News for hackers. Stuff that matters.
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by koakuma-chan on Friday July 11, 2025 @07:02AM (#44530755)
is 100 tflops a lot?
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by saagarjha on Friday July 11, 2025 @07:34AM (#44530967)
It's like 5-10% here
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by irrelative on Friday July 11, 2025 @11:07AM (#44532999)
Correct, this is the actual headline too. 100 tflops sure seems like it'd be more than that, but here we are.

If the headline was "FB8 is ~7% faster when kernel name has 'cutlass' in it...", it wouldn't seem sensational.

Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by saagarjha on Friday July 11, 2025 @04:02PM (#44536305)
I think the interesting part is that it improves performance measurably at all, not the actual number. These people are trying to hit 90+% MFU (though most don't reach it) so this does actually translate to many millions of dollars for them.
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by progx on Friday July 11, 2025 @07:43AM (#44531023)
5060 ti +~15%
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by brightmood on Friday July 11, 2025 @07:15AM (#44530841)
yea
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by HideousKojima on Friday July 11, 2025 @09:40AM (#44532027)
According to Terminator 3 Skynet used a mere 60 TFLOPS
Re:FP8 is ~100 tflops faster when the kernel name has "cutla... by IAmBroom on Friday July 11, 2025 @10:51AM (#44532800)
How much is that in jiggawatts per parsec?