NVIDIA A100 40GB SXM4 vs NVIDIA RTX PRO 6000 Blackwell Workstation Edition, AI & Machine Learning Comparison
NVIDIA RTX PRO 6000 Blackwell Workstation Edition wins 143 of 145 benchmarks, averaging 72.9% faster.
NVIDIA A100 40GB SXM4's numbers are anchored estimates calibrated against our measured cards, pending first-party measurement. Treat small gaps as ties.
What the numbers say
The gap is widest in Swin2SR 4x Upscaler, where NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by 206% (17.25 vs 52.83 images/min); the closest fight is Qwen2.5 7B served (29% apart); VRAM decides part of this one: NVIDIA RTX PRO 6000 Blackwell Workstation Edition runs 144 of our 12 AI workloads while the other card runs 135, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA A100 40GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| Qwen3-4B tok/s | 193.14 | 362.81 | -47% |
| Llama-3.1-8B tok/s | 157.26 | 258.37 | -39% |
| Qwen2.5-Coder-14B tok/s | 86.2 | 147.74 | -42% |
| Qwen3-32B tok/s | 43.8 | 70.13 | -38% |
| Llama-3.3-70B tok/s | 0 | 34.87 | n/a |
| Stable Diffusion XL images/min | 18.576 | 27.94 | -34% |
| Z-Image Turbo images/min | 9.975 | 15.6 | -36% |
| FLUX.1 dev images/min | 4.264 | 6.921 | -38% |
| FLUX.1 Kontext dev images/min | 1.993 | 3.086 | -35% |
| Qwen-Image-Edit images/min | 0 | 2.64 | n/a |
| LTX-Video (distilled) frames/s | 8.95 | 16.36 | -45% |
| DeepSeek-R1 Distill Llama 8B tok/s | 157.16 | 235.07 | -33% |
| DeepSeek-R1 Distill 1.5B tok/s | 320.78 | 635.1 | -49% |
| DeepSeek-R1 Distill 14B tok/s | 86.39 | 135.41 | -36% |
| DeepSeek-R1 Distill 7B tok/s | 159.09 | 256.04 | -38% |
| Gemma 3 12B tok/s | 88.97 | 138.09 | -36% |
| Gemma 3 4B tok/s | 174.14 | 297.73 | -42% |
| Gemma 4 12B tok/s | 87.14 | 137.73 | -37% |
| Llama 3.2 1B tok/s | 532.08 | 892.74 | -40% |
| Llama 3.2 3B tok/s | 261.24 | 435.63 | -40% |
| Mistral 7B v0.3 tok/s | 165.9 | 247.33 | -33% |
| Mistral Small 24B tok/s | 62.42 | 93.55 | -33% |
| Phi-4 14B tok/s | 98.65 | 142.68 | -31% |
| Phi-4 Mini 3.8B tok/s | 236.71 | 391.04 | -39% |
| Qwen2.5-Coder 7B tok/s | 159.9 | 256.11 | -38% |
| Qwen3 0.6B tok/s | 417.88 | 788.97 | -47% |
| Qwen3 1.7B tok/s | 340.76 | 575.77 | -41% |
| Qwen3 14B tok/s | 90.01 | 139.01 | -35% |
| Qwen3 30B A3B tok/s | 169.1 | 308.74 | -45% |
| Qwen3 8B tok/s | 146.21 | 225.95 | -35% |
| SmolLM3 3B tok/s | 246.25 | 406.45 | -39% |
| Codestral 22B tok/s | 63.74 | 92.7 | -31% |
| DeepSeek-R1 Distill 32B tok/s | 43.13 | 66.06 | -35% |
| Devstral Small 24B tok/s | 62.4 | 93.52 | -33% |
| Dolphin 2.9.1 Yi 1.5 34B tok/s | 42.97 | 63.74 | -33% |
| Dolphin Mistral 24B Venice tok/s | 62.46 | 93.67 | -33% |
| Dolphin X1 8B tok/s | 156.3 | 239.4 | -35% |
| Dolphin 3.0 Llama 3.1 8B tok/s | 156.55 | 239.28 | -35% |
| Dolphin 3.0 R1 Mistral 24B tok/s | 62.4 | 93.62 | -33% |
| Gemma 3 27B tok/s | 47.47 | 72.1 | -34% |
| Qwen2.5-Coder 32B tok/s | 43.14 | 66.07 | -35% |
| Qwen3-Coder 30B A3B tok/s | 173.47 | 317.68 | -45% |
| QwQ 32B tok/s | 43.02 | 66.07 | -35% |
| StarCoder2 15B tok/s | 80.25 | 122.09 | -34% |
| FLUX.1 Schnell images/min | 28.44 | 42.41 | -33% |
| Z-Image Turbo (1024px) images/min | 18.7 | 31.43 | -41% |
| BiRefNet images/min | 611.56 | 1445.92 | -58% |
| Depth Anything V2 Large images/min | 536.74 | 1328.11 | -60% |
| Depth Anything V2 Small images/min | 613.42 | 1374.58 | -55% |
| SAM ViT-Base images/min | 696.77 | 1079.89 | -35% |
| SAM ViT-Huge images/min | 146.81 | 193.47 | -24% |
| Swin2SR 4x Upscaler images/min | 17.25 | 52.83 | -67% |
| Dolphin X1 Trinity Nano 6B tok/s | 153.27 | 282.02 | -46% |
| gpt-oss-20b tok/s | 205.1 | 375.09 | -45% |
| Olmo-3.1-32B-Think tok/s | 43.67 | 65.71 | -34% |
| Dolphin-Mistral-24B-Venice-Edition tok/s | 62.56 | 93.71 | -33% |
| DeepSeek-Coder-V2-Lite tok/s | 205.42 | 352.52 | -42% |
| DeepSeek-R1-0528-Qwen3-8B tok/s | 145.84 | 226.85 | -36% |
| EVA-Qwen2.5-14B-v0.2 tok/s | 86.48 | 135.83 | -36% |
| Qwen2.5 1.5B LoRA train tok/s | 6349.7 | 15753.9 | -60% |
| Qwen2.5 7B LoRA train tok/s | 3670.8 | 6041.8 | -39% |
| SmolLM2 1.7B LoRA train tok/s | 6938.8 | 16255.4 | -57% |
| TinyLlama 1.1B LoRA train tok/s | 7471 | 18556.4 | -60% |
| gemma-2-2b-it-abliterated tok/s | 232.27 | 416.68 | -44% |
| gemma-2-2b tok/s | 234.68 | 417.07 | -44% |
| gemma-2-9b tok/s | 101.97 | 156.29 | -35% |
| gemma-3-1b tok/s | 310.69 | 567.1 | -45% |
| gemma-3-270m tok/s | 574.24 | 965 | -40% |
| GLM-4.7-Flash tok/s | 119.1 | 215.61 | -45% |
| SmolLM3-3B tok/s | 247.66 | 409.73 | -40% |
| KAT-Coder-V2.5-Dev tok/s | 137.31 | 253.35 | -46% |
| LFM2.5-1.2B tok/s | 572.69 | 989.91 | -42% |
| Llama-2-7B tok/s | 172.53 | 263.84 | -35% |
| Llama-3.2-3B-Instruct-uncensored tok/s | 260.83 | 440.15 | -41% |
| Meta-Llama-3.1-8B tok/s | 156.68 | 237.69 | -34% |
| Phi-4-mini tok/s | 237.38 | 393.81 | -40% |
| Mistral-7B-Instruct-v0.1 tok/s | 166.39 | 252.55 | -34% |
| Mistral-7B-Instruct-v0.2 tok/s | 166.8 | 253.34 | -34% |
| Mistral-7B-Instruct-v0.3 tok/s | 167.06 | 253.2 | -34% |
| Ornith-1.0-35B tok/s | 131.77 | 237.7 | -45% |
| Ornith-1.0-9B tok/s | 126.79 | 201.69 | -37% |
| Phi-3.5-mini tok/s | 222.53 | 362.79 | -39% |
| Qwen-AgentWorld-35B-A3B tok/s | 131.16 | 234.82 | -44% |
| Qwen3-0.6B tok/s | 408.39 | 806.77 | -49% |
| Qwen3-1.7B tok/s | 341.85 | 593.5 | -42% |
| Qwen3-14B tok/s | 90.07 | 139.45 | -35% |
| Qwen3-8B tok/s | 145.96 | 226.88 | -36% |
| Qwen2.5-0.5B tok/s | 517.09 | 905.27 | -43% |
| Qwen2.5-1.5B tok/s | 315.59 | 636.39 | -50% |
| Qwen2.5-14B tok/s | 86.07 | 135.66 | -37% |
| Qwen2.5-32B tok/s | 43.23 | 66 | -35% |
| Qwen2.5-3B tok/s | 240.48 | 398.72 | -40% |
| Qwen2.5-7B tok/s | 159.81 | 256.63 | -38% |
| Qwen2.5-Coder-1.5B tok/s | 322.96 | 650.4 | -50% |
| Qwen2.5-Coder-3B tok/s | 241.51 | 401.24 | -40% |
| AI21-Jamba-Reasoning-3B tok/s | 231.15 | 400.02 | -42% |
| Codestral 22B (Q3_K_M) tok/s | 48.37 | 107.67 | -55% |
| DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored tok/s | 156.62 | 239.45 | -35% |
| DeepSeek-R1 Distill 14B (Q3_K_M) tok/s | 65.33 | 149.81 | -56% |
| DeepSeek-R1-Distill-Qwen-32B-abliterated tok/s | 43.07 | 66.03 | -35% |
| dolphin-2.9-llama3-8b tok/s | 153.34 | 239.29 | -36% |
| Gemma 3 12B (Q3_K_M) tok/s | 69.7 | 153.93 | -55% |
| GLM-4.7-Flash-REAP-23B-A3B tok/s | 109.1 | 199.14 | -45% |
| Josiefied-Qwen3-8B-abliterated-v1 tok/s | 146.8 | 227.15 | -35% |
| Hermes-3-Llama-3.2-3B tok/s | 261 | 439.65 | -41% |
| L3-8B-Stheno-v3.2 tok/s | 156.55 | 239.84 | -35% |
| LFM2.5-8B-A1B tok/s | 351.47 | 617.09 | -43% |
| Mistral-Nemo-Instruct-2407 tok/s | 106.33 | 164.28 | -35% |
| Mistral Small 24B (Q3_K_M) tok/s | 45.34 | 108.88 | -58% |
| NemoMix-Unleashed-12B tok/s | 105.05 | 164.21 | -36% |
| Nemotron-3-Nano-30B-A3B tok/s | 192.44 | 325.43 | -41% |
| Hermes-4-14B tok/s | 89.97 | 139.56 | -36% |
| phi-2 tok/s | 228.1 | 410.12 | -44% |
| Phi-4 14B (Q3_K_M) tok/s | 80.71 | 165.43 | -51% |
| Qwen3-30B-A3B tok/s | 167.1 | 311.76 | -46% |
| Qwen3-4B-Instruct-2507 tok/s | 192.89 | 333.54 | -42% |
| Qwen1.5-0.5B tok/s | 497.73 | 1025.58 | -51% |
| Qwen2-1.5B tok/s | 322.09 | 648.49 | -50% |
| Uncensored tok/s | 85.82 | 135.68 | -37% |
| Qwen2.5-Coder-0.5B tok/s | 542.36 | 910.13 | -40% |
| Qwen2.5-Coder-14B-Instruct-abliterated tok/s | 86.34 | 135.88 | -36% |
| Qwen2.5-Coder 32B (Q3_K_M) tok/s | 31.04 | 75.78 | -59% |
| Qwen2.5-Coder-7B-Instruct-abliterated tok/s | 159.76 | 257.07 | -38% |
| Qwen3 30B A3B (Q3_K_M) tok/s | 142.78 | 309.61 | -54% |
| Qwen3-4B-Instruct-2507 tok/s | 193.39 | 334.71 | -42% |
| Qwen3-4B-Thinking-2507 tok/s | 189.89 | 334.65 | -43% |
| SmolLM2-135M tok/s | 544.06 | 1252.77 | -57% |
| Cydonia-24B-v4.3 tok/s | 62.41 | 93.76 | -33% |
| Qwen2.5 1.5B served serve tok/s | 4111.1 | 5786.7 | -29% |
| Qwen2.5 7B served serve tok/s | 1769.6 | 2282.8 | -22% |
| SmolLM2 1.7B served serve tok/s | 3865.2 | 5642 | -31% |
| TinyLlama 1.1B served serve tok/s | 3221.3 | 7678.7 | -58% |
| Qwen3-Coder-Next-abliterated tok/s | 0 | 210.47 | n/a |
| Laguna-XS-2.1 tok/s | 0 | 0 | n/a |
| Nanbeige4.2-3B tok/s | 0 | 0 | n/a |
| Qwen3-Coder-Next tok/s | 0 | 205.69 | n/a |
| Qwen3-Next-80B-A3B-Thinking tok/s | 0 | 203.29 | n/a |
| Qwen2.5-72B tok/s | 0 | 29.65 | n/a |
| Qwen3-Coder-Next tok/s | 0 | 211.5 | n/a |
| Qwen3-Next-80B-A3B-Thinking tok/s | 0 | 214.04 | n/a |
| Qwen3-Next-80B-A3B tok/s | 0 | 210.17 | n/a |
| DeepSeek-R1-Distill-Llama-70B tok/s | 0 | 32.52 | n/a |
| Hermes-4-70B tok/s | 0 | 32.51 | n/a |
| Llama-3.3-70B-Instruct-abliterated tok/s | 0 | 32.53 | n/a |
| Meta-Llama-3.1-70B tok/s | 0 | 32.52 | n/a |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA A100 40GB SXM4 is faster on 3 of 12; NVIDIA RTX PRO 6000 Blackwell Workstation Edition on 9.
| Workflow | NVIDIA A100 40GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| 50-image depth pass | 10 s | 4 s | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 2.32x faster |
| 24-frame storyboard | 3 min | 3.6 min | NVIDIA A100 40GB SXM4 1.20x faster |
| 500-image masking run | 3.5 min | 2.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.33x faster |
| 60-second AI short film | 4 min | 4.5 min | NVIDIA A100 40GB SXM4 1.14x faster |
| 6-panel comic page | 4.8 min | 5.1 min | NVIDIA A100 40GB SXM4 1.06x faster |
| Character sheet, 12 poses | 6.3 min | 6.1 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.02x faster |
| Full codebase review | 11.7 min | 6.8 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.72x faster |
| 200-product catalogue cutout | 12.2 min | 4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 3.02x faster |
| 40-product photo shoot | 22.2 min | 16.7 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.33x faster |
| 40-product shoot, start to finish | 24.8 min | 17.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.41x faster |
| 100-photo restoration batch | 50.2 min | 33.4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.50x faster |
| 100-photo restore and enlarge | 56 min | 35.3 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.59x faster |
Specifications compared
| NVIDIA A100 40GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | |
|---|---|---|
| VRAM | 40GB | 96GB |
| Architecture | Ampere | Blackwell |
| Memory bandwidth | 1555 GB/s | 1792 GB/s |
| Boost clock | 1,410 MHz | 2,617 MHz |
| TDP | 400 W | 600 W |
| Launch MSRP | $12,000 | $8,565 |
| Release | 2020-05-14 | 2025-03-18 |
FAQ
- Which is better for ai & machine learning: NVIDIA A100 40GB SXM4 or NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA RTX PRO 6000 Blackwell Workstation Edition performs better for ai & machine learning, winning 143 of 145 benchmarks in our suite with an average 72.9% advantage.
- What are the main hardware differences between NVIDIA A100 40GB SXM4 and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA A100 40GB SXM4 has 40GB VRAM and a 400W TDP, while NVIDIA RTX PRO 6000 Blackwell Workstation Edition has 96GB VRAM and a 600W TDP.
- Does VRAM matter more than speed between NVIDIA A100 40GB SXM4 and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- For AI, yes, NVIDIA RTX PRO 6000 Blackwell Workstation Edition fits 13 more of our 12 workloads than NVIDIA A100 40GB SXM4. A model that exceeds VRAM doesn't run slower, it doesn't run at all, so the card that fits the model wins that workload outright.
- Where is the biggest performance difference between NVIDIA A100 40GB SXM4 and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- Swin2SR 4x Upscaler: NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by roughly 206% (17.25 vs 52.83 images/min) in our testing.
NVIDIA A100 40GB SXM4 full review · NVIDIA RTX PRO 6000 Blackwell Workstation Edition full review · All AI & Machine Learning rankings