Advokatens 8x V100-server går ner på 4-bitars precision – DeepSeek V4.1 Flash får nytt liv
2026-11-14 • Kategori: Hårdvara
TL;DR: Advokat uppgraderar privat AI-server med 8x V100 och 1TB RAM, byter till 4-bitars precision för DeepSeek V4.1 Flash trots något lägre token-hastighet, och löser kraschproblem.
A lawyer with a private AI server built on eight used Nvidia V100 GPUs returned with an update, moving about 80% of its work up from 2-bit to 4-bit NVFP4 precision and bringing in 1TB of RAM and an RTX 3090. The server uses DeepSeek V4.1 Flash with a legal harness, the workflows and checks he wraps around the model, for his law practice. The cost? Performance, with code output dropping from 160–215 to 150–180 tokens per second, although the crashes that hit the new build are gone. The culprit behind the crashes, originally blamed on PCIe, was likely a bad PSU.Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.