NVIDIA Tesla P40

Compare NVIDIA Tesla P40 against another machine in the Arena

Reference: Nvidia Tesla on Wikipedia

Class
GPU
Memory
24 GB GDDR5
Runs (Q4_K_M)
—
Bandwidth
346 GB/s
TDP
250 W
Released
2016-09-01
Price (US)
$225 used · $329 new as of 2026-08
Price (Canada)
CA$339.49 used sourced, as of 2026-08
Sourcing and disambiguation notes

Passive-cooled server card: needs a 3rd-party blower shroud + high-static-pressure fan and an EPS-to-PCIe (CPU 8-pin) power adapter to run outside a server chassis. No video output. Extremely poor native FP16 (~1/64 of FP32 rate on Pascal), so GGUF/INT8-style quantized inference (llama.cpp) is the only practical path, not FP16 frameworks. No flash-attention kernel support. Price source: gpudojo.com price tracker (eBay-sourced, refreshed twice daily), accessed 2026-08-22. New-old-stock (NOS) 'new' figure of $329 is via Amazon, per the same gpudojo.com tracker cited above (used $225 via eBay). Canadian price: eBay.ca active listing, pre-owned 24GB GDDR5, plus ~$29 CAD shipping (price given is pre-tax item price only).

Specification sources

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 24 GB; models that fit are listed largest first.