3 ms·
Except by all accounts they succeeded. I believe they have the fastest matmul on nvidia chips in the industry
by mempko 1y ago
Except by all accounts they succeeded. I believe they have the fastest matmul on nvidia chips in the industry
- timmg 1y agoI was under the impression that their uptake it slow or non-existant. Am I wrong on that?
- ozgrakkurt 1y agoIs it really faster than cublas?
- melodyogonna 1y agoIn some things yes. They're mostly identical in performance though
- saagarjha 1y agoCUTLASS would like to have a word with you.
- fooblaster 1y agoevidence?
- chrislattner 1y agoModular/Mojo is faster than NVIDIA's libraries on their own chips, and open source instead of binary blob. See the 4 part series that culimates in https://www.modular.com/blog/matrix-multiplication-on-blackwell-part-4---breaking-sota https://www.modular.com/blog/matrix-multiplication-on-blackw... for Blackwell for example.
- fooblaster 1y agothanks