Back to AIBriefs
AnalysisDevelopers

llama.cpp build b9455 boosts speed on dual RTX 3090

llama.cpp build b9455 boosts speed on dual RTX 3090View on Reddit

User reports improved performance with llama.cpp build b9455 on 2x3090 GPUs, previously getting 30-50 tk/s. The new build appears to close the gap with vLLM for tensor-parallel inference.

·
14 days ago
llama.cpp build b9455 boosts speed on dual RTX 3090 — AIBriefs