Session Details |
|
|
Name: |
|
Processor Technologies for HPC & AI - Convergence or Divergence? |
|
Time: |
|
Tuesday, June 20, 2017 11:00 am - 12:30 pm |
|
|
Room: |
|
Panorama 1 Messe Frankfurt |
|
|
Breaks: | 12:30 pm - 01:45 pm Lunch |
|
Chair: |
|
Satoshi Matsuoka, Tokyo Institute of Technology |
|
|
Abstract: |
|
The third wave of “rebirth” of AI is driven by significant advances in computing capabilities that HPC has taken a central part in progression over the years. With many orders of magnitude increase in expensive algorithms such as deep neural networks have now become pragmatic whereas it was unthinkable 20 years ago even with supercomputers of that era. As such AI has leveraged HPC, and given the sustained importance to the society. Some speculate that HPC will in turn leverage AI for its existence, just as it has leveraged other IT industries such as PCs, web, and gaming, leading to convergence of AI and HPC. At the same time, however, some claim that characteristics of processing that is particular to AI may cause divergence with HPC, differing in emphasis compared to traditional simulation workloads such as PDEs. This session invites three leading HPC processor vendors, NVIDIA, Intel and Fujitsu, who have also been openly active in developing processors that are suited to AI/Machine learning workloads, and will have them present their views on whether mainstream HPC today and AI will converge, or diverge in the future. |
|
|
Presentations: |
|
Dawn of Smart Supercomputing: How HPC & AI Are Engaging in a Virtuous Cycle That Will Leave Both Forever Transformed & Immeasurably Improved 11:00 am - 11:30 am |
|
|
Steve Oberlin, NVIDIA |
|
| |
|
|
Achieving AI Scale with HPC 11:30 am - 12:00 pm |
|
|
Pradeep Dubey, Intel |
|
| |
|
|
Fujitsu HPC & AI Processors 12:00 pm - 12:30 pm |
|
|
Takumi Maruyama, Fujitsu |
|
|