BREAKING NEWS🚨: NVIDIA has mainly only shown love to Blackwell Ultra & Rubin with its decode-specialized software, but has shown mostly none to the GPU Poors on Hopper or Blackwell 😭 Luckily, @mgoin_ cares about the proletarians too and got his agent swarm to write 10,000 LoC of ultra-fused Hopper kernels that improve decode speeds by up to 38%!
Agent swarm writes Hopper decode kernels that speed up decoding up to 38%
AImgoin used an agent swarm to write 10,000 lines of fused Hopper kernels that improve decode speeds by up to 38%. The post says NVIDIA's decode-specialized software has focused on Blackwell Ultra and Rubin rather than Hopper or Blackwell.
Source details
SemiAnalysis · x.comPublished · added here
Original title: BREAKING NEWS🚨: NVIDIA has mainly only shown love to Blackwell Ultra & Rubin with its decode-specialized software, but has shown mostly ...