Decommissioned NVIDIA enterprise GPUs are one of the last remaining sources of idle VRAM. K80 with 24GB of GDDR5 sells for $60, P100-16GB for around $75 and V100-16GB for under $200. Shortcomings and all, it’s important to understand if we can utilize these widely available cards in the modern era. This project to benchmark tesla GPUs has been in the works for almost a year and over the winter I was able to spend many kilowatt hours heating up my studio with GPUs.
The rapid ascent of the LLM into the collective consciousness has sent the big players into a frenzy over datacenter GPUs. This is putting accelerating downward pressure on the price of all used compute GPUs, even the historically pricey stuff. P100s can be had for ~$100, V100 16GB are selling for ~$500, any day now the lower VRAM Ampere cards are going to drop below $1000…
We need a new cooler that can support any of these 300W GPUs, can be installed as densely as the cards, and is easy on the ears for homelab use. Pictured below are four prototype coolers installed on Tesla P100s:
Four PCBs installed on Four P100s on the Open Bench Table build.
Edit: This project was completed hackathon-style in a matter of days. I’ve been working to optimize the design and sell kits, follow along here.
Here’s a (long winded) video overview of this project:
Background
Rendered desperate for VRAM by a forthcomingnow released! stylegan-related project, I recently had to wade thermistor first into the concernedly hot and strange world of GPUs without video outputs to design a high performance cooler for the NVIDIA Tesla K80.
Too esoteric to game on, and too power hungry to mine cryptocurrencies with, the K80 (allegedly the ‘The World’s Most Popular GPU’) can be had for under $250 USD on ebay, a far cry from it’s imperial MSRP of $5000. By my math, the card is one of the most cost-efficient ways to avail one’s self of video ram by the dozen of gigabytes.
This sounds great on paper, but actually getting one of these configured to do useful work is a kind of a project in, and of itself. I’ll eventually get to this in the aforementioned upcoming post. Today’s topic however, is upstream of all that: the task of keeping these things cool.