NVIDIA Tesla P40
The NVIDIA Tesla P40 is a datacenter GPU (Pascal, GP102) with 24 GB of GDDR5 — in the homelab it's mainly prized as a cheap source of VRAM for local LLM inference. Note that it's passively cooled and has no display output, so it needs active airflow (a shroud/fan) and an EPS-to-PCIe power adapter for roughly 250 W. It is not suited to gaming. We track the P40 on eBay.de and score every listing against its own price history — don't confuse it with the smaller Tesla P4 or the P400.
Aggregated from 17 daily medians. Sources: eBay.de Browse API + Kleinanzeigen.de. We save every price as a snapshot on each ingest — the data baseline grows daily.
Current listings
No NVIDIA Tesla P40 listed right now — other Gpu deals:
808409-001 HPE Nvidia Tesla M6 Mezzanine GPU Adapter
NVidia Tesla A16 64GB Ampere Server VDI AI KI GPU PCIe x16 4.0 250W 5120 CUDA
HP nVidia Tesla A16 64GB GDDR6 Computing Grafikkarte 4x GPU PCIe x16 4.0 P48409-
Dell nVidia Tesla A16 0D1P1T 64GB GDDR6 Computing Grafikkarte 4x GPU PCIe x16 4.
Frequently asked questions
Is the Tesla P40 enough for running local LLMs?
Yes — its 24 GB of VRAM fits many quantized models entirely on the card, at a low used price. The Pascal architecture is slower than current GPUs and weaker at FP16, but it remains a popular entry point for self-hosting LLMs.
What should I check before buying a used P40?
Cooling and power. The card is passive and needs a fan shroud plus an EPS-to-PCIe adapter (~250 W), and neither is always included. Confirm those parts come with it, or budget for the extras.
Can I use the Tesla P40 for gaming or as a regular graphics card?
No. The P40 has no display output and is built purely as a compute GPU. You'd need a second GPU or an iGPU for video output, and it isn't intended for gaming.