NVIDIA Tesla P100 16GB (PCIe)

Compare NVIDIA Tesla P100 16GB (PCIe) against another machine in the Arena

Class: GPU
Memory: 16 GB HBM2
Runs (Q4_K_M): —
Bandwidth: 732 GB/s
TDP: 250 W
Released: 2016-06-01
Price (US): $78 used · $94 new as of 2026-08
Price (Canada): CA$184.22 used sourced, as of 2026-08

Sourcing and disambiguation notes

Also sold as an SXM2 module with the same 732 GB/s HBM2, which needs a server carrier board and does not fit a standard desktop PCIe slot; a separate SXM2 llama.cpp result exists but is not merged here, to avoid mixing form factors. No video output, and passive cooling on most pulled units. Real-world decode on this chip is oddly bandwidth-underutilised in llama.cpp, with contributors reporting 41-43% memory-controller utilisation, so it under-performs its bandwidth advantage over the P40. The Canadian figure is for the PCIe card; cheaper SXM2 listings exist around CA$113 but need an adapter board, so the PCIe price is used as representative.

Specification sources

https://gpudojo.com/tesla-p100 — press, 2026-08-22
https://cputronic.com/en/gpu/nvidia-tesla-p100-pcie-16-gb — press, 2026-08-22

Measurements

Decode is token generation — the speed you feel while an answer streams. Prefill is prompt processing — the wait before it starts. Why bandwidth predicts decode speed.

What verified, single-source and estimated mean, and the same rows with every filter and sort in the benchmarks explorer.

Speed over time

What it can run

Fit is arithmetic, not a measurement — how it is computed. Against 16 GB; models that fit are listed largest first.