

NVIDIA RTX PRO 6000 Blackwell Server Edition wins 5 of 9 benchmarks, averaging 35.1% faster.
Both cards were measured first-party on our bench, same suite, same test rig.
The gap is widest in FLUX.1 dev, where NVIDIA RTX PRO 6000 Blackwell Server Edition leads by 232% (2.036 vs 6.75 images/min); the closest fight is Qwen2.5-Coder 14B (11% apart); VRAM decides part of this one: NVIDIA RTX PRO 6000 Blackwell Server Edition runs 12 of our 12 AI workloads while the other card runs 7, models that don't fit score zero.
| Benchmark | NVIDIA GeForce RTX 5090 | NVIDIA RTX PRO 6000 Blackwell Server Edition | Difference |
|---|---|---|---|
| Qwen3 4B tok/s | 375.6 | 315.72 | +19% |
| Llama 3.1 8B tok/s | 268.14 | 233.8 | +15% |
| Qwen2.5-Coder 14B tok/s | 149.75 | 135.07 | +11% |
| Qwen3 32B tok/s | 71.15 | 63.9 | +11% |
| Llama 3.3 70B tok/s | 0 | 32.08 | n/a |
| Stable Diffusion XL images/min | 21.08 | 26.32 | -20% |
| Z-Image Turbo images/min | 10.725 | 14.775 | -27% |
| FLUX.1 dev images/min | 2.036 | 6.75 | -70% |
| Qwen-Image-Edit images/min | 0 | 2.62 | n/a |
How long each card takes to finish a complete pipeline, not just one model. NVIDIA GeForce RTX 5090 is faster on 1 of 2; NVIDIA RTX PRO 6000 Blackwell Server Edition on 1.
| Workflow | NVIDIA GeForce RTX 5090 | NVIDIA RTX PRO 6000 Blackwell Server Edition | Difference |
|---|---|---|---|
| Full codebase review | 6.7 min | 7.4 min | NVIDIA GeForce RTX 5090 1.11x faster |
| 24-frame storyboard | 12.5 min | 2.3 min | NVIDIA RTX PRO 6000 Blackwell Server Edition 5.43x faster |
| NVIDIA GeForce RTX 5090 | NVIDIA RTX PRO 6000 Blackwell Server Edition | |
|---|---|---|
| VRAM | 32GB | 96GB |
| Architecture | Blackwell (GB202) | Blackwell |
| Memory bandwidth | 1792 GB/s | 1792 GB/s |
| Boost clock | 2,407 MHz | 2,610 MHz |
| TDP | 575 W | 600 W |
| Launch MSRP | $1,999 | $8,565 |
| Release | 2025-01-30 | 2025-03-18 |
NVIDIA GeForce RTX 5090 full review · NVIDIA RTX PRO 6000 Blackwell Server Edition full review · All AI & Machine Learning rankings