Lawyer upgrades eight-V100 private AI server to run 80% of DeepSeek V4.1 Flash at 4-bit NVFP4

A lawyer running a private AI server on eight used Nvidia V100 GPUs moved about 80% of its DeepSeek V4.1 Flash legal workload from 2-bit to 4-bit NVFP4 precision, adding 1TB of RAM and an RTX 3090. Code output dropped from 160–215 to 150–180 tokens per second, but the crashes that hit the new build are gone, with a bad PSU rather than PCIe the likely culprit.