Head-to-head for local LLM inference. The honest comparison: VRAM, bandwidth, compute, and which of the 30 catalog models actually fit on each.
// showing 12 of 30 models; differing fits first
Gaudi 3 has 48 GB more.
Gaudi 3 by +10%.
Gaudi 3 by +86% TFLOPS.
Tied: 27 of 30 fit on each.
// sign in with github to leave a comment. threads live in the repo's discussions tab.