undefined | Better HN

0 pointsrfoo1y ago0 comments

They have H800s which have exactly same memory bandwidth and max FLOPS.

0 comments

What about NVLink? Does it plays a role here?

rfooOP1y ago

For FlashMLA? No. The code here runs on one GPU only and do not have a builtin communication part.

But for the training it does. You need to communicate gradient changes between GPUs.

j / k navigate · click thread line to collapse

What about NVLink? Does it plays a role here?

rfooOP1y ago

For FlashMLA? No. The code here runs on one GPU only and do not have a builtin communication part.

But for the training it does. You need to communicate gradient changes between GPUs.

j / k navigate · click thread line to collapse