readygpu

AMD Radeon RX 7900 XT vs NVIDIA Tesla P40

For running models locally, priced in the UK. At 8k context the first runs 53 models fully in memory, the second 61.

In short

The NVIDIA Tesla P40 has more memory for models: 24 GB against 20 GB. At 8k context the NVIDIA Tesla P40 runs 61 of the models we track fully in memory and the AMD Radeon RX 7900 XT 53; only the NVIDIA Tesla P40 holds Gemma 4 31B. On the models both run fully in memory, the AMD Radeon RX 7900 XT is about 2.1× as fast as the NVIDIA Tesla P40 (estimated, median across models). In the UK the AMD Radeon RX 7900 XT costs £789 at Amazon and the NVIDIA Tesla P40 £783 at Amazon: the NVIDIA Tesla P40 is £6 cheaper.

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

AMD Radeon RX 7900 XT

£789

Amazon

Add to build
NVIDIA Tesla P40

£783

Amazon

Add to build

Specifications

SpecAMD Radeon RX 7900 XTNVIDIA Tesla P40
GPURadeon RX 7900 XTTesla P40
Memory20 GB GDDR624 GB GDDR5
Memory bus320-bit384-bit
Memory bandwidth800 GB/s346 GB/s
FP16 compute103 TFLOPS11.8 TFLOPS
SoftwareROCmCUDA
Board power315 W250 W
PCIe generationGen 4Gen 3
PCIe lanesx16x16

What each runs

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

ModelAMD Radeon RX 7900 XTNVIDIA Tesla P40
Qwen 3.8 27B24tok/sQ4_K_M10tok/sUD-Q6_K
Qwen 3.6 27B25tok/sQ4_K_M9tok/sQ6_K
Gemma 4 31Boffload9tok/sQ5_K_M
Qwen 3.5 27B25tok/sQ4_K_M9tok/sQ6_K
Qwen 3.6 35B-A3Boffload62tok/sQ4_K_M
Gemma 4 26B-A4B78tok/sQ5_K_M32tok/sUD-Q6_K
GLM 4.7 Flash76tok/sQ4_K_M32tok/sQ5_K_M
Devstral Small 2 24B24tok/sQ5_K_M10tok/sQ6_K
Qwen 3.5 35B-A3Boffload62tok/sQ4_K_M
Nemotron 3.5 Lightning 30B-A3Boffload56tok/sMXFP4
Qwen3 Coder 30B-A3B100tok/sQ4_K_M44tok/sQ5_K_M
gpt-oss 20B100tok/sMXFP449tok/sMXFP4
Qwen3 32Boffload10tok/sQ4_K_M
Gemma 4 12B34tok/sQ8_016tok/sQ8_0
Qwen 3.5 9B45tok/sQ8_022tok/sQ8_0
Gemma 3 27B25tok/sQ4_K_M9tok/sQ6_K
Mistral Small 3.2 24B24tok/sQ5_K_M10tok/sQ6_K
DeepSeek R1 Distill Qwen 32Boffload10tok/sQ4_K_M
Ministral 3 14B28tok/sQ8_014tok/sQ8_0
Qwen3 14B26tok/sQ8_013tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B44tok/sQ8_022tok/sQ8_0
Qwen3 8B44tok/sQ8_022tok/sQ8_0
Phi-4 14B26tok/sQ8_012tok/sQ8_0
Qwen 3.5 4B92tok/sQ8_045tok/sQ8_0
Ministral 3 8B43tok/sQ8_021tok/sQ8_0
Granite 4.2 8B41tok/sQ8_020tok/sQ8_0
Gemma 4 E4B52tok/sQ8_025tok/sQ8_0
LFM2.5 8B-A1B180tok/sQ8_088tok/sQ8_0
Llama 3.1 8B57tok/sQ6_K28tok/sQ6_K
Qwen2.5 0.5B697tok/sQ8_0340tok/sQ8_0
Qwen3.5 0.8B472tok/sQ8_0230tok/sQ8_0
Qwen3.5 2B283tok/sQ5_K_M138tok/sQ5_K_M
LFM3 1.7B225tok/sQ8_0110tok/sQ8_0
DeepSeek R1 Distill Qwen 1.5B207tok/sQ8_0101tok/sQ8_0
MiniCPM5 2B145tok/sQ8_071tok/sQ8_0
LFM2.5 2.6B130tok/sQ8_064tok/sQ8_0
LFM2.5 VL 3B130tok/sQ8_064tok/sQ8_0
Spark X2.5 4B130tok/smlx-4bit63tok/smlx-4bit
LFM2.5 24B A2B119tok/sQ5_K_M53tok/sQ6_K
granite vision 4.1 4b103tok/sQ8_050tok/sQ8_0
Ministral 3 3B Instruct 251297tok/sQ8_047tok/sQ8_0
granite 4.2 3b96tok/sQ8_047tok/sQ8_0
Mellum2.1 12B A2.5B Thinking94tok/sQ8_046tok/sQ8_0
FrogNano 4B 260989tok/sQ8_043tok/sQ8_0
Kimi VL A3B Thinking 250679tok/sQ8_038tok/sQ8_0
Kimi VL A3B75tok/smlx-8bit37tok/smlx-8bit
DeepSeek R1 Distill Qwen 7B51tok/sQ8_025tok/sQ8_0
GLM 4 9B 041451tok/smlx-5bit25tok/smlx-5bit
DeepSeek R1 Distill Llama 8B46tok/sQ8_022tok/sQ8_0
Ministral 3 8B Reasoning 251243tok/sQ8_021tok/sQ8_0
GLM 4.6V Flash43tok/sQ8_021tok/sQ8_0
GLM 4.1V 9B Thinking36tok/smlx-8bit18tok/smlx-8bit
Ministral 3 14B Reasoning 251228tok/sQ8_014tok/sQ8_0
DeepSeek R1 Distill Qwen 14B25tok/sQ8_012tok/sQ8_0
Muse Glimmer 30B25tok/sQ4_K_M9tok/sQ6_K
Devstral Small 250724tok/sQ5_K_M12tok/sQ5_K_M
Magistral Small 250624tok/sQ5_K_M10tok/sQ6_K
Magistral Small 250724tok/sQ5_K_M12tok/sQ5_K_M
Magistral Small 250924tok/sQ5_K_M12tok/sQ5_K_M
granite 4.2 30boffload9tok/sQ5_K_M
Qwen3.8 27B OBLITERATEDoffload10tok/smlx-4bit
Qwen 3.5 122B-A10Boffloadoffload
Qwen3 Coder Nextwon't runoffload
Mistral Small 4 119Boffloadoffload
GLM 4.5 Airoffloadoffload
Qwen3 Next 80B-A3Bwon't runoffload
Kimi Linear 48B-A3Boffloadoffload
Llama 3.3 70Boffloadoffload
DeepSeek R1 Distill Llama 70Boffloadoffload
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
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Zerowon't runwon't run
DeepSeek V3won't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.1 Terminuswon't runwon't run
DeepSeek V3.2won't runwon't run
DeepSeek V4 Flash 0731won't runwon't run
DeepSeek V4 Flash Vision Expwon't runwon't run
Devstral 2 123B Instruct 2512won't runwon't run
GLM 4.5won't runwon't run
GLM 4.5Vwon't runwon't run
GLM 4.6won't runwon't run
GLM 4.6Vwon't runwon't run
GLM 4.7won't runwon't run
GLM 5won't runwon't run
GLM 5.1won't runwon't run
GLM 5.2won't runwon't run
GLM 5.3won't runwon't run
Kimi Dev 72Bwon't runwon't run
Kimi K2won't runwon't run
Kimi K2 Instruct 0905won't runwon't run
Kimi K2 Thinkingwon't runwon't run
Kimi K2.5won't runwon't run
MiniMax M2won't runwon't run
MiniMax M2.1won't runwon't run
MiniMax M2.5won't runwon't run
MiniMax M3won't runwon't run
Mistral Medium 3.5 128Bwon't runwon't run

Questions

Which is better for local AI, the AMD Radeon RX 7900 XT or the NVIDIA Tesla P40?
Choose the NVIDIA Tesla P40 to run bigger models. Choose the AMD Radeon RX 7900 XT for speed. Choose the NVIDIA Tesla P40 to spend less.
Which can run bigger models, the AMD Radeon RX 7900 XT or the NVIDIA Tesla P40?
At 8k context the NVIDIA Tesla P40 runs 61 of the models we track fully in memory and the AMD Radeon RX 7900 XT 53; only the NVIDIA Tesla P40 holds Gemma 4 31B.
Which is faster for local AI, the AMD Radeon RX 7900 XT or the NVIDIA Tesla P40?
On the models both run fully in memory, the AMD Radeon RX 7900 XT is about 2.1× as fast as the NVIDIA Tesla P40 (estimated, median across models).
What is the price difference between the AMD Radeon RX 7900 XT and the NVIDIA Tesla P40?
In the UK the AMD Radeon RX 7900 XT costs £789 at Amazon and the NVIDIA Tesla P40 £783 at Amazon: the NVIDIA Tesla P40 is £6 cheaper.

More comparisons

More hardware for local AI