Two Tesla Referral P100 16GB PCIe cards. Same PH400 / 699-2H400-0201-530 cards. Both work. Selling because they will not fit as a pair in my half-height case. My loss, your gain.
16GB each, 32GB together. Not SXM2. Fine for a local LLM box if you have a real case and a modular PSU.
This is not bare cards. You get a working printed shroud + blower, a manual fan controller, and the power adapters already figured out. That is the extra value. One card uses twin PCIe 8-pin leads. The other uses a CPU / EPS 8-pin to Tesla Referral adapter, plus a StarTech EPS8EXT 8-inch extension. You still need your own cooler for the second card. I will pack the cooler and the wiring separate from the cards so nothing snaps in the box.
Modular PSU is required if you want to use the CPU-power route. Do not jam a PCIe 8-pin into a CPU 8-pin.
I have almost 600 response at 100% positive. Cards, cooler, knobs, and power setup all worked before they came apart for packing. Used server cards, looks like the photos.
Bought separately you are looking at well over $300 in cards and necessary accessories. I would rather move the whole kit.
You need a full-size case, two long slots, and another blower for card two. These run hot, so I had mine running at 150w.
On my Linux box a single card was doing about 24 tokens per second on Qwen 3.8. Two cards will not magically double that, but you get the headroom for a better quant and more context.
Don't delay while GPU prices continue to skyrocket! Buy this enthusiast bundle from a trusted seller today!
