10-12-2026, 07:19 AM
Quote:My first guess would be this: Before rebuilding anything, I'd test one clean baseline with DeepSeek and compare from there. Once you know the plain pa
Appreciate the detail. That matches what I saw on dual 3090. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
2x3090 • llama.cpp • too much VRAM is still not enough
