fluid” may alternatively be referred to as “cooling fluid”. Unless otherwise noted, the integrated cooling assemblies described herein may be used with any desired fluid, e.g., liquid, gas, and/or vapor-phase coolants, such as water, glycol, etc. In some embodiments, the coolant fluid(s) may contain additives to enhance the conductivity of the coolant fluid(s) within the integrated cooling assemblies. The additives may comprise, for example, nanoparticles of various types, such as carbon nanotubes, graphene, and/or metal oxides. The concentration of these nanoparticles within the coolant fluid may be less than 1%, less than 0.2%, or less than 0.05%. The coolant fluids may also contain a small amount of glycol or glycols (e.g., propylene glycol, ethylene glycol, etc.) to reduce frictional shear stress and drag coefficient in the coolant fluid(s) within the integrated cooling assembly. In some embodiments the coolant fluid may contain entirely glycol or glycols.
Exemplary fluids available for use in the various thermal solution embodiments include: water (either purified or deionized), a glycol (e.g., ethylene glycol, propylene glycol), glycols mixed with water (e.g., ethylene glycol mixed with water (EGW) or propylene glycol mixed with water (PGW)), dielectric fluids (e.g. fluorocarbons, polyalphaolefin (PAO), isoparaffins, synthetic esters, or very high viscosity index (VHVI) oils), or mineral oils. Additionally, depending upon design and operating conditions, these fluids may be used in single-phase liquid, single-phase vapor, two-phase liquid/vapor or two-phase solid/liquid. By adjusting the fluid selection and the relative fluid concentrations in the fluid mixtures, it is possible to alter the thermohydraulic and heat transfer properties by altering the temperatures where phase change occurs, enabling meeting design temperature and pressure conditions for the component being cooled or warmed and the thermal solution being deployed. Additionally, different combinations of the fluid phases may be employed in various hybrid configurations to meet the particular cooling or warming needs of a respective implementation and still be within the scope of the contemplated embodiments.
Additionally, in some embodiments part or all the cooling is provided by gases. Exemplary gases include atmospheric air and/or one or more inert gases such as nitrogen. Atmospheric air may be taken to mean the mixture of different gases in Earth's atmosphere made up of about 78% nitrogen and 21% oxygen.
Depending on the design needs of a thermal solution system using the disclosed embodiments, engineered dielectric coolant fluids may be used. As used herein, a dielectric coolant fluid is a fluid that is thermally conductive but not electrically conductive. Some examples of dielectric fluids used for cooling semiconductors include: 3M™ Fluorinert™ Liquid FC-40-A non-flammable, dielectric fluid that can be used in direct contact with live electronics; 3M™ Novec™ Engineered Fluids-A non-flammable, dielectric fluid that can be used in direct contact with live electronics; Galden® PFPE (perfluoropolyether) products used as heat transfer fluids; EnSolv Fluoro HTF-A solvent with a high boiling point and low pour point that can be used for semiconductor wafer cooling. It is understood that in the selection of the coolant fluid, system design aspects such as operating temperatures and pressures, fluid flow rates, fluid viscosity, and other properties will require evaluation when selecting the appropriate coolant fluid.
In some embodiments, the coolant fluids may contain microparticles and/or nano-particle additives to enhance the conductivity of the coolant fluid within the integrated cooling assemblies. Nanofluids are engineered fluids prepared by suspending the nano-sized (1-100 nm) particles of metals/non-metals and their oxide(s) with a base/conventional fluid. The suspension of high thermal conductivity metals/non-metals and their oxides nano-particles enhances the thermal conductivity and heat transfer ability, etc. of the base fluid. The additives to the underlying coolant fluid may comprise for example, nano-particles of carbon nanotube, nano-particles of graphene, or nano-particles of metal oxides. When the coolant fluid contains microparticles, the microparticles are typically 10 microns or less in diameter. Silicon oxide microparticles may be used.
The volume concentration of these micro or nano-particles within the coolant fluid may be less than 1%, less than 0.2%, or less than 0.05%. Depending upon the liquid and micro/nano-particle type chosen for the coolant fluid, higher volume concentrations of 10% or less, 5% or less, or 2% or less may be used. The coolant fluids may also contain small amounts of glycol or glycols (e.g. propylene glycol, ethylene glycol etc.) to reduce frictional shear stress and drag coefficient in the coolant fluid within the integrated cooling assembly. The availability of different base fluids (e.g., water, ethylene glycol, mineral or other stable oils, etc.) and different nanomaterials provide a variety of nanomaterial options for nanofluid solutions to be used in the various embodiments. These nanomaterial option groups such as aforementioned metals (e.g., Cu, Ag, Fe, Au, etc.), metal oxides (e.g., TiO2, Al2O3, CuO, etc.), carbons (e.g. CNTS, graphene, diamond, graphite . . . etc.), or a mixture of different types of nanomaterials. Metal nano-particles (Cu, Ag, Au . . . ), metal oxide nano-particles (Al2O3, TiO2, CuO), and carbon-based nano-particles are commonly employed elements. Silicon oxide nano-particles may also be used. Using coolant fluids with micro and/or nano-particles when practicing the various embodiments disclosed herein can result in increased heat removal efficiencies and effectiveness.
The fluid control design aspects of specific embodiments may require the nanofluids to be magnetic to facilitate either movement or cessation of movement of the fluids within the semiconductor structures. Magnetic nanofluids (MNFs) are suspensions of a non-magnetic base fluid and magnetic nano-particles. Magnetic nano-particles may be coated with surfactant layers such as oleic acid to reduce particle agglomeration and/or settling. Magnetic nano-particles used in MNFs are usually made of metal materials (ferromagnetic materials) such as iron, nickel, cobalt, as well as their oxides such as spinel-type ferrites, magnetite (Fe3O4), and so forth. The magnetic nano-particles used in MNFs typically range in size from about 1 to 100 nanometers (nm).
This disclosure describes embodiments involving the architecture of system and component elements that can be employed to provide for the cooling of semiconductor components, packaging, and boards. However, those skilled in the art will appreciate the disclosed components and arrangements can be deployed and used in scenarios where component heat up or thermal warm up is desired for a component that is currently outside the low end of the desired operational range. Components that are outside the low end of their operational range can, if started in a cold environment, experience thermal warping or cracking up to and including thermal overexpansion and contact separation that may impair the successful operation of the system. Therefore, in these scenarios, the architectures and embodiments disclosed herein can be used where the indirect thermal solutions supporting them are repurposed or operated in a hybrid configuration to provide warming fluids or heat transfer media to accomplish the warm-up or heat-up scenario. These scenarios are controlled by systems not shown here to bring temperatures up at a speed or timing that enables the materials to avoid the excessive thermal expansion or unequal thermal expansion that may occur among the materials of the semiconductor or packaging being serviced by the thermal solution. Once the component or packaging is brought up into the normal operating range, it can be safely started and brought to a useful operational state.
Considering the warm-up or heat-up embodiments introduced above, the balance of this disclosure and terms used should be viewed in a light that also considers the design option for such warm-up or heat-up. Thus, where terms such as cooling channel, cooling chamber volume, and cooling port are used, for example, such terms could also be considered as a thermal control channel, a thermal control volume, or a thermal control port, respectively. A person of skill would understand that heat flux or heat transfer would go in a different direction, but the design concepts are similar and can be successfully employed in the various embodiments.
In some embodiments, a cooling channel is a liquid cooling channel, and a liquid may flow through the liquid cooling channel. In some embodiments, the liquid may comprise a water and/or glycol (e.g., propylene glycol, ethylene glycol, and mixtures thereof).
As described below, coolant fluid flowing through a cold plate may be used to control the temperature of semiconductor devices. The fluid flowing across the surface of the semiconductor device absorbs heat and conducts heat away from the semiconductor device.
Semiconductor devices that target Artificial Intelligence applications (e.g., AI chips) can have multiple GPU cores in close proximity to HBMs. Placing multiple GPU cores and HBMs on a two-dimensional (2D) layout or plane can be preferable for thermal considerations, but can result in large interconnect distances. Stacking multiple GPU cores and HBMs can shorten interconnect distances, but can result in thermal issues. Advantageously, by placing GPU cores and HBMs on both sides of an interposer and using liquid cooling on both sides, an interconnect distance can be shortened and thermal issues can be reduced. The interposer may be attached to substrates such as a PCB or a package substrate.
FIGS. 1-7 relate to various aspects of a cooling assembly, which may be applied towards cooling a configuration with semiconductor devices attached to an interposer and cold plates on both sides of the interposer. FIG. 8 may schematically illustrate an integrated cooling assembly with two sided cooling. In some embodiments, a cold plate may be attached to a plurality of semiconductor devices on a first side of the interposer, and at least one cold plate may be attached to at least one semiconductor device on a second side of the interposer (e.g., as shown in FIGS. 9A-9D) or directly bonded to the interposer (e.g., as shown in FIG. 9D). In some embodiments, a cold plate may be attached to a plurality of semiconductor devices on a first side of an interposer, and a cold plate may be attached to a plurality of semiconductor devices on a second side of the interposer (e.g., as shown in FIGS. 9E-9G). FIGS. 10A-10B schematically illustrate a method that can be used to manufacture a system panel with two sided cooling. In some embodiments, at least one cold plate may be attached to at least one semiconductor device (e.g., one or more semiconductor devices 804, 806, and 808 may have a corresponding cold plate in FIGS. 9A-9D).
Although channels of a particular shape, number, or type (e.g., exposed or not exposed to a backside of semiconductor device or interposer) may be shown or described in embodiments the present disclosure, any suitable shape, number, or channel type channel may be used.
Although semiconductor devices (e.g., chips, power delivery chips, etc.) and cold plates may be shown or described in embodiments of the present disclosure as being attached using direct bonding or hybrid bonding, any suitable technique or method (e.g., solder, adhesive, etc.) may be used.
Although a particular configuration for attaching semiconductor devices to interposers (e.g., hybrid bonded), cold plates to semiconductor devices (e.g., directly bonded), or substrates to other substrates (e.g., surface mount technology (SMT), solder balls), any suitable method of attachment may be used (e.g., direct bonding, hybrid bonding, solder, adhesive, etc.)
Although various figures in the present disclosure may show a particular configuration (e.g., type and number) of semiconductor devices attached to interposer, any suitable configuration of semiconductor devices may be used in embodiments of the present disclosure. FIG. 8 may show semiconductor devices 804 and 814 as GPUs and semiconductor devices 806, 808, 816, 818 as HBMs. In some embodiments, the second semiconductor devices 816 and 818 may be dummy chips (e.g., bottom cooling or second cold plates 908, 910 is for heat removal of the top or first semiconductor devices 804, 806, and 808). FIG. 9D may show a semiconductor device 816 as a dummy semiconductor device 816 and there may be no semiconductor device 818. FIG. 9A may show semiconductor devices 816 and 818 as XPUs. A semiconductor device may be any suitable type of semiconductor device such as a GPU, HBM, or a dummy chip. FIG. 8 may show a configuration including a semiconductor device 814 (e.g., GPU) whereas FIGS. 9A-9D may show a configuration that does not include semiconductor device 814. FIG. 9D may show a configuration that does not include a semiconductor device 818. FIG. 9E shows a single semiconductor device 814 (e.g., HBM) may be replaced with two semiconductor devices 814A and 814B (e.g., HBMs). A configuration may have fewer or additional semiconductor devices on each side of the interposer, or may have no semiconductor devices on a side of the interposer (e.g., FIG. 9D where semiconductor device 816 is also removed).
Although various figures in the present disclosure may show cold plates with channels directly in contact with or exposed to a backside of a semiconductor device (e.g., FIG. 8 shows cold plate 906 and semiconductor devices 804, 806, and 808), and cold plates with channels that are not in direct contact with or exposed to a backside of a semiconductor device (e.g., FIG. 8 shows cold plate 910 and semiconductor device 816, cold plate 908 and semiconductor device 818), any suitable configuration may be used in embodiments of the present disclosure.
Although various figures in the present disclosure may show a configuration of a single cold plate attached to multiple chips (e.g., cold plate 906 and semiconductor devices 804, 806, and 808), a configuration of a single chip having a single cold plate may be used. Although various figures in the present disclosure may show a configuration of a single chip having a single cold plate (e.g., cold plate 910 and semiconductor device 816, cold plate 908 and semiconductor device 818), a configuration having multiple chips attached to a single cold plate may be used.
Although various figures (e.g., FIGS. 9A-9G) in the present disclosure may show a single channel on a backside of the semiconductor device, there may be multiple channels on a backside of the device (e.g., as shown in FIG. 4, FIG. 6B).
FIG. 1 is a schematic side view of a device package 10 and a heat sink 22 attached to the device package 10. The device package 10 typically includes a package substrate 12, a first device 14, a device stack 15, a heat spreader 18, and first TIM layers 16A, 16B thermally coupling the first device 14 and the device stack 15 to the heat spreader 18. The device package 10 is thermally coupled to the heat sink 22 through a second TIM layer 20. The TIM layers 16A, 16B, 20 facilitate thermal contact between components in the device package 10 and between the device package 10 and the heat sink 22.
As heat flux density increases with increasing power density in advanced semiconductor devices, the cumulative thermal resistance of the system illustrated in FIG. 1 is increasingly problematic as heat cannot be dissipated quickly enough to allow semiconductor devices to run at optimal power. Consequently, the energy efficiency of semiconductor devices is reduced. Furthermore, heat is transferred between semiconductor devices within the device package 10, as shown with heat transfer path 24 (illustrated as a dashed line), where heat may be undesirably transferred from the first device 14 having a high heat flux, such as a central processing unit (CPU) or a graphical processing unit (GPU), to the device stack 15 having low heat flux, such as memory, through the heat spreader 18.
For example, as shown in FIG. 1, each device package component and the respective interfacial boundaries therebetween have a corresponding thermal resistance that forms heat transfer path 26 (illustrated by arrow 26 in FIG. 1). The right-hand side of FIG. 1 illustrates the heat transfer path 26 as a series of thermal resistances R1-R8 between a heat source and a heat sink. Here, R1 is the thermal resistance of the bulk semiconductor material of the first device 14. R3 and R7 are the thermal resistances of the first TIM layers 16A, 16B and the second TIM layer 20, respectively. R5 is the thermal resistance of the heat spreader 18. R2, R4, R6, and R8 represent the thermal resistance at the interfacial region of the components (e.g., contact resistances). In a typical cooling system, R3 and R7 may account for 80% or more of the cumulative thermal resistance of the heat transfer path 26, and R5 may account for 5% or more. R1 of the first device 14 and R2, R4, R6, and R8 of the interfaces account for the remaining cumulative thermal resistance. Accordingly, embodiments described herein provide for integrated cooling assemblies embedded within a device package. The embedded cooling assemblies shorten the thermal resistance path between a semiconductor device and a heat sink and reduce thermal communication between semiconductor devices disposed in the same device package, such as described in relation to the figures below.
FIG. 2A is a schematic plan view of an example of a system panel 100, in accordance with embodiments of the present disclosure. Generally, the system panel 100 includes a printed circuit board (PCB) 102, a plurality of device packages 201 mounted to the PCB 102, and a plurality of coolant lines 108 fluidly coupling each of the device packages 201 to a coolant source 110. It is contemplated that coolant fluid may be delivered to each of the device packages 201 in any desired fluid phase, e.g., liquid, vapor, gas, or combinations thereof, and may flow out from each device package 201 in the same phase or a different phase. In some embodiments, the coolant fluid is delivered to the device packages 201 and returned therefrom as a liquid, whereby the coolant source 110 may comprise a heat exchanger or chiller to maintain the coolant fluid at a desired temperature. In other embodiments, the coolant fluid may be delivered to the device packages 201 as a liquid, vaporized to a vapor within the device packages 201, and returned to the coolant source 110 as a vapor. In those embodiments, the device packages 201 may be fluidly coupled to the coolant source 110 in parallel, and the coolant source 110 may include or further include a compressor (not shown) for condensing the received vapor to a liquid form.
FIG. 2B is a schematic partial sectional side view of a portion of the system panel 100 of FIG. 2A. As shown, each device package 201 is fluidly coupled to the plurality of coolant lines 108 and is disposed in a socket 114 of the PCB 102 and connected thereto using a plurality of pins 116, or by other suitable connection methods, such as solder bumps (not shown). The device package 201 may be seated in the socket 114 and secured to the PCB 102 using a mounting frame 106 and a plurality of fasteners 112, e.g., compression screws, collectively configured to exert a relatively uniform downward force on the upward facing edges of the device package 201. The uniform downward force ensures proper pin contact between the device package 201 and the socket 114.
FIG. 2C is a schematic exploded isometric view of an example device package 201, in accordance with embodiments of the present disclosure. Generally, the device package 201 includes a package substrate 202, an integrated cooling assembly 203 disposed on the package substrate 202, and a package cover 208 disposed on a peripheral portion of the package substrate 202. Suitable materials that may be used in the package cover 208 include copper, aluminum, metal alloys, etc. The package cover 208 extends over the integrated cooling assembly 203 so that the integrated cooling assembly 203 is disposed between the package substrate 202 and the package cover 208. The integrated cooling assembly 203 typically includes a semiconductor device 204 and a cold plate 206 bonded to the semiconductor device 204. Although the lateral dimensions (or footprint) of the cold plate 206 are shown to be the same or similar to the lateral dimensions (or footprint) of the semiconductor device 204, the footprint of the cold plate 206 may be smaller or larger in one or both directions when compared to the footprint of the semiconductor device 204.
As shown, the device package 201 further includes a sealing material layer 222 that forms a coolant fluid impermeable barrier between the package cover 208 and the integrated cooling assembly 203 that prevents leaking of the coolant fluid outside of the cooling assembly and prevents coolant fluid from reaching an active side 218 (discussed below in relation to FIG. 3) of the semiconductor device 204 and causing damage thereto. In some embodiments, the sealing material layer 222 comprises an adhesive material that reliably attaches the package cover 208 to the integrated cooling assembly 203. In some embodiments, the sealing material layer 222 comprises a polymer or epoxy material that extends upwardly from the package substrate 202 to encapsulate and/or surround at least a portion of the semiconductor device 204. In some embodiments, the sealing material layer 222 may also comprise conductive material, e.g., solder. In other embodiments, the sealing material layer 222 is formed from a molding compound, e.g., a thermoset resin, that when polymerized, forms a hermetic seal between the package cover 208 and the cold plate 206. Here, the coolant fluid is delivered to the cold plate 206 through openings 222A disposed through the sealing material layer 222. As shown, the openings 222A are respectively in registration and fluid communication with inlet and outlet openings 212 of the package cover 208 thereabove and inlet and outlet openings 206A in the cold plate 206 therebelow.
It will be understood that the openings are shown in a section view. The openings may have any cross-sectional shape that allows fluid to flow therethrough (e.g., rectangular, square, hexagonal or circular cross-sections). For example, the inlet and outlet openings 206A of the cold plate 206 may form an elongated shape extending from one side of the cold plate 206 to another side of the cold plate 206. For example, the inlet and outlet openings 206A may form any shape having a length greater than a width in the X-Y plane (e.g., a rectangular or a trapezoidal shape). A shape in the X-Y plane of the openings 222A disposed through the sealing material layer 222 may be substantially the same as the shape of the inlet and outlet openings 206A of the cold plate 206 in the same place. Furthermore, it will be understood that references to an opening throughout the present disclosure refer to an opening defined by a sidewall (e.g., opening sidewall), unless otherwise indicates.
In some embodiments, gaps formed between the inside walls of the package cover 208 and the integrated cooling assembly 203 may be filled (partially or completely) with a molding material 223. The molding material 223 may encapsulate the integrates cooling assembly 203 to improve structural stability, for example.
The package substrate 202 can include a rigid material, such as an epoxy or resin-based laminate, that supports the integrated cooling assembly 203 and the package cover 208. The package substrate 202 may include conductive features disposed in or on the rigid material that electrically couples the integrated cooling assembly 203 to a system panel, such as the PCB 102.
FIG. 3 is a schematic sectional view in the X-Z plane of the device package 201 taken along line A-A′ of FIG. 2C. As illustrated in FIG. 3, the semiconductor device 204 includes the active side 218 that includes device components, e.g., transistors, resistors, and capacitors, formed thereon or therein, and a non-active side, here the semiconductor device backside 220, opposite the active side 218. As shown, the active side 218 is positioned adjacent to and facing towards the package substrate 202. The active side 218 may be electrically connected to the package substrate 202 by use of conductive bumps 219, which are encapsulated by a first underfill layer 221 disposed between the semiconductor device 204 and the package substrate 202. The first underfill layer 221 may comprise a cured polymer resin or epoxy, which provides mechanical support to the conductive bumps 219 and protects against thermal fatigue. In some embodiments, the active side 218 may be electrically connected to another package substrate, another active die, or another passive die (e.g., interposer) using hybrid bonding or conductive bumps 219. For example, the package substrate 202 may be a semiconductor substrate, a semiconductor interposer, a glass substrate, a glass interposer, a PCT or a combination thereof. The cold plate 206 may be disposed above the package substrate 202 with the semiconductor device 204 disposed therebetween. For example, the semiconductor device 204 (and the first underfill layer 221) may be disposed between the cold plate 206 and the package substrate 202. In some embodiments, the cold plate 206 may be disposed directly on the package substrate 202.
Here, the cold plate 206 comprises a top portion 234 and a sidewall 240 (e.g., a perimeter sidewall defining a perimeter of the cold plate 206) extending downwardly from the top portion 234 to the backside 220 of the semiconductor device 204. The top portion 234, the perimeter sidewall 240, and the backside 220 of the semiconductor device 204 collectively define a coolant channel 210 therebetween. The cold plate 206 comprises cavity dividers 230 extending downwardly from the top portion 234 towards the backside 220 of the semiconductor device 204. The cavity dividers 230 may alternatively be referred to as support features 230, which provide structural support to the integrated cooling assembly 203. The cavity dividers 230 may extend laterally and in parallel between an inlet opening 206A of the cold plate 206 and an outlet opening 206A of the cold plate 206 to define plural coolant channels 210 therebetween. It should be appreciated that, the cold plate 206 may comprise one cavity divider 230 which forms two coolant channels (e.g., one coolant channel on either side of the cavity divider 230) by means of the cavity divider 230 and portions of the perimeter sidewall 240. More specifically, coolant channels 210 may be formed between the cavity divider 230 and a portion of the perimeter sidewall 240 extending parallel to or in the same general direction as the cavity divider 230. Alternatively, in other embodiments, the cold plate 206 may comprise plural cavity dividers 230, for example two cavity dividers, five cavity dividers, or six cavity dividers (as illustrated in FIG. 4). In such examples, the cold plate 206 comprises more than two coolant channels 210, for example three coolant channels, four coolant channels, seven coolant channels, or more, defined between the cavity dividers 230 and/or the cavity divider(s) 230 and the perimeter sidewall 240. In some embodiments, at least one of the cavity dividers 230 may extend discontinuously between the inlet opening 206A and the outlet opening 206A (in the X-axis direction) to form a discontinuous cavity divider. A discontinuous cavity divider may be formed of plural segments between which coolant fluid may flow. The segments of a discontinuous cavity divider may have the same or different lengths in the X-axis direction. One or more segments may form a post.
The cavity dividers 230 comprise cavity sidewalls 232 which form surfaces of corresponding coolant channels 210. In embodiments where plural cavity dividers 230 extend in parallel to each other, cavity sidewalls 232 of adjacent cavity dividers 230 are opposite (e.g., facing) each other. In embodiments comprising a single cavity divider 230, a first cavity sidewall may be opposite (e.g., face) a first portion of the perimeter sidewall 240 extending parallel to and facing the first cavity sidewall. A second cavity sidewall may be opposite (e.g., face) a second portion of the perimeter sidewall 240 extending parallel to and facing the second cavity sidewall. The first portion of the perimeter sidewall 240 may be an opposite side of the cold plate 206 to the second portion of the perimeter sidewall 240. For example, in embodiments where the cold plate 206 is rectangular, first and second opposing sides of the rectangular cold plate 206 form the first and second portions of the perimeter sidewall 240.
The cavity dividers 230 may be continuous cavity dividers which extend continuously (e.g., in the X-axis direction) between the inlet opening 206A and the outlet opening 206A of the cold plate 206.
With reference to FIG. 3, coolant channels 210 may be defined by:
the backside 220 of the semiconductor device 204, which forms lower coolant channel surfaces;
portions of the perimeter sidewall 240 extending in the Y-axis direction, which form end surfaces of the coolant channels 210;
the cavity sidewalls 232, which form inner surfaces of the coolant channels 210 in the X-axis direction; and
portions of the perimeter sidewall 240 extending in the X-axis direction, which form outer surfaces of the coolant channels 210 in the X-axis direction.
As shown in FIG. 4 and described in further detail below, the cavity sidewalls 232 can be formed at an acute angle with respect to the backside 220 of the semiconductor device 204 such that upper portions of opposing (e.g., facing) cavity sidewalls 232 meet. Therefore, the cavity sidewalls 232 and the backside 220 of the semiconductor device 204 collectively define a triangular cross-section of the coolant channel 210. However, it will be understood that the coolant channel 210 may be formed with different shaped cross-sections. For example, one or more coolant channels may be formed with trapezoidal, rectangular, or semi-circular cross-section, or a combination thereof.
In some embodiments, the backside 220 of the semiconductor device 204 comprises a corrosion protective layer (not shown). The corrosion protective layer may be a continuous layer disposed across the entire backside 220 of the semiconductor device 204, such that the cold plate 206 is attached thereto. Beneficially, the corrosion protective layer provides a corrosion-resistant barrier layer, thus preventing undesired corrosion of the semiconductor device 204 (e.g., the semiconductor substrate material which might otherwise be in direct contact with coolant fluid flowing through a coolant chamber volume 210).
One or more coolant chamber volumes may include one or more coolant channels. The coolant channels may extend between a single inlet opening and a single outlet opening of the cold plate 206, such that the coolant chamber volume(s) and/or coolant channel(s) share the same inlet and outlet openings. In other embodiments, multiple inlet and/or outlet openings may be coupled to the coolant chamber volume(s).
In embodiments having plural coolant chamber volumes and/or plural coolant channels, each coolant chamber volume and/or coolant channel may be connected between a separate inlet opening and a separate outlet opening. In such embodiments, the coolant fluid may be directed to the separate inlet openings and from the separate outlet openings using a manifold disposed above the openings in the Z-axis direction. In some embodiments, a gasket may be used to seal a gap between the manifold and the cold plate inlet/outlet openings. The gasket may be made of rubber (e.g., neoprene, nitrile, ethylene propylene diene monomer, or silicon rubber) or similar such material. For example, the gasket may be an o-ring. The gasket may be attached between a lower surface of the manifold and an upper surface of the cold plate facing the manifold using an adhesive. The gasket may provide a water tight seal to direct coolant fluid from the manifold into the cold plate inlet/outlet openings while preventing coolant fluid from leaking onto exterior surfaces of the integrated cooling assembly 203. In some embodiments, the manifold is attached to one or more cold plates using one or more corresponding gaskets.
Referring to FIG. 4, a height h in the Z-axis direction of the coolant chamber volume(s) and or coolant channel(s) may be greater than 100 μm, 100 μm-1000 μm, or 100 μm-700 μm. A width w in the Y-axis direction of each coolant channel 210 may be greater than 100 μm, 100 μm-1000 μm, or 100 μm-700 μm. For example, the width of each coolant channel 210 may be greater than the height h thereof. In some embodiments, the width w of a coolant channel 210 may, at the widest portion, which may be taken as a base of the triangular shape of the coolant chamber channels 210 shown in FIG. 4, range from 0.2 mm to 5 mm. More specifically, the width w of a coolant channel 210 may range from 0.5 to 1.5 mm. The width w of the coolant chamber volumes(s) and/or coolant channel(s) may also be between 1 and 5 mm.
A cross-section of the coolant chamber volume(s) and/or coolant channel(s) in the Y-Z plane is wide enough to allow for a pressure drop of 0-20 psi, 3-15 psi, or 4-10 psi.
In some embodiments, preparing a desired surface roughness of the sidewalls of the coolant chamber volume(s) and/or coolant channels may include depositing an organic layer on a photoresist layer after cold plate features have been etched to form a micro-masking layer, such as between 1 to 30 nm. The micro-masking layer may be dry etched to form the desired surface roughness, such as between 0.1 to 3.0 nm. Advantageously, providing sidewalls with surface roughness increases the likelihood of fluid being directed towards and contacting the backside 220 of the semiconductor device 204 (e.g., by disrupting a hydrodynamic boundary layer of fluid between the sidewall and the coolant fluid).
With reference to FIG. 3, the cold plate 206 is attached to the backside 220 of the device 204 without the use of an intervening adhesive. For example, the cold plate 206 may be directly bonded to the backside 220 of the device 204, such that the cold plate 206 and the backside 220 of the device 204 are in direct contact. For example, in some embodiments, one or both of the cold plate 206 and the backside 220 of the semiconductor device 204 may comprise a dielectric material layer, e.g., a first dielectric material layer 224A and a second dielectric material layer 224B respectively, and the cold plate 206 is directly bonded to the backside 220 of the semiconductor device 204 through bonds formed between the dielectric material layers 224A, 224B. In some embodiments, one of the cold plate 206 or the backside 220 of the semiconductor device 204 may comprise a thin bonding dielectric layer (e.g., silicon nitride, etc.) and other element(s) may not include any such explicit bonding dielectric layer (or can have only a native oxide layer). The first and second dielectric material layers 224A, 224B may be continuous or non-continuous. For example, the first dielectric material layer 224A may be disposed only on lower surfaces of the cold plate 206 facing the backside 220 of the semiconductor device 204. With reference to FIG. 4, described below, portions of the first dielectric material layer 224A may be disposed only on lower surfaces of the cavity dividers 230 (e.g. support features 230) and the perimeter sidewall 240. Beneficially, directly bonding the cold plate 206 to the semiconductor device 204, as described above, reduces the thermal resistance therebetween and increases the efficiency of heat transfer from the semiconductor device 204 to the cold plate 206. In particular, thermal resistance is reduced by directly bonding lower surfaces of the cavity dividers 230 facing the semiconductor device 204 to the backside 220 of the semiconductor device 204.
FIG. 4 is a schematic sectional view in the Y-Z plane of the integrated cooling assembly 203. In FIG. 4, the cold plate 206 comprises a patterned side that faces towards the semiconductor device 204 and an opposite side that faces towards the package cover 208 (not shown). The patterned side comprises a coolant chamber volume having plural coolant channels 210, which extend laterally (along the X-axis direction in FIG. 4) between the inlet and outlet openings of the cold plate 206. Each coolant channel 210 comprises cavity sidewalls that define a corresponding coolant channel 210. Portions of the cold plate 206 between the cavity sidewalls 232 form the support features 230 (e.g., cavity dividers 230). The support features 230 (e.g., cavity dividers 230) provide structural support to the integrated cooling assembly 203 and disrupt laminar fluid flow (e.g., due to surface roughness of the sidewalls) at the interface of the coolant and the device backside 220, resulting in increased heat transfer therebetween. Furthermore, by introducing plural coolant channels 210 to define separate coolant flow paths, an internal surface area of the cold plate 206 is increased, which further increases the efficiency of heat transfer.
In FIG. 4, arrows 228A and 228B illustrate two different heat transfer paths in the integrated cooling assembly 203. A first heat transfer path illustrated by arrow 228A shows heat generated by the semiconductor device 204 transferring directly from the semiconductor material of the semiconductor device 204 to coolant fluid flowing through the cold plate 206. A second heat transfer path illustrated by arrows 228B shows heat generated by the semiconductor device 204 being transferred from semiconductor material (e.g., silicon material) of the semiconductor device 204 to semiconductor material (e.g., silicon material) of the cold plate 206 structure, propagated throughout the semiconductor material of the cold plate 206 structure (shown as dashed lines), and being transferring into coolant fluid flowing through the cold plate 206. A thermal resistance of the first and second heat transfer paths 228A, 228B is illustrated by heat transfer path 228C, which is shown as thermal resistance R1 between a heat source and a cold plate. Here, R1 is the thermal resistance of the bulk semiconductor material of the semiconductor device 204. It can be seen that the heat transfer path 228C of the integrated cooling assembly 203 is reduced compared to the heat transfer path 26 of the device package 10 of FIG. 1, due to the direct bonding discussed above.
In some embodiments, the cold plate 206 may be attached to the semiconductor device 204 using a hybrid bonding technique, where bonds are formed between the dielectric material layers 224A, 224B and between metal features, such as between first metal pads and second metal pads, disposed in the dielectric material layers 224A, 224B. Advantageously, by using hybrid bonding techniques, interconnections may be formed between the cold plate 206 and the semiconductor device 204 using the first and second metal pads.
Suitable dielectrics that may be used as the dielectric material layers 224A, 224B include silicon oxides, silicon nitrides, silicon oxynitrides, silicon carbon nitrides, metal-oxides, metal-nitrides, silicon carbide, silicon oxycarbides, silicon oxycarbonitride, diamond-like carbon (DLC), or combinations thereof. In some embodiments, one or both of the dielectric material layers 224A, 224B are formed of an inorganic dielectric material, e.g., a dielectric material substantially free of organic polymers. Typically, one or both of the dielectric layers are deposited to a thickness greater than the thickness of a native oxide, such as about 1 nanometer (nm) or more, 5 nm or more, 10 nm or more, 50 nm or more, or 100 nm or more. In some embodiments, one or both of the layers are deposited to a thickness of 3 micrometers or less, 1 micrometer or less, 500 nm or less, such as 100 nm or less, or 50 nm or less. The dielectric layer material and thickness may be optimized for lower thermal resistance between the semiconductor device and the cold plate.
The cold plate 206 may be formed of any suitable material that has sufficient structural strength to withstand the desired pressures of coolant flowing into the coolant chamber volume 210. For example, the cold plate 206 may be formed of semiconductor material like silicon or other materials like glass. In other examples, the cold plate 206 may be formed of a material selected from a group comprising polymers, metals, ceramics, or composites thereof. In some embodiments, the cold plate 206 may be formed of stainless steel (e.g., from a stainless steel metal sheet) or a sapphire plate.
In some embodiments, the cold plate 206 may be formed of a bulk material having a substantially similar CTE to the bulk material of the substrate 202 and/or the semiconductor device 204, where the CTE is a fractional change in length of the material (in the X-Y plane) per degree of temperature change. In some embodiments, the CTEs of the cold plate 206, the substrate 202, and/or the semiconductor device 204 are matched so that the CTE of the substrate 202 and/or the semiconductor device 204 is within about +/−20% or less of the CTE of the cold plate 206, such as within +/−15% or less, within +/−10% or less, or within about +/−5% or less when measured across a desired temperature range. In some embodiments, the CTEs are matched across a temperature range from about −60° C. to about 100° C. or from about −60° C. to about 175° C. In one example embodiment, the matched CTE materials each include silicon.
In some embodiments, the cold plate 206 may be formed of a material having a substantially different CTE from the semiconductor device 204, e.g., a CTE mismatched material. In such embodiments, the cold plate 206 may be attached to the semiconductor device 204 by a compliant adhesive layer (not shown) or a molding material that absorbs the difference in expansion between the cold plate 206 and the semiconductor device 204 across repeated thermal cycles.
The package cover 208 shown in FIGS. 2C and 3 generally comprises one or more vertical or sloped sidewall portions 208A and a lateral portion 208B that spans and connects the sidewall portions 208A. The sidewall portions 208A may extend upwardly from a peripheral surface of the package substrate 202 to surround the device 204 and the cold plate 206 disposed thereon. The lateral portion 208B may be disposed over the cold plate 206 and is typically spaced apart from the cold plate 206 by a gap corresponding to the thickness of the sealing material layer 222. The sealing material may be an adhesive or a gasket. In some embodiments, instead of or as well as the sealing material layer 222, a gasket may be used to seal a gap between the package cover 208 and the cold plate inlet/outlet openings. The gasket may be made of rubber (e.g., neoprene, nitrile, ethylene propylene diene monomer, or silicon rubber) or similar such material. For example, the gasket may be an o-ring. The gasket may be attached between a lower surface of the package cover 208 and an upper surface of the cold plate facing the package cover 208 using an adhesive. The gasket may provide a water tight seal to direct coolant fluid from the package cover 208 into the cold plate inlet/outlet openings while preventing coolant fluid from leaking onto exterior surfaces of the integrated cooling assembly 203. In some embodiments, the package cover 208 is attached to one or more cold plates using one or more corresponding gaskets.
Coolant is circulated through the coolant chamber volume 210 through the inlet and outlet openings 212 of the package cover 208 formed through the lateral portion 208B. The inlet and outlet openings 206A of the cold plate 206 may be in fluid communication with the inlet and outlet openings 212 of the package cover 208 through the inlet and outlet openings 222A formed in the sealing material layer 222 disposed therebetween. In certain embodiments, coolant lines 108 (FIGS. 2A-2B) may be attached to the device package 201 by use of connector features formed in the package cover 208, such as threads formed in the sidewalls of the inlet and outlet openings 212 of the package cover 208 and/or protruding features 214 that surround the inlet and outlet openings 212 and extend upwardly from a surface of the lateral portion 208B.
Typically, the package cover 208 is formed of semi-rigid or rigid material so that at least a portion of the downward force exerted on the package cover 208 by the mounting frame is transferred to a supporting surface of the package substrate 202 and not transferred to the cold plate 206 and the semiconductor device 204 therebelow. In some embodiments, the package cover 208 is formed of a thermally conductive metal, such as aluminum or copper. In such embodiments, the package cover 208 functions as a heat spreader that redistributes heat from one or more electronic components of the semiconductor device 204. In some embodiments, the package cover 208 and/or a manifold (such as the manifold discussed above) may consist of or comprise a thermally insulating material or materials. In such embodiments, the package cover 208 and/or the manifold may function as a thermal insulator to retain heat or cold. In some embodiments, the package cover 208 and/or the manifold may be insulating to minimize or reduce the flow of thermal energy (e.g., thermal flux) between components (e.g., semiconductor devices, semiconductor device stacks, device packages, etc.). For example, the package cover 208 and/or the manifold may minimize or reduce the flow of thermal energy between a first semiconductor device and a second semiconductor device. In another example, the package cover 208 and/or the manifold may minimize or reduce the flow of thermal energy between a first semiconductor device stack and a second semiconductor device stack. In another example, the package cover 208 and/or the manifold may minimize or reduce the flow of thermal energy between a first device package and a second device package. In another example, the package cover 208 and/or the manifold may minimize or reduce the flow of thermal energy between a semiconductor device and a semiconductor device stack. In another example, the package cover 208 and/or the manifold may minimize or reduce the flow of thermal energy between a semiconductor device of a device package and a second device package.
It should be noted that the direction in which the coolant fluid flows through the cold plate 206 may be controlled depending on the relative locations of the inlet and outlet openings. For example, the coolant fluid may flow from left to right in the device package 201 of FIG. 3 when the inlet openings 212, 222A, 206A of the package cover 208, the sealing material layer 222, and the cold plate 206, respectively, are located on the left-hand side of the device package 201 and the outlet openings 212, 222A, 206A of the package cover 208, the sealing material layer 222, and the cold plate 206, respectively, are located on the right-hand side of the device package 201. Alternatively, the coolant fluid may flow from right to left in the device package 201 illustrated in FIG. 3 when the outlet openings 212, 222A, 206A of the package cover 208, the sealing material layer 222, and the cold plate 206 are located on the left-hand side of the device package 201 and the inlet openings 212, 222A, 206A of the package cover 208, the sealing material layer 222, and the cold plate 206 are located on the right-hand side of the device package 201. Although only one set of inlet and outlet openings is shown and described here, additional inlet and outlet openings may also be provided at various locations on the package cover 208, the sealing material layer 222, and the cold plate 206.
An example flow path of the coolant fluid through the coolant chamber volume 210 may be as follows:
1. Coolant fluid enters the coolant chamber volume 210 through the inlet openings 212, 222A, and 206A.
2. Coolant fluid flows across the inside surfaces of the cold plate 206 and absorbs heat generated by the semiconductor device 204, which has dissipated into the cold plate 206 structure. The coolant fluid may also flow directly across the backside 220 of the semiconductor device 204 to absorb heat energy directly from the semiconductor device 204. The coolant chamber volume 210 may additionally have various channels formed to direct the coolant fluid flow from inlet opening(s) to outlet opening(s) and facilitate heat extraction from the semiconductor device 204 by the coolant fluid. In some embodiments, the coolant fluid may be in direct contact with the backside 220 of the semiconductor device 204 or via one or more substrate or layers between the coolant fluid or backside 220 of the semiconductor device 204.
3. Coolant fluid exits the coolant chamber volume 210 through outlet openings 212, 222A, and 206A.
It will be understood from the above flow path that heat is extracted without introducing an unnecessary thermal resistance (e.g., a TIM disposed between the backside 220 of the semiconductor device 204 and the cold plate 206) between the backside 220 of the semiconductor device 204 and the cold plate 206.
FIG. 5 is a schematic side sectional view in the X-Z plane of an example of a multi-component device package 501 that includes a cold plate 506 directly bonded to the backside surfaces of two or more devices 501A, 501B. The multi-component device package 501 may be similar to the device package 201 described above, and therefore the description of similar features is omitted for brevity. In some embodiments, the two or more devices 501A and 501B are singulated from reconstituted wafers and then bonded to the cold plate 506. As shown, the device package 501 includes a package substrate 502, an integrated cooling assembly 503 and a package cover 508. The integrated cooling assembly 503 may include a plurality of devices 501A (one shown) that may be singulated and/or disposed in a vertical device stack 501B (one shown). The cold plate 506 may be attached to each of the devices 501A and device stack 501B, e.g., by the direct bonding methods described herein or other methods including adhesive. In some embodiments, the device 501A may comprise a processor, and the device stack 501B may comprise a plurality of memory devices, such as a high bandwidth memory (HBM) comprising a bottom logic die and a plurality of memory (e.g., DRAM) dies stacked on the logic die. Here, the device 501A and the device stack 501B are disposed in a side-by-side arrangement on the package substrate 502 and are in electrical communication with one another through conductive elements formed in, on, or through the package substrate 502. Here, the cold plate 506 is sized to provide a bonding surface for attachment to both the device 501A and the device stack 501B but may otherwise be the same or substantially similar to other cold plates described herein. In some embodiments, the lateral dimensions (or footprint) of the cold plate 506 may be smaller or larger than the combined lateral dimensions (or footprint) of both the device 501A and the device stack 501B. In some embodiments, one or more sidewalls of the cold plate 506 may be aligned or offset to the vertical sidewalls of the device 501A and the device stack 501B (including inside or outside their footprint). In some embodiments, more than one cold plate 506 may be bonded. For example, separate cold plates may be bonded to the device 501A and the device stack 501B.
FIG. 6A is a schematic sectional view in the X-Z plane of an integrated cooling assembly 603 and FIG. 6B is a schematic sectional view in the Y-Z plane of the integrated cooling assembly 603, in accordance with embodiments of the present disclosure. The integrated cooling assembly 603 may be similar to the integrated cooling assembly 203 described above, and therefore the description of similar features is omitted for brevity.
In FIG. 6A, a width of a cold plate 606 in a first direction is greater than a width of a semiconductor device 604 the first direction. The first direction may be taken to be a direction perpendicular to a second direction in which perimeter sidewall extends. With reference to FIGS. 6A and 6B, the second direction is the Z-axis direction and the first direction is either the X-axis or the Y-axis direction. As shown, the width of the cold plate 606 is greater than the width of the semiconductor device 604 in both the X-axis direction and the Y-axis direction. In embodiments of FIGS. 6A and 6B where the semiconductor device 604 has a rectangular footprint, the cold plate 606 may extend beyond all four sidewalls of the semiconductor device 604. However, it will be understood that the width of the cold plate 606 may be greater than the width of the semiconductor device 604 in either the X-axis direction or the Y-axis direction.
In order to provide a cold plate 606 having a width greater than a width of the semiconductor device 604, a structural substrate 600 having substantially the same width (in the X-axis direction and/or the Y-axis direction) as the cold plate 606 is provided between the cold plate 606 and the semiconductor device 604. The structural substrate 600 provides structural rigidity to overhanging portions of the cold plate 606 and also closes portions of coolant channels 610 in the overhanging portions which would otherwise be exposed. The structural substrate 600 may be attached between the cold plate 606 and the semiconductor device 604 using direct bonding techniques described herein.
Advantageously, by increasing the width of the cold plate 606 in the X-axis direction and/or the Y-axis direction, as described above, additional coolant channels 610 may be introduced to the cold plate 606 in order to increase the efficiency of thermal cooling.
In some embodiments, only portions of coolant channels 610 in the overhanging portions may be closed by the structural substrate 600, while portions of coolant channels 610 vertically adjacent to the semiconductor device 604 may be exposed to a backside of the semiconductor device 604. For example, portions of coolant channels 610 vertically adjacent to the semiconductor device 604 may be exposed by etching openings in horizontally aligned portions of the structural substrate 600.
FIG. 7 is a flow diagram showing a method 70 of forming an integrated cooling assembly, according to embodiments of the present disclosure. Generally, the method 70 includes bonding a first substrate comprising one or more cold plates 206 to a second substrate comprising one or more semiconductor devices 204, and singulating one or more integrated cooling assemblies 203 from the bonded first and second substrates. For example, a wafer (bare or reconstituted wafer) comprising one or more cold plates 206 can be directly bonded to another wafer (bare or reconstituted wafer) comprising one or more semiconductor devices 204.
It will be understood that the first substrate may be a cold plate die or part of a wafer of cold plates. Further, the second substrate may be a semiconductor device die or part of a wafer of semiconductor devices 204. Therefore, the method 70 may include die-to-die direct bonding (e.g., cold plate die to semiconductor device die), wafer-to-die direct bonding (e.g., cold plate die to semiconductor device wafer, or cold plate wafer to semiconductor device die), and wafer-to-wafer direct bonding (e.g., cold plate wafer to semiconductor device wafer). It will be understood that the singulation step (discussed in relation to block 74, below) may not be required for a die-to-die direct bonding operation.
For simplicity, the following description is focused on forming one integrated cooling assembly 203 comprising one cold plate 206 and one semiconductor device 204. However, as mentioned above, in some embodiments, the first substrate may comprise plural cold plates 206 and the second substrate may comprise plural semiconductor devices 204, such that plural integrated cooling assemblies 203 may be formed from the first and second substrates.
At block 72, the method 70 includes directly bonding the first substrate (e.g., a monocrystalline silicon wafer) comprising a cold plate 206 to the second substrate (e.g., a monocrystalline silicon wafer) comprising a semiconductor device 204. By direct bonding, it is meant that the bond is effected without an intervening adhesive.
In some embodiments, the first substrate may be etched using a patterned mask layer formed on its surface to form features of the cold plate 206. An anisotropic etch process may be used, which uses inherently differing etch rates for the silicon material as between {100} plane surfaces and {111} plane surfaces when exposed to an anisotropic etchant.
In some embodiments, the etching process is controlled to where a ratio of the etch rate in the {100} plane to the etch rate in the {111} plane is between about 1:10 and about 1:200, such as between about 1:10 and about 1:100, for example between about 1:10 and 1:50, or between about 1:25 and 1:75. Examples of suitable anisotropic wet etchants include aqueous solutions of potassium hydroxide (KOH), ethylene diamine and pyrocatechol (EPD), ammonium hydroxide (HN4OH), hydrazine (N2H4), or tetra methyl ammonium hydroxide (TMAH). The actual etch rates of the silicon substrate depend on the concentration of the etchant in the aqueous solution, the temperature of the aqueous solution, and a concentration of the dopant in the substrate (if any). Typically, the mask layer is formed of a material that is selective to anisotropic etch compared to the underlying monocrystalline silicon substrate. Examples of suitable mask materials include silicon oxide (SixOy) or silicon nitride (SixNy). In some embodiments, the mask layer has a thickness of about 100 nm or less, such as about 50 nm or less, or about 30 nm or less. The mask layer may be patterned using any suitable combination of lithography and material etching patterning methods.
The second substrate may include a bulk material, and a plurality of material layers disposed on the bulk material. The bulk material may include any semiconductor material suitable for manufacturing semiconductor devices, such as silicon, silicon carbide, silicon germanium, germanium, group III-V semiconductor materials, group II-VI semiconductor materials, or combinations thereof. While some high-performance processors like CPUs, GPUs, neural processing units (NPUs), and tensor processing units (TPUs) are typically made out of silicon, some other high power density (hence substantial heat-generating) devices may comprise silicon carbide or gallium nitride, for example. In some embodiments, the second substrate may include a monocrystalline wafer, such as a silicon wafer, a plurality of device components formed in or on the silicon wafer, and a plurality of interconnect layers formed over the plurality of device components. In other embodiments, the second substrate may comprise a reconstituted substrate, e.g., a substrate formed from a plurality of singulated devices embedded in a support material. In some embodiments, each semiconductor device may have its own individual cold plate fabricated through a reconstitution process.
The bulk material of the second substrate may be thinned after the semiconductor device 204 is formed using one or more backgrinding, etching, and polishing operations that remove material from the backside. Thinning the second substrate may include using a combination of grinding and etching processes to reduce the thickness (in the Z-direction) to about 450 μm or less, such as about 200 μm or less, or about 150 μm or less or about 50 μm or less. After thinning, the backside 220 may be polished to a desired smoothness using a chemical mechanical polishing (CMP) process, and the dielectric material layer may be deposited thereon. In some embodiments, the dielectric material layer may be polished to a desired smoothness to prepare the second substrate for the bonding process. In some embodiments, the method 70 includes forming a plurality of metal features in the dielectric material layer in preparation for a hybrid bonding process, such as by use of a damascene process.
In some embodiments, the active side of the second substrate is temporarily bonded to a carrier substrate (not shown) before the thinning process. When used, the carrier substrate provides support for the thinning operation and/or for the thinned material to facilitate substrate handling during one or more of the subsequent manufacturing operations described herein.
Here, the method 70 may include forming dielectric layers on one or both the first and second substrates, and directly bonding includes forming dielectric bonds between a first dielectric material layer of the first substrate and a second dielectric material layer of the second substrate (or forming direct dielectric bonds between one substrate (which may have a native oxide layer at its bonding surface) and a dielectric material layer of the other substrate). Direct bonding processes join dielectric layers by forming strong chemical bonds (e.g., covalent bonds) between the dielectric layers.
Generally, directly bonding the surfaces (of the dielectric material layers formed on the first and second substrates) includes preparing, aligning, and contacting the surfaces. Examples of dielectric material layers include silicon oxide, silicon nitride, silicon oxynitride, and silicon carbonitride. Preparing the surfaces may include smoothing the respective surfaces to a desired surface roughness, such as between 0.1 to 3.0 nm RMS, activating the surfaces (e.g., a “very slight etch” using plasma or wet chemical treatment as taught in U.S. Pat. No. 6,902,987) to weaken or open chemical bonds in the dielectric material, and terminating the surfaces with a desired species (e.g., also as described in U.S. Pat. No. 6,902,987). Smoothing the surfaces may include polishing the first and second substrates using a CMP process. Simultaneously activating and terminating the surfaces with a desired species may include exposing the surfaces to radical species formed in a plasma. The bond interface between the bonded dielectric layers can include a higher concentration of materials from the activation and/or last chemical treatment processes compared to the bulk of the bonding layers. For example, in some embodiments that utilize a nitrogen plasma for activation that terminates the bonding surface with a nitrogen-containing species, a nitrogen concentration peak can be formed at the bond interface. In some embodiments, the nitrogen concentration peak may be detectable using secondary ion mass spectroscopy (SIMS) techniques. In various embodiments, for example, a nitrogen termination treatment (e.g., exposing the bonding surface to a nitrogen-containing plasma) can replace OH groups of a hydrolyzed (OH-terminated) surface with NH2 molecules, yielding a nitrogen-terminated surface. In embodiments that utilize an oxygen plasma for activation, an oxygen concentration peak can be formed at the bond interface between non-conductive bonding surfaces. Such an oxygen concentration peak will be more detectable when the bonding layers do not contain oxygen, such as layers containing silicon nitride or silicon carbon nitride.
In some embodiments, the plasma is formed using a nitrogen-containing gas, e.g., N2, and the terminating species includes nitrogen, or nitrogen and hydrogen. In some embodiments, fluorine may also be present within the plasma. In some embodiments, the surfaces may be activated using a wet cleaning or etching process, e.g., by exposing the surfaces to an aqueous ammonia solution (e.g., ammonium hydroxide). In some embodiments, the dielectric bonds may be formed using a dielectric material layer deposited on only one of the first and second substrates, but not on both. In those embodiments, the direct dielectric bonds may be formed by contacting the deposited dielectric material layer of one of the first and second substrates directly with a bulk material surface (or such a surface with a native oxide) of the other substrate.
Directly forming direct dielectric bonds between the first and second substrates at block 72 may include bringing the prepared and aligned surfaces into direct contact at a temperature less than 150° C., such as less than 100° C., for example, less than 30° C., or about room temperature, e.g., between 20° C. and 30° C. Without intending to be bound by theory, in the case of directly bonding surfaces terminated with nitrogen and hydrogen (e.g., NH2 groups), it is believed that a chemical bond is formed in part from the nitrogen species, wherein hydrogen gas byproducts (H2 gas) of the chemical reaction diffuse away from the interfacial bonding surfaces. In some embodiments, the direct bond is strengthened using an anneal process, where the substrates are heated to and maintained at a temperature of greater than about 30° C. and less than about 450° C., for example, greater than about 50° C. and less than about 250° C., or about 150° C., for a duration of about 5 minutes or more, such as about 15 minutes. Typically, the bonds will strengthen over time even without the application of heat. Thus, in some embodiments, the method does not include heating the substrates.
In embodiments where the first and second substrates are bonded using hybrid dielectric and metal bonds, the method 70 may further include planarizing or recessing the metal features below the dielectric field surface before contacting and bonding the dielectric material layers. After the dielectric bonds are formed, the first and second substrates may be heated to a temperature of 150° C. or more and maintained at the elevated temperature for a duration of about 1 hour or more, such as between 8 and 24 hours, to form direct metallurgical bonds between the metal features.
Suitable direct dielectric and hybrid bonding technologies that may be used to perform aspects of the methods described herein include ZiBond® and DBI®, each of which are commercially available from Adeia Holding Corp., San Jose, CA, USA.
At block 74, the method 70 includes singulating at least one integrated cooling assembly 203 from the bonded first and second substrates. Singulation after bonding may impart distinctive structural characteristics on the integrated cooling assembly 203 as the cold plate 206 has the same perimeter as the semiconductor device 204 bonded thereto. Thus, the sidewalls (e.g., side surfaces) of the cold plate 206 are typically flush with the sidewalls (e.g., side surfaces) of the semiconductor device 204 about their common perimeters. In some embodiments, the cold plate 206 is singulated from the first substrate using a process that cuts or divides the first substrate in a vertical plane, i.e., in the Z-direction. In those embodiments, the side surfaces of the cold plate 206 are substantially perpendicular to the backside 220 of the semiconductor device 204, i.e., a horizontal (X-Y) plane of an attachment interface between the semiconductor device 204 and the cold plate 206. In some embodiments, the cold plate 206 is singulated using a saw or laser dicing process.
At block 76, the method 70 may include connecting the integrated cooling assembly 203 to the package substrate 202 and sealing a package cover 208 comprising inlet and outlet openings 212 to the integrated cooling assembly 203 by use of a sealing material layer 222, such as a molding compound that is cured.
At block 78, the method 70 may include, before or after sealing the package cover 208 to the integrated cooling assembly 203, forming inlet and outlet openings 222A in the sealing material layer 222 to fluidly connect the inlet and outlet openings 212 of the package cover 208 to the cold plate 206.
FIG. 8 schematically illustrates an integrated cooling assembly 800 with two-sided cooling, according to some embodiments. The integrated cooling assembly 800 has semiconductor devices (e.g., semiconductor devices 804, 806, 808, 814, 816, and 818) on both sides (e.g., a first side 803 and a second side 813) of an interposer 802. The semiconductor devices are cooled with cold plates (e.g., cold plate 820 and cold plate 830) on each side of the interposer 802. By placing the six semiconductor devices in a stacked configuration (e.g., three semiconductor devices 804, 806, and 808 on a top side or first side 803 of the interposer 802, three semiconductor devices 814, 816, and 818 on a bottom side or second side 813 of the interposer 802), a communication distance between the semiconductor devices (e.g., GPUs and the HBMs) can be reduced compared to a configuration where all six semiconductor devices are in a same plane on top (e.g., on a top side 803) of the interposer 802.
The interposer 802 has a first side 803 and a second side 813. The first side is opposite the second side 813 of the interposer 802 (e.g., a top side and bottom side of the interposer 802). In some embodiments, the interposer 802 may be a passive device comprising interconnects and vias that reroute connections to the semiconductor devices. The interposer 802 may communicate with a PCB board (e.g., substrate 918 as shown in FIGS. 9A-9G). After communication signals reach the interposer from a PCB board, the communication distances between the semiconductor devices (e.g., GPUs and HBMs) can be shorter through the interposer 802 than compared to a configuration where all the semiconductor devices are on one side of the interposer 802.
First semiconductor devices 804, 806, and 808 are attached to the first side 803 of the interposer 802. Second semiconductor devices 814, 816, and 818 are attached to the second side 813 of the interposer 802. The first semiconductor devices 804, 806, 808 and second semiconductor devices 814, 816, 818 may be directly bonded or hybrid bonded to the interposer 802.
The semiconductor devices 804 and 814 may be GPUs, and the semiconductor devices 806, 808, 816, and 818 may be HBMs or stacked DRAMs. A top or bottom GPU (e.g., semiconductor device 804, semiconductor device 814) may communicate with all HBMs on top and bottom of the interposer (e.g., semiconductor devices 806, 808, 816 and 818). A top GPU (e.g., semiconductor device 804) may communicate with a bottom GPU (e.g., semiconductor device 814). In some embodiments, the GPUs (e.g., semiconductor devices 804 and 814) on different sides of an interposer 802 may communicate with each other through vias (e.g., thru-silicon vias or TSVs) in the interposer 802. The GPUs (e.g., semiconductor devices 804 and 814) may communicate with HBMs on a same side or an opposite side of the interposer 802 through vias (e.g., TSVs) or interconnects in the interposer 802. The interposer 802 may be connected to one or more substrates (e.g., as shown in FIGS. 9A-9G).
A first cold plate 820 is attached to the first semiconductor devices 804, 806, and 808. A second cold plate 830 is attached to the second semiconductor devices 814, 816, and 818. The first cold plate 820 may be directly bonded or hybrid bonded to the to the first semiconductor devices 804, 806, and 808. The second cold plate 830 may be directly bonded or hybrid bonded to the second semiconductor devices 814, 816, and 818.
Although FIG. 8 shows semiconductor devices 804 and 814 as GPUs, and semiconductor devices 806, 808, 816, and 818 as HBMs (e.g., stacked DRAM), the semiconductor devices 804, 806, 808, 814, 816, and 818 may be any suitable type of semiconductor device in any suitable type of order or arrangement. In some embodiments, there may be only one GPU (e.g., semiconductor device 804) and the bottom semiconductor devices may all be HBMs (e.g., semiconductor devices 814, 816, and 818), and the GPU (e.g., semiconductor device 804) may communicate with all the HBMs (e.g., 806, 808, 814, 816, and 818). In some embodiments, semiconductor devices 816 and 818 may be XPUs and there may be no semiconductor device 814. In some embodiments, there may not be second semiconductor devices 814, 816, and 818, and a second cold plate 830 may be directly bonded to the interposer 802.
FIGS. 9A-9G schematically illustrates examples of system panels with two-sided cooling, according to some embodiments. In FIGS. 9A-9D, the cross section shown may include a cross section along a width of a channel. For example, coolant fluid may flow from an inlet (e.g., arrows pointing downwards into the openings or inlets of the cold plate 906 in FIGS. 9A-9D) and into the channel (e.g., represented by “x” in the channel, indicating flow into the page in FIGS. 9A-9D). The coolant fluid may flow through the channel, and exit at openings or outlets of the cold plate 906 (e.g., similar to what is shown in FIGS. 9A-9D, except arrows pointing upwards out of the openings or outlets of the cold plate 906 to indicate coolant fluid exiting the cold plate 906). In FIGS. 9F-9G, the cross section shown may include a cross section along a length of a channel. For example, coolant fluid may flow from an inlet (e.g., top left opening) of cold plate 954, through the channel (e.g., top channel), and exit out the outlet (e.g., top right opening) of the cold plate 954. Coolant fluid may flow from an inlet (e.g., top left opening) of the cold plate 954, to an outlet (e.g., bottom left opening) of the cold plate 954, to an inlet (e.g., top left opening) of the cold plate 956, through the channel (e.g., bottom channel), and exit out an outlet (e.g., top right opening) of the cold plate 956, to an inlet (e.g., bottom right opening) of the cold plate 954, to an outlet (top right opening) of cold plate 954.
FIG. 9A shows an integrated cooling assembly 900 embedded or disposed in a first substrate 914 (e.g., package substrate) attached to a second substrate 918 (e.g., PCB). The integrated cooling assembly 900 comprises an interposer 902, first semiconductor devices 804, 806, 808 on a first side (e.g., top side) of the interposer 902, second semiconductor devices 816 and 818 on a second side (e.g., opposite side, bottom side) of the interposer 902, a first cold plate 906 directly bonded to backsides of the first semiconductor devices 804, 806, 808, and second cold plates 908, 910 directly bonded to backsides of second semiconductor devices 816, 818, respectively. Cooling channels are in a top portion and a bottom portion of the integrated cooling assembly 900 (e.g., both sides of the interposer 902). A top coolant channel (e.g., coolant channel in a first cold plate 906) can be connected to a bottom coolant channel (e.g., coolant channel in a second cold plate 908, 910) through openings in the interposer 902. The interposer 902 may be attached to and electrically connected to the first substrate 914 through conductive bumps 912 (e.g., microbump connections, solder balls, etc.). The first substrate 914 may be attached to and electrically connected to the second substrate 918 through conductive bumps 916 (e.g., microbump connections, solder balls, etc.). The conductive bumps 912 may be about 300 microns, less than about 500 microns, 200-400 microns, or 100-500 microns. Although a specific number of conductive bumps may be shown throughout various figures of embodiments of the present disclosure, any suitable number of conductive bumps may be used.
The coolant channel at the bottom of the integrated cooling assembly 900 may be embedded in the first substrate 914. The first substrate 914 may comprise a dielectric material. The first substrate 914 may be a multi-layer substrate or a package substrate. A thickness of the first substrate 914 may be about 1 mm, 1-2 mm, or 1-5 mm. The first substrate 914 may have a first side 913 and a second side 915 opposite the first side 913. Openings may be disposed or etched in a first side 913 of the first substrate 914. A depth of the opening may be about 300 microns, 500 microns, 300-500 microns, greater than about 300 microns, less than 500 microns.
The integrated cooling assembly 900 may comprise the semiconductor devices 804, 806, 808, 816, and 818 attached to the interposer 902 and the cold plates 906, 908, and 910 attached to the respective semiconductor devices. A method may include forming the integrated cooling assembly 900 by attaching (e.g., directly bonding, hybrid bonding) the semiconductor devices 804, 806, 808, 816, and 818 to the interposer 902, and attaching (e.g., directly bonding, hybrid bonding) the cold plates 906, 908, 910 to the respective semiconductor devices. The method may further include attaching the integrated cooling assembly 900 to the first substrate 914, and attaching the first substrate 914 to the second substrate 918. Surface mount technology (SMT) or soldering may be used to attach the integrated cooling assembly 900 to the first substrate 914 or to attach the first substrate 914 to the second substrate 918.
The second cold plates 908, 910 may be hanging in the opening of the first substrate 914 as shown in FIG. 9A. For example, an air gap may separate sides of the second cold plates 908, 910 and the first substrate 914. In some embodiments, the opening of the first substrate 914 may be underfilled or an adhesive may be used to attach the cold plate 908 or 910 in the opening of the first substrate 914.
A spacer 905 may be between the cold plate 906 and the cold plate 908. The spacer 905 may comprise adhesive. The spacer 905 may comprise a spacer material 905a and adhesive 905b. For example, the spacer 905 may comprise a gasket (e.g., o-ring) and adhesive on top and bottom of the gasket to couple the gasket to the cold plates 906 and 908.
A spacer 904 may be on both sides of the interposer 902. The spacer 904 may be between the interposer 902 and the cold plate 906. The spacer 904 may have a thickness similar to the thickness of the semiconductor device 806 or 816. In some embodiments, the spacer 904 may comprise a spacer material and adhesive (e.g., similar to what is shown for spacer 905 as spacer material 905a and adhesive 905b). In some embodiments, the spacer 904 may comprise adhesive.
In some embodiments, the semiconductor devices 806 and 808 may be HBMs, semiconductor device 804 may be a GPU. The semiconductor devices 816 and 818 may be HBMs. In some embodiments, the semiconductor devices 816 and 818 may be XPUs. In some embodiments, the semiconductor devices 816 and 818 may be dummy silicon or dummy semiconductor device (e.g., used to improve cooling). These configurations of semiconductor devices or any suitable configuration may be applied to any of those described throughout the present disclosure.
In FIG. 9A, on the right side of the integrated cooling assembly 900, coolant fluid may flow into an inlet opening in a first cold plate 906, through a channel (e.g., along an x-axis, flow into the page, top side of interposer 902, backside of semiconductor device 808), and exit up out of an outlet opening in the cold plate 906 (e.g., on another side of the cold plate). Coolant fluid may flow into an inlet opening in a first cold plate 906, down to an inlet opening in a second cold plate 908, through a channel (e.g., along an x-axis, flow into the page, bottom side of interposer 902), up through an outlet opening in the second cold plate 908, up to the first cold plate 906, and exit up out of an outlet opening in the cold plate 906 (e.g., on another side of the cold plate 906). This flow may be similar to the flow in integrated cooling assemblies 920 and 930 of FIGS. 9B and 9C. The flow for integrated cooling assembly 940 of FIG. 9D may be similar, except a second cold plate 928 is used instead of a second cold plate 908 of FIGS. 9A-9C and the cold plate 928 is directly attached to the interposer.
The flow from cold plate 906 to cold plate 908 can be made through spacer 905 (e.g., spacer material 905a and adhesive 905b as shown in FIG. 9A) and or through an interposer (e.g., through interposer 922 and spacers 904 as shown in FIGS. 9B and 9C). A flow from cold plate 908 to cold plate 906 can be made through spacer 905 (e.g., similar to spacer material 905a and adhesive 905b as shown in FIG. 9A) and through an interposer (e.g., similar to interposer 922 and spacers 904 as shown in FIGS. 9B and 9C).
In a middle of the integrated cooling assembly 900, coolant fluid may flow into an inlet opening in the first cold plate 906, through a channel (e.g., along an x-axis, flow into the page, top side of interposer 902, backside of semiconductor device 804), and exit up out of an outlet opening in the cold plate 906 (e.g., on another side of the cold plate). This flow may be similar to integrated cooling assemblies of FIGS. 9B-9D.
On the left side of the integrated cooling assembly 900, coolant fluid may flow into an inlet opening in a first cold plate 906, through a channel (e.g., along an x-axis, flow into the page, top side of interposer 902, backside of semiconductor device 806), and exit up out of an outlet opening in the cold plate 906 (e.g., on another side of the cold plate). Coolant fluid may flow into an inlet opening in a first cold plate 906, down to an inlet opening in a second cold plate 908, through a channel (e.g., along an x-axis, flow into the page, bottom side of interposer 902), up through an outlet opening in the second cold plate 908, up to the first cold plate 906, and exit up out of an outlet opening in the cold plate 906 (e.g., on another side of the cold plate 906). This flow may be similar to integrated cooling assemblies of FIGS. 9B-9D. The flow from cold plate 906 to cold plate 910 can be made through the interposer (e.g., interposer 902, 922, 932) and spacers 904 (e.g., adhesive) as shown in FIGS. 9A-9D. A flow from cold plate 910 to cold plate 906 can be made through the interposer (e.g., interposer 902, 922, 932) and spacers 904 (e.g., adhesive) similar to those shown in FIGS. 9A-9D.
FIG. 9B shows an integrated cooling assembly 920 with an interposer 922 similar to the interposer 902 of FIG. 9A, except interposer 922 is wider. The wider interposer 922 enables more connections from the interposer to the first substrate 914. The interposer 922 is attached and electrically connected to the first substrate 914 in a central portion and a peripheral portion of the interposer 922. FIG. 9B shows electrical connections being made through conductive bumps 912 in a central portion of the interposer 922 to the first substrate 914 and through conductive bumps 924 in a peripheral portion of the interposer 922 (e.g., outside a region corresponding to a footprint of the semiconductor devices 804, 806, and 808). The wider interposer 922 has openings through the interposer 922 connecting first coolant channels the first cold plate 906 to the second coolant channels in the second cold plates 908 and 910, respectively, through the interposer 922.
FIG. 9C shows electrical connections can be brought from the top of the integrated cooling assembly 930 through the semiconductor device 804. For example, electrical connections (e.g., power and ground) can be brought from a top cooling side (e.g., through first cold plate 906, in between coolant channels may be interconnects, I/O from a backside of the semiconductor device 804) and transferred to interposer 932 through TSVs 934 (e.g., pass through power TSVs) in the semiconductor device 804. Although four TSVs 934 are shown in FIG. 9C, any suitable number of TSVs may be used. In some embodiments, electrical connections (e.g., power and ground signal) can enter through the cooling side (e.g., cold plate 906) instead of through a PCB side (e.g., from second substrate 918). A GPU (e.g., semiconductor device 804) may be redesigned to accommodate for more TSVs.
FIG. 9D shows an integrated cooling assembly 940, in which the bottom cooling may be used for cooling the first semiconductor devices 804, 806, and 808. In some embodiments, second semiconductor device 816 may be a dummy semiconductor device or dummy chip (e.g., silicon), and a cold plate 910 (e.g., second cold plate) is attached to the dummy chip. In some embodiments, there may be no second semiconductor device 818, and a cold plate 928 (e.g., third cold plate) may be attached (e.g., directly bonded or hybrid bonded) to the interposer 942.
FIG. 9E shows an integrated cooling assembly 950 with a cold plate (e.g., first cold plate 954, second cold plate 956) attached to multiple semiconductor devices on each side of an interposer 952. The first semiconductor devices 804, 806, 808 are disposed on a central portion of the first side 951 of the interposer 952. The interposer 952 is attached to a first substrate 957 (e.g., shim) at the periphery of a second side 953 of the interposer 952 and connected via conductive bumps 955. The first substrate 957 may be two separate substrates placed on at least two sides of an interposer 952. The first substrate 957 may be a frame (e.g., four sides with a center portion removed where the second cold plate 956 may fit). The first substrate 957 may have three sides (e.g., corresponding to three edges of the interposer with a center portion removed). The I/O connections to the interposer 952 may be made on the sides (e.g., right and left sides, three sides, four sides, periphery, etc.) of the interposer 952. The first substrate 957 is connected to a second substrate 959 via conductive bumps 958. The first substrate 957 may be used to meet the height limitation. For example, the conductive bumps 955 may not be sufficient in height, and the first substrate 957 may be used to add height to the second substrate 959 for the cold plate 956 (e.g., second cold plate). A benefit of having a long channel (e.g., cooling across multiple semiconductor devices 816, 814, and 818) is that a second GPU may be placed at the bottom of the interposer (e.g., semiconductor device 814). There may be fewer I/O connections for this type of configuration. The configuration in FIG. 9E does not include a central portion with connections made through a first substrate to a second substrate (e.g., PCB). The configuration in FIG. 9E shows connections made through the edge portion of the interposer 952 through the conductive bumps 955, first substrate 957, conductive bumps 958, and the second substrate 959.
In some embodiments, a GPU (e.g., semiconductor device 804) may be surrounded by HBMs. For example, instead of another GPU (e.g., semiconductor device 814) on a bottom portion of the interposer 952 there may be multiple HBMs (e.g., semiconductor devices 814A and 814B).
FIGS. 9F-9G show variations to the configuration of FIG. 9E in which an interposer 952 is placed directly on a substrate 962 without an intervening substrate 957 (e.g., shim). FIG. 9F shows an integrated cooling assembly 960 with a substrate 962 that is etched. Substrate 962 may have an opening for the cold plate 956. The configuration in FIG. 9F shows connections made through the edge portion of the interposer 952 through the conductive bumps 964, to the substrate 962. FIG. 9G shows an integrated cooling assembly 970 with an interposer 952 across two substrates 972 and 974. In some embodiments, the substrate may be a deeply etched substrate (e.g., substrate 972 and substrate 974 may be of a same substrate, similar to the substrate 962 of FIG. 9F except with a deeper opening so that interposer 952 may act as a bridge). The interposer 952 is attached and electrically connected to substrate 972 and 974 through conductive bumps 976.
FIGS. 10A-10C show example methods that can be used to manufacture an integrated cooling assembly with two sided cooling, according to some embodiments. FIG. 10A may show a method 1000 for manufacturing an integrated cooling assembly with two sided cooling. FIG. 10B may show a method 1010 used to couple a coolant channel in a first cold plate to a second cold plate through an interposer. FIG. 10C may show a method 1020 for coupling a coolant channel in a first cold plate to a second cold plate outside of an interposer.
FIG. 10A shows a method 1000 for manufacturing an integrated cooling assembly with two sided cooling, in which a cold plate may be directly bonded a semiconductor device attached to an interposer or directly bonded to an interposer. At block 1001, the method 1000 includes providing an interposer having a first side and a second side opposite the first side. In some embodiments, providing an interposer includes forming holes (e.g., openings) through the interposer. For example, FIGS. 9A-9G show interposers 902, 922, 932, 942, 952 may include one or more holes. In some embodiments, the interposer may not have holes through the interposer. For example, in FIG. 9A, a channel between top cold plate 906 and bottom cold plate 908 is formed outside of an interposer 902 of FIG. 9A. In some embodiments, the interposer 902 of FIG. 9A may not include a hole (e.g., the left side of FIG. 9A may be similar to the right side of FIG. 9A and channel between top cold plate 906 and bottom cold plate 910 may be formed outside of an interposer 902).
In some embodiments, providing an interposer may comprise attaching (e.g., directly bonding, hybrid bonding, soldering, etc.) semiconductor devices to one or more sides of the interposer. For example, a plurality of semiconductor devices (e.g., semiconductor devices 804, 806, and 808) may be attached to a first side of the interposer (e.g., interposer 942 of FIG. 9D). In some embodiments, a plurality of first semiconductor devices (e.g., semiconductor devices 804, 806, and 808) may be attached to a first side of the interposer, and a plurality of second semiconductor devices (e.g., semiconductor devices 814, 816, and 818) may be attached to a second side of the interposer. There may be at least one semiconductor device attached to a second side of the interposer (e.g., semiconductor device 816 is attached to the interposer 942 of FIG. 9D). In some embodiments, there may be no semiconductor devices attached to a second side of the interposer (e.g., similar to embodiment FIG. 9D on right side, where there is no semiconductor device between interposer 942 and cold plate 928 and where the left side of FIG. 9D may be similar to the right side of FIG. 9D, with no semiconductor device 816 between the interposer 942 and cold plate 910).
At block 1002, the method 1000 includes directly bonding a first cold plate to at least one semiconductor device attached to the first side of an interposer. In some embodiments, the first cold plate is directly bonded to a plurality of semiconductor devices attached to the first side of the interposer. For example, the cold plate 906 may be directly bonded to a plurality of semiconductor devices 804, 806, and 808 in FIGS. 9A-9D. For example, first cold plate 954 may be directly bonded to a plurality of semiconductor devices 804, 806, and 808 in FIGS. 9E-9G. In some embodiments, the first cold plate may be hybrid bonded to a plurality of semiconductor devices attached to the first side of the interposer.
At block 1003, the method 1000 includes attaching at least one second cold plate to the second side of the interposer. In some embodiments, attaching at least one second cold plate to the second side of the interposer comprises directly bonding a second cold plate to the second side of the interposer. For example, the cold plate 928 is directly bonded to the second side of the interposer 942 in FIG. 9D. In some embodiments, a plurality of second cold plates may be bonded to a second side of an interposer (e.g., similar to embodiment FIG. 9D, where there is no semiconductor device 816 between the interposer 942 and cold plate 910, and a cold plate similar to or the same as cold plate 910 is directly bonded to the second side of the interposer 942). In some embodiments, one or more second cold plates may be hybrid bonded to the second side of the interposer 942.
In some embodiments, attaching at least one second cold plate to a second side of the interposer comprises directly bonding a second cold plate to a second semiconductor device. In some embodiments, a plurality of first semiconductor devices (e.g., semiconductor devices 804, 806, and 808) may be attached to a first side of the interposer, and a plurality of second semiconductor devices (e.g., semiconductor devices 814, 816, and 818) may be attached to a second side of the interposer. There may be at least one semiconductor device attached to a second side of the interposer (e.g., two semiconductor devices 816, 818 are attached to the interposer 902, 922, 932 of FIGS. 9A-9C, one semiconductor device 816 is attached to the interposer 942 of FIG. 9D, etc.). In some embodiments, a second cold plate may be attached to a second semiconductor device (e.g., FIG. 9D, cold plate 910 may be directly bonded to second semiconductor device 816). In some embodiments a plurality of second cold plates may be directly bonded to a plurality of second semiconductor devices (e.g., cold plate 910, 908 may be directly bonded to second semiconductor device 816, 818 respectively). The method may include directly bonding at least one second cold plate to at least one second semiconductor device. In some embodiments, a second cold plate may be directly bonded to a plurality of second semiconductor devices (e.g., second cold plate 956 may be directly bonded to second semiconductor devices 814, 816, and 818 of FIGS. 9E-9G). In some embodiments, one or more second cold plates may be hybrid bonded to the one or more second semiconductor devices.
At block 1004, the method 1000 includes providing a substrate. A substrate may comprise a package substrate, a substrate with one or more openings, a frame, or a shim. Providing a substrate may comprise forming one or more openings in a substrate (e.g., substrate 914 has openings for cold plates 908, 910 or cold plates 928, 910 in FIGS. 9A-9D, substrate 962 has an opening for cold plate 956 in FIG. 9F, etc.). In some embodiments, the openings in the substrate may be formed using any suitable technique to remove material from a substrate or pattern a substrate (e.g., etching, wet etch, dry etch, etc.). In some embodiments, providing a substrate may comprise providing a substrate with an opening through the substrate (e.g., a frame or substrate 957 of FIG. 9E). In some embodiments, two or more substrates may be provided (e.g., shims or substrates 957 of FIG. 9E, substrates 972 and 974 of FIG. 9G).
At block 1005, the method 1000 includes attaching the interposer to the substrate. The interposer may be attached to the substrate using conductive bumps or solder bumps (e.g., interposer 902, 922, 923, 942 attached to substrate 914 using conductive bumps 912 in FIGS. 9A-9D, interposer 952 attached to substrate 962 using conductive bumps 964 in FIG. 9F). The interposer may be attached one or more substrates (e.g., frame, shims, two or more substrates, etc.) using conductive bumps (e.g., interposer 952 attached to one or more substrates 957 using conductive bumps 955 in FIG. 9E). The interposer may be attached to two or more substrates using conductive bumps (e.g., interposer 952 attached to substrates 972 and 974 using conductive bumps 976 in FIG. 9G). In some embodiments, the method includes attaching the substrate to a PCB. In some embodiments, the substrate is attached to a PCB (e.g., first substrate 914 attached to second substrate 918 or PCB using conductive bumps 916 in FIGS. 9A-9D, one or more first substrates 957 is attached to second substrate 959 using conductive bumps 958 in FIG. 9E).
FIG. 10B may show a method 1010 used to couple a coolant channel in a first cold plate to a second cold plate through an interposer. At block 1011, the method 1010 includes forming a hole (e.g., opening) through the interposer. The hole or opening may be formed using any suitable technique (e.g., etching, wet etch, dry etch, etc.). In some embodiments, block 1011 may be included in block 1001 of FIG. 10A. At block 1012, the method 1010 includes coupling at least one coolant channel in the first cold plate to at least one coolant channel of a second cold plate using a first spacer between the first cold plate and the interposer and using a second spacer between the second cold plate and the interposer. For example, spacers 904 are shown in FIGS. 9A-9D to connect a coolant channel in the first cold plate 906 to a coolant channel in the second cold plate 910, or spacers 904 are used to connect a coolant channel in cold plate 954 to cold plate 956 in FIGS. 9E-9G. In some embodiments, a first spacer may be placed between the first cold plate and the interposer prior to the first cold plate being directly bonded to at least one semiconductor device (e.g., block 1002 of FIG. 10A) and a second spacer may be placed between the second cold plate and the interposer prior to attaching at least one second cold plate to the second side of the interposer (e.g., block 1003 of FIG. 10A).
FIG. 10C may show a method 1020 for coupling a coolant channel in a first cold plate to a coolant channel in a second cold plate outside of an interposer. At block 1021, the method may include providing a spacer material. The spacer material may comprise adhesive (e.g., spacer 905 of FIG. 9A). The spacer material may comprise a gasket (e.g., o-ring) and adhesive may be used to couple the gasket to the cold plates (e.g., spacer material 905a and adhesive 905b of FIG. 9A). At block 1022, the method may include attaching the first cold plate to the second cold plate using adhesive between the spacer material and portions of the first cold plate and the second cold plate. For example, in FIG. 9A, spacer material 905a (e.g., material used in a gasket or o-ring such as neoprene, nitrile, ethylene propylene diene monomer, or silicon rubber, etc.) is used to couple a coolant channel in a first cold plate 906 to a second cold plate 908. The spacer material 905a may be attached to the first cold plate 906 using adhesive 905b, and the spacer material 905a may be attached to the second cold plate 908 using adhesive 905b.
The embodiments discussed above are intended to be illustrative and not limiting. One skilled in the art would appreciate that individual aspects of the cooling assemblies, device packages, and methods discussed herein may be omitted, modified, combined, and/or rearranged without departing from the scope of the disclosure.Source: ipg260505.zip (2026-05-05)