Tesla P100 F16 results surprisingly low

timo.mustamaki's Avatar

timo.mustamaki

20 Mar, 2024 01:37 PM

Hello!

I am trying to undestand why the F16 performance of my Tesla P100 is really no higher than the F32 result. Nvidia's datasheet https://images.nvidia.com/content/tesla/pdf/nvidia-tesla-p100-PCIe-datasheet.pdf implies that the P100 should have twice the half precision performance compared to full precision but according to my Geekbench ML score https://browser.geekbench.com/ml/v0/inference/360072 this is not the case.

Any insight? And please, none of that "Pascal is really old" stuff. I know it is. If the whitepaper says the performance should be there, why can it not be seen in practice?

Reply to this discussion

Internal reply

Formatting help / Preview (switch to plain text) No formatting (switch to Markdown)

Attaching KB article:

»

Attached Files

You can attach files up to 10MB

If you don't have an account yet, we need to confirm you're human and not a machine trying to post spam.

Keyboard shortcuts

Generic

? Show this help
ESC Blurs the current field

Comment Form

r Focus the comment reply box
^ + ↩ Submit the comment

You can use Command ⌘ instead of Control ^ on Mac