Darshan (@neuronfitting) header
Darshan (@neuronfitting) profile picture

making gpu's that go brrrrr @AgraniLabs

Joined December 2024
Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)Photo by Darshan (@neuronfitting)
from making gpu's go brrr to making gpu's that go brrr I am joining @AgraniLabs as an Performance Architect Intern
1
4
220
In hindsight it's obvious that the fastest path to solving Millennium Prize Problems was to play lots of video games in order for GPUs to improve, so that ML could develop and language models could get trained. Gamers, thank you for your valuable contribution to mathematics!
93
666
76
9,688
151,803
Replying to @henrylhtsang
That what torch said
1
24
Replying to @say_lee_x
Link reset?
22
1) what?
India is playing a whole new game!
36
Replying to @maharshii
Nvidia doesn't have fp6 cores.... Where is the speedup comming from?
49
Wow
The only thing faster than GB300 Dynamo + vLLM is Vera Rubin Dynamo + vLLM 🏎️💨. The first DeepSeek V4.1 Flash results on @NVIDIA Vera Rubin NVL72 are here, powered by Dynamo + vLLM on @SemiAnalysis_'s AgentX benchmark. Compared with GB300 NVL72 running Dynamo + vLLM, InferenceX's perf ruler shows: • 63.4× total token throughput per chip at 400 tok/s/user • 93.3× at 600 tok/s/user
1
108
Replying to @zephyr_z9
Nvlink is the moat
261
Replying to @maharshii
You can look into sgl-nvidia-megakerles repo Its a nice learning materal for PDL, they have got good gains
1
2
182
Replying to @AryaTschand
"and the insights should generalize to any hybrid transformer + MoE model architecture" -- why is that? can't think about it from first principles
23
claude --remote-control is the best thing this world has every produced; i can't switch to codex just because of this feature!
1
2
190
Darshan retweeted
A good reminder that just because you cuda doesn’t mean you shuda.
38
217
24
2,778
79,984
A man who pioneered parallel processing cannot be expected to run a single thread
261
2,779
459
31,774
1,059,704
Replying to @KheteshAkoliya
We'll mog chinese olympics
16
Public infra: Ahmedabad >>> Bangalore
1
2
100
Replying to @cudagdb
Isn't that just a modern day megakernel The thing which nvida-sglang-megakernel repo is doing
45
Literally me after reading sglang PR which adds megakernel support for some chinese model to improve performance on inferncex
1
41