Abstract

As system performance requirements for high performance computing (HPC) systems become more demanding, the need to increase component packaging density to shorten interconnect distances becomes more stringent. One technique for accomplishing this requirement is to implement 3-dimensional heterogeneous integration of system components. In an earlier publication, we described the design of a processor-memory module for a high performance computing (HPC) application space using a 3D integration (3DI) approach [1]. The design was based on interconnection and power delivery requirements for a processor-memory module capable of supporting 64 full-duplex 30 Gb/second SerDes, routing for 800 processor-to-memory pins, an integrated multi-tiered power delivery network, and a thermal management solution capable of dissipating a nominal processor heat flux of 100 W/cm2. Using thermal test chips (TTC), we designed and assembled a 3D processor-memory module with an integrated power delivery network to investigate interconnect density, integration, testability, and rework issues with 3D integrated packaging in an HPC environment. The technologies selected—semi-rigid flex, power connectors, land grid array (LGA) attachment with an anisotropic film, and cold plate-based cooling—are all commercially available, which were adapted for the test module. We were able to fabricate and conduct thermal testing of this design. This paper includes an overview of our HPC 3DI thermal test vehicle (3DI TTV) design, and compares test results between measured and simulated temperatures for the TTCs used to emulate both the memory and the processor. Unexpected differences were observed between the measured and simulated results at a corner location on the TTC. After ruling out device and test equipment issues, we discovered a silicon defect that, although it could not be modeled using our standard computational fluid dynamics (CFD) methods, appeared to explain the measured results. A rudimentary finite element analysis (FEA) analysis agreed more closely with the measured results, indicating the need for awareness of possible limitations with assumptions used in our CFD analysis.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.