Feat/more gpus#1213
Conversation
garloff
added
enhancement
|
Good luck obtaining them 😂 |
|
Added to agenda of next SIG Std/Cert on Thursday (2026-06-18) for discussion. Do tests and tooling need to be adjusted as well? |
I think the "u" needs to be added here: |
|
Apparently I wasn't awake ... |
|
GPU providers: Please have a look and let us know whether the updated table makes sense for you. |
|
Feedback welcome, also from others of course ... |
|
I believe this should be a non-brainer: We evolve our implementation hints (and for the blackwell ultra also our standard) along the most recent hardware. This keeps it viable and consistent. |
|
@garloff I think this is undisputed. We wanted to give operators the chance to weigh in (and maybe dispute it for the first time), but (apparently) we can't pressure them to do so... |
This way we keep the tables current and helpful. Signed-off-by: Kurt Garloff <kurt@garloff.de>
The tensor cores are different enough to justify this. Main use case for GPUs in clouds is AI and the tensor cores do make a difference there. Signed-off-by: Kurt Garloff <kurt@garloff.de>
Signed-off-by: Kurt Garloff <kurt@garloff.de>
Signed-off-by: Kurt Garloff <kurt@garloff.de>
mbuechse
left a comment
mbuechse
left a comment
There was a problem hiding this comment.
No one opted to weigh in. And I don't see a reason to block this any longer.
3 participants
Adding one letter
ufor Blackwell Ultra (GB300) and a lot more examples in the GPU table.Added a disclaimer that providers need to use real values and not rely on data sheet fed table values.