The Minutes of Dylan Patel on gpu performance https://minutesof.com/dylan-patel/on/gpu-performance/ Everything Dylan Patel has said on gpu performance: 4 verbatim quotes between March 2025 and July 2026, each with a timestamp and a link to the recording… en Sun, 30 Aug 2026 16:22:19 +0000 Patel claims Nvidia achieves 50% sustained performance on common 8k gemm while AMD gets only 30%. https://minutesof.com/q/f0d9f7af-e5f2-4a3f-9331-63fe03e874a1/ https://minutesof.com/q/f0d9f7af-e5f2-4a3f-9331-63fe03e874a1/ “if you just take a eight k by eight k by eight k gem, very common shape and you run it on the map mode unit of AMD and Nvidia, you get like 50% sustained performance on Nvidia and you get like 30% sustained performance on AMD.” — TensorWave Thu, 30 Apr 2026 18:18:14 +0000 gpu performance Patel says H100 cannot physically reach advertised 2,000 teraflops, maxing at 1,300 teraflops for matrix opera https://minutesof.com/q/7eb3f152-085c-4b25-9dd3-33f9d606c0a6/ https://minutesof.com/q/7eb3f152-085c-4b25-9dd3-33f9d606c0a6/ “In fact, it's physically impossible to get 2,000 teraflops out of the H100. Even though the advertises it, at most you can get 1,300 FP8 floating point eight teraflops.” — Clockwork Fri, 21 Nov 2025 17:18:51 +0000 gpu performance Patel quantifies Hopper performance improved 30-40% over 2024 with Blackwell showing similar gains in months. https://minutesof.com/q/ec5cbc83-8dcb-4132-b787-4f7c4639cd91/ https://minutesof.com/q/ec5cbc83-8dcb-4132-b787-4f7c4639cd91/ “Even Hopper in January 24 to December 2024, you still had like a 30%, 40% performance improvement. And likewise for Blackwell, you've seen a similar sort of improvement, right?” — Together AI Fri, 03 Oct 2025 20:23:30 +0000 gpu performance Patel says GPUs in a 16,000 GPU cluster show up to 10% speed variation from manufacturing differences. https://minutesof.com/q/e4cb87fa-869e-4b13-8b85-bd4d3fb6cb18/ https://minutesof.com/q/e4cb87fa-869e-4b13-8b85-bd4d3fb6cb18/ “But even within a cluster of like 16 k GPUs, GPUs will be slower by up to 10%. Right? So there there's a lot of variation even in the manufacturing.” — Prime Intellect AI Tue, 25 Mar 2025 18:21:45 +0000 gpu performance