AMD GPUs Go Brrr(hazyresearch.stanford.edu)
222 points by vinhnx | 71 comments
First screen, captured .
Looks similar
We bought the whole GPU, so we're damn well going to use the whole GPU(hazyresearch.stanford.edu)
472 points by sydriax | 94 comments
Stanford Hazy Research blog post introducing a throughput-optimized megakernel for tensor-parallel Llama-70B inference on H100 GPUs that overlaps compute, memory, and communication.
Reasonable GPUs
27 points by frognumber | 20 comments
Related stories
We were wrong about GPUs(fly.io)
834 points by mxstbr | 529 comments
Fly.io blog post reflecting on how the company's bet on offering GPU machines to developers did not pan out as expected.
Using AI to reverse engineer hardware(blogs.gnome.org)
224 points by ErenayDev | 198 comments
Blog post by Christian F.K. Schaller, a Red Hat developer, about using AI tools to reverse engineer hardware, presumably for Linux driver development.
The GPU, not the TPM, is the root of hardware DRM(mjg59.dreamwidth.org)
399 points by DvdGiessen | 408 comments
Blog post by Matthew Garrett arguing that GPUs, not TPMs, are the real root of hardware DRM; the capture shows a Dreamwidth captcha page instead.
TPUs vs. GPUs and why Google is positioned to win AI race in the long term(uncoveralpha.com)
375 points by vegasbrianc | 278 comments
A Substack deep dive from UncoverAlpha on Google's TPU chips, covering their technical design, strategic position versus GPUs, and financial implications for AI inference.
We bought the whole GPU, so we're damn well going to use the whole GPU(hazyresearch.stanford.edu)
472 points by sydriax | 94 comments
Stanford Hazy Research blog post introducing a throughput-optimized megakernel for tensor-parallel Llama-70B inference on H100 GPUs that overlaps compute, memory, and communication.
What every developer should know about GPU computing(codeconfessions.substack.com)
422 points by Anon84 | 149 comments
A Substack primer by Abhinav Upadhyay explaining GPU architecture and computing concepts for software developers familiar mainly with CPUs.