readygpu

Intel Arc B580 vs NVIDIA Tesla P40

For running models locally, priced in Germany. At 32k context the first runs 10 models fully in memory, the second 26.

In short

The NVIDIA Tesla P40 has more memory for models: 24 GB against 12 GB. At 32k context the NVIDIA Tesla P40 runs 26 of the models we track fully in memory and the Intel Arc B580 10; only the NVIDIA Tesla P40 holds Qwen 3.8 27B. On the models both run fully in memory, the Intel Arc B580 is about 1.2× as fast as the NVIDIA Tesla P40 (estimated, median across models). In Germany the Intel Arc B580 costs 349 € at eBay and the NVIDIA Tesla P40 369 € at eBay (used): the Intel Arc B580 is 20 € cheaper.

Updated . Speeds are estimates; how they are worked out.

Intel Arc B580

349 €

eBay

Add to build
NVIDIA Tesla P40

369 €

eBay, used (asking)

Add to build

Specifications

SpecIntel Arc B580NVIDIA Tesla P40
GPUArc B580Tesla P40
Memory12 GB GDDR624 GB GDDR5
Memory bus192-bit384-bit
Memory bandwidth456 GB/s346 GB/s
FP16 compute116.5 TFLOPS11.8 TFLOPS
SoftwareSYCL / oneAPICUDA
Board power190 W250 W
PCIe generationGen 4Gen 3
PCIe lanesx8x16

What each runs

Estimated decode speed with the most precise quant that fits each one est.

ModelIntel Arc B580NVIDIA Tesla P40
Qwen 3.8 27Boffload10tok/sUD-Q5_K_M
Qwen 3.6 27Boffload10tok/sQ5_K_M
Gemma 4 31Boffload10tok/sQ4_K_M
Qwen 3.6 35B-A3Boffload50tok/sUD-Q4_K_M
Qwen 3.5 27Boffload10tok/sQ5_K_M
Gemma 4 26B-A4Boffload29tok/sUD-Q6_K
Qwen 3.5 35B-A3Boffload50tok/sQ4_K_M
GLM 4.7 Flashoffload24tok/sQ5_K_M
Devstral Small 2 24Boffload10tok/sQ5_K_M
gpt-oss 20Boffload40tok/sMXFP4
Qwen3 Coder 30B-A3Boffload26tok/sQ4_K_M
Gemma 4 12B21tok/sQ6_K16tok/sQ8_0
Qwen 3.5 9B21tok/sQ8_020tok/sQ8_0
Gemma 3 27Boffload10tok/sQ5_K_M
Mistral Small 3.2 24Boffload10tok/sQ5_K_M
Ministral 3 14Boffload11tok/sQ8_0
Qwen3 14Boffload10tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B21tok/sQ5_K_M16tok/sQ8_0
Qwen3 8B21tok/sQ5_K_M16tok/sQ8_0
Phi-4 14Boffload10tok/sQ8_0
Qwen 3.5 4B41tok/sQ8_038tok/sQ8_0
Ministral 3 8B22tok/sQ5_K_M16tok/sQ8_0
Granite 4.2 8B21tok/sQ4_K_M15tok/sQ8_0
Gemma 4 E4B25tok/sQ8_023tok/sQ8_0
LFM2.5 8B-A1B47tok/sQ8_048tok/sQ8_0
Llama 3.1 8B21tok/sQ6_K20tok/sQ6_K
Qwen 3.5 122B-A10Bwon't runoffload
Qwen3 Coder Nextwon't runoffload
Mistral Small 4 119Bwon't runoffload
GLM 4.5 Airwon't runoffload
Qwen3 Next 80B-A3Bwon't runoffload
Nemotron 3.5 Lightning 30B-A3Boffloadoffload
Kimi Linear 48B-A3Boffloadoffload
Qwen3 32Boffloadoffload
DeepSeek R1 Distill Qwen 32Boffloadoffload
DeepSeek V4 Flashwon't runwon't run
GLM 5.3 Flashwon't runwon't run
Qwen 3.8 Flash Nextwon't runwon't run
MiniMax M2.7won't runwon't run
Nemotron 3 Super 120B-A12Bwon't runwon't run
gpt-oss 120Bwon't runwon't run
Llama 3.3 70Bwon't runwon't run

Questions

Which is better for local AI, the Intel Arc B580 or the NVIDIA Tesla P40?
Choose the NVIDIA Tesla P40 to run bigger models. Choose the Intel Arc B580 for speed. Choose the Intel Arc B580 to spend less.
Which can run bigger models, the Intel Arc B580 or the NVIDIA Tesla P40?
At 32k context the NVIDIA Tesla P40 runs 26 of the models we track fully in memory and the Intel Arc B580 10; only the NVIDIA Tesla P40 holds Qwen 3.8 27B.
Which is faster for local AI, the Intel Arc B580 or the NVIDIA Tesla P40?
On the models both run fully in memory, the Intel Arc B580 is about 1.2× as fast as the NVIDIA Tesla P40 (estimated, median across models).
What is the price difference between the Intel Arc B580 and the NVIDIA Tesla P40?
In Germany the Intel Arc B580 costs 349 € at eBay and the NVIDIA Tesla P40 369 € at eBay (used): the Intel Arc B580 is 20 € cheaper.

More comparisons