How the efficiency of supercomputers has changed over the past decade
Source: Koomey Analytics, in collaboration with AMD
Jonathan Koomey*, Zachary Schmidt*, and Samuel Naffziger†
* Koomey Analytics; †Advanced Micro Devices, Inc.
A new paper analyzes data from the industry site Top 500 to evaluate the efficiency of supercomputers over the past decade. It also compares how the efficiency and performance of a recently announced supercomputer, scheduled to be completed in 2021, compares to a simple extrapolation of those historical trends.
Technical paper link
Experts at the Table: Extensible instruction-set architecture is drawing attention from across the industry and supply chain.
It’s time to move up in abstraction again as a complexity overwhelms a key approach.
Hybrid bonding opens up a big improvement in die-to-die performance, but getting there is not trivial.
The goal is to use fab-like methods in the lab, but that’s not easy.
Semiconductor aging has moved from being a foundry issue to a user problem. As we get to 5nm and below, vectorless methodologies become too inaccurate.
Efforts underway to develop 7nm, DRAM, 3D NAND, and EUV domestically as trade war escalates.
Lots of new technologies, problems and uncertainty as device scaling continues.
Manufacturing chips at future nodes is possible from a technology standpoint, but that’s not the only consideration.
Why future nodes will require new equipment and approaches.
New approaches aim for better performance, more flexibility — and for some, lower cost.
Programmability shifts some of the burden from hardware engineers to software developers for ML applications.
Manufacturing issues grow as cameras become more sophisticated.
Leave a Reply