摘要:
A method and an apparatus for cooling, preferably within an enclosure, a diversity of heat-generating components, with at least some of the components having high-power densities and others having low-power densities. Heat generated by the essentially relatively few high-power-density components, such as microprocessor chips for example, is removed by direct liquid cooling, whereas heat generated by the more numerous low-power or low-watt-density components, such as memory chips for example, is removed by liquid-assisted air cooling in the form of a closed loop comprising a plurality of heating and cooling zones that alternate along the air path.
摘要:
Systems and methods for providing memory modules with multiple hub devices. Exemplary systems include a cascade-interconnect memory system with a memory bus, a memory controller and a memory module. The memory controller is in communication with the memory bus for generating, receiving and responding to memory access requests. The memory module includes a first hub device with three or more ports and a second hub device with three or more ports. A first port on the first hub device is in communication with the memory controller via the memory bus, a second port on the first hub device is in communication with a first set of memory devices, and a third port on the first hub device is cascade connected to a first port on the second hub device. A second port on the second hub device is in communication with a second set of memory devices and a third port on the second hub device supports a cascaded connection to a subsequent hub device in the memory system.
摘要:
A high density circuit package includes a pair of planar packages, the planar packages exhibiting front and back surfaces and positioned back-to-back in the high density circuit package. Each planar package includes a flexible circuit carrier having a plurality of circuit chips mounted thereon. Front and back planar metallic heat sinks sandwich the circuit carriers, at least one of the heat sinks contacting a surface of the chips mounted on the sandwiched circuit carriers. Each heat sink is provided with air flow apertures formed in its planar surface and adjacent to each circuit chip. A circuit card interconnects with the circuit carriers in an interconnection region and is pluggable into a female connector. The planar metallic heat sinks and circuit carriers are mechanically packaged so as to provide a planar arrangement which aligns the apertures in both the front and rear heat sinks. A pair of planar packages are mechanically connected in a back-to-back arrangement so that the apertures therebetween are aligned. The associated circuit cards are also back-to-back oriented so as to enable their joint interconnection into the female connector.
摘要:
Wiring topology and hierarchy of transmission line impedances for connecting I/O of semiconductor devices together. This arrangement gives smooth signal shapes for signal rise and fall times as fast as one (1) nsec or faster. The design uses a balanced multi-way branched net with increasing impedance until very close to the end, of the net, where it is a balanced multi-way branched net with unterminated ends. Thus, coming out of the signal driver is a single impedance transmission line. This single impedance transmission line (A) then branches into two impedance transmission lines (B), each having an impedance higher than the single line impedance that feeds it. These lines (B) are used to drive electronic modules. After entering the electronic modules through a connector, each of these lines (B) then branch into two transmission lines (C) having a yet higher impedance value. Each of these lines (C) finally ends in a cluster of four transmission lines (D. Each of these final lines (D) connects to electronic devices on the electronic module.
摘要:
This invention relates to three dimensional packaging of integrated circuit chips into stacks to form cuboid structures. Between adjacent chips in the stack, there is disposed an electrical interconnection means which is a first substrate having a plurality of conductors one end of which is electrically connected to chip contact locations and the other end of which extends to one side of the chip stack to form a plurality of pin-like electrical interconnection assemblies. The pin-like structures can be formed from projections of the first substrate having an electrical conductor on at least one side thereof extending from this side. Alternatively, the pin-like structures can be formed from conductors which cantilever from both sides of an edge of the first substrate and within which corresponding conductors from both sides are aligned and spaced apart by the first substrate thickness. The spaces contain solder and form solder loaded pin-like structures. The pin-like structures can be directly solder bonded to conductors on a second substrate surface or the pin-like structures can be adapted for insertion into apertures in a second substrate. The second substrate provides a means for electrically inter-connecting a plurality of these cuboids. Preferably, the first and second substrates are circuitized flexible polymeric films. The second substrate is disposed on a third substrate, such as a PC board, with a resilient material therebetween which permits a heat sink to be pressed into intimate contact with an opposite side of the cuboid structures.
摘要:
An interconnecting structure in which circuit and componentry bearing cards are retained in contact with an intermediate card that has component and circuitry areas, and socket insertion contacts.
摘要:
A method is described for aligning a contact pattern on an electronic device held by a first movable support, with a bond site pattern on a lead frame held by a second movable support. The method includes the steps of: (a) creating and storing models of a chip's contact pattern and a lead frame's bond site pattern; (b) imaging the electronic device and lead frame; (c) determining the position of contacts on said electronic device and reorienting the contact pattern model to a best fit with the imaged contact position; (d) determining the position of each bond site on the imaged lead frame and reorienting the bond site model to a best fit with the imaged bond site position; (e) determining positional differences between the reoriented lead frame and contact pattern models; and (f) generating signals to reorient the first and second movable supports to minimize the positional differences when they are moved into a bonding positon. A machine is described for performing the above method wherein each of three main movable elements of the machine is assigned a dimensional axis in which its travel is non-adjustable, so that other movable elements can be calibrated thereagainst.
摘要:
A Multi-Petascale Highly Efficient Parallel Supercomputer of 100 petaOPS-scale computing, at decreased cost, power and footprint, and that allows for a maximum packaging density of processing nodes from an interconnect point of view. The Supercomputer exploits technological advances in VLSI that enables a computing model where many processors can be integrated into a single Application Specific Integrated Circuit (ASIC). Each ASIC computing node comprises a system-on-chip ASIC utilizing four or more processors integrated into one die, with each having full access to all system resources and enabling adaptive partitioning of the processors to functions such as compute or messaging I/O on an application by application basis, and preferably, enable adaptive partitioning of functions in accordance with various algorithmic phases within an application, or if I/O or other processors are underutilized, then can participate in computation or communication nodes are interconnected by a five dimensional torus network with DMA that optimally maximize the throughput of packet communications between nodes and minimize latency.
摘要翻译:具有100 petaOPS规模计算的多Petascale高效并行超级计算机,其成本,功耗和占地面积都在降低,并且允许从互连角度来看处理节点的最大封装密度。 超级计算机利用了VLSI的技术进步,实现了许多处理器可以集成到单个专用集成电路(ASIC)中的计算模型。 每个ASIC计算节点包括利用集成到一个管芯中的四个或更多个处理器的片上系统ASIC,每个处理器具有对所有系统资源的完全访问,并且使得处理器能够对诸如计算或消息传递I / O 并且优选地,根据应用内的各种算法阶段实现功能的自适应分割,或者如果I / O或其他处理器未被充分利用,则可以参与计算或通信节点通过五维环面网络互连 使用DMA来最大限度地最大化节点之间的分组通信的吞吐量并最小化等待时间。
摘要:
According to one embodiment of the present invention, a system for network communication includes an M dimensional grid of node groups, each node group including N nodes, wherein M is greater than or equal to one and N is greater than one and each node comprises a router and intra-group links directly connecting each node in each node group to every other node in the node group. In addition, the system includes inter-group links directly connecting each node in each node group to a node in each neighboring node group in the M dimensional grid.
摘要:
According to one embodiment of the present invention, a system for network communication includes an M dimensional grid of node groups, each node group including N nodes, wherein M is greater than or equal to one and N is greater than one and each node comprises a router and intra-group links directly connecting each node in each node group to every other node in the node group. In addition, the system includes inter-group links directly connecting each node in each node group to a node in each neighboring node group in the M dimensional grid.