$200 'socketed' Nvidia AI GPU for servers hacked into a PCIe card with custom PCB and 3D-printed cooling — modded Tesla V100 SMX data center GPU runs AI LLMs and is more efficient than many modern midrange offerings in AI inference

A Hardware Haven video showed a $100 Nvidia Tesla V100 server GPU being converted to PCIe with a custom adapter and 3D-printed cooling.

In tests on a Ryzen system, it delivered strong LLM and NVR inference performance, including high token throughput in Ollama and Frigate.

At similar power limits, the modded V100 was more efficient than an RTX 3060 and competitive with newer consumer GPUs.

The result highlights how older datacenter cards can be a low-cost option for local AI workloads thanks to their large VRAM and solid inference speed.