User shares budget multi-GPU build for local inference

View on Reddit

Reddit user upgraded their budget build for local inference with 2x RTX 3090 (24GB each) and an Intel Arc A770 (16GB), totaling 64GB VRAM for under $850. The all-second-hand setup runs large LLMs across three GPUs.