I got 2.2x more tokens per second from llama.cpp on Intel Arc | Hacker News - But Better

I got 2.2x more tokens per second from llama.cpp on Intel Arc grigio.org

I spent a weekend benchmarking llama.cpp on my Xiaomi Book Pro 14. The machine has Intel's Core Ultra X7 358H and Arc B390 integrated graphics, with 30 GiB of unified memory. That last part matters. It means the GPU…

4 points•grigio•2 days ago•0 comments•

#cpp #arc #llama #tokens #intel #intel arc #spent #weekend

0 comments

No comments yet.

Related stories

Show HN: Jeva.cpp – a llama.cpp fork with JEV-compatible API for all LLMsgithub.com
Hacker News · 2 points · 2 days ago
Show HN: Llama.cpp fork with 2-4x multiGPU speed for MoE models bigger than VRAMgithub.com
Hacker News · 1 points · 5 days ago
Arc and Scheme in Emacs Lisprepl.it
Lobsters · 10 points · over 7 years ago
Llama.cpp Under the Hoodcppdepend.com
Hacker News · 2 points · 8 days ago
Using Llama-cpp-Python grammars to generate JSONtil.simonwillison.net
Hacker News · 1 points · 8 days ago
Custom Models in Oh My Pi: vLLM, Llama.cpp, SGLang and Moredoug.sh
Hacker News · 2 points · 5 days ago