I Put a Datacenter GPU in My Gaming PC for £200
Summary
This article documents a DIY upgrade: fitting a Tesla V100 SXM2 datacenter GPU into a gaming PC via an SXM2-to-PCIe adapter to run a local 27B LLM model. It covers hardware mods, NVLink potential, performance metrics (about 32 tokens per second with 32GB total VRAM across two GPUs), and the Linux/NixOS software setup required to make it work. The piece also discusses practical drawbacks like fan noise and ACPI quirks, offering a compelling case for affordable local AI inference.