google-deepmind/alphatensor

Clock Speed Set to 1530MHz when V100 has a max boost clock of 1380 MHz

afzalxo opened this issue · 2 comments

In your GPU benchmark, you set the persistence mode to ON and then lock the GPU clocks to 1530,1530 as follows:

process = subprocess.Popen(
      'sudo nvidia-smi --lock-gpu-clocks=1530,1530'.split(' '),
      stdout=subprocess.PIPE)

How does that work with V100 GPU with a base clock of 1245 MHz and boost clock of 1380 MHz? To my understanding V100S has a higher clock, but GCP offers V100, not V100S.

Credit to Nigel Tufnel. Dials set to 11.

I wonder what would happen if you cranked it up to full 100?
Would Google cloud GPU combust?