NVIDIA GTC Keynote

  • Thread starter Deleted member 162929
  • Start date
1679418087967.png

1679418248645.png
 
Not sure how exaggerated the claim is, but :
Nvidia claims its new approach enables 500 DGX H100 systems wielding 4,000 Hopper GPUs to do the same amount of work as 40,000 CPU-based servers, but do so 40X faster and with 9X less power. Nvidia claims this reduces the computational lithography workload to produce a photomask from several weeks down to eight hours.

That starting to sound like some form of acceleration of the loop, GPU used to make better GPU faster, those faster made better GPUs than used to make better GPUs faster.

Maybe it would remove what look like still a lot of guesswork and surprise in the process, with just testing something tomorrow morning instead of not because it would have took more than 2 weeks and a small fortune (seem like it could cost $330,000 just from raw computer power to do it at $0.1khw in the past, without counting cooling them and all that come with it)
 
Last edited:
You must log in or register to reply here.
Back
Top