NVIDIA GeForce RTX 5090 D vs NVIDIA RTX PRO 5000 Blackwell Comparison

The Verdict

The data presents a clear split between two very different GPUs that share the same underlying chip. The NVIDIA RTX PRO 5000 Blackwell wins zero head-to-head benchmarks against the GeForce RTX 5090 D, but that does not make it the wrong choice. The RTX PRO 5000 sits at the 98th percentile of all GPUs, while the RTX 5090 D sits at the 92nd. The RTX 5090 D leads in every recorded test, with its largest advantage in the 3DMark Steel Nomad DX12 test, where it scores 14,326 versus 9,579.5, a 33.1% lead. In Geekbench OpenCL, the RTX 5090 D scores 310,674 versus 254,116, an 18.2% advantage. In Geekbench Vulkan, the margin is 25%, with scores of 376,915 versus 282,631.

However, the RTX PRO 5000's average benchmark score of 182,109 is far higher than the RTX 5090 D's 77,712. This discrepancy comes from the test suite composition. The RTX PRO 5000 has three recorded results, all from heavy compute and synthetic workloads. The RTX 5090 D has ten recorded results, including several Passmark tests that score low relative to the Geekbench and 3DMark numbers, dragging down its average. The RTX PRO 5000's closest rivals in the database are the A100 SXM4 80 GB, RTX 5000 Ada Generation, RTX 4090 D, and A100 SXM4 40 GB, with delta values between -2.7% and +2.3%. The RTX 5090 D's nearest rivals are mobile and older parts, including the RX 6650M XT and Tesla P100 variants, which is a consequence of the passmark scores compressing its average.

For buyers, the decision hinges on workload. The RTX 5090 D wins every direct comparison, but it consumes 575 W versus 300 W for the RTX PRO 5000. The suggested PSU requirement is 950 W versus 700 W. The RTX PRO 5000 doubles the memory capacity at 48 GB versus 32 GB, and it carries a launch MSRP of 5,099 USD. The RTX 5090 D has a launch MSRP of 2,299 USD. The RTX PRO 5000 is designed for workstation tasks; the RTX 5090 D is a GeForce part. The recorded data shows that for raw speed, the RTX 5090 D wins. For capacity and efficiency, the RTX PRO 5000 holds ground.

Where Each One Wins

The RTX 5090 D wins in every benchmark where both are measured. These three tests are the only direct comparisons available. The 3DMark Steel Nomad DX12 test, which stresses modern game workloads, shows a 33.1% lead for the RTX 5090 D. That is the largest margin in any test. The Geekbench Vulkan test shows a 25% lead, and the Geekbench OpenCL test shows an 18.2% lead. The RTX 5090 D also has a much broader benchmark footprint, with additional PassMark results: DirectX 10 at 231, DirectX 11 at 371, DirectX 12 at 219, DirectX 9 at 434, G2D at 1,487, G3D at 44,065, and GPU Compute at 28,396. These numbers are not available for the RTX PRO 5000, so no comparison can be drawn.

The RTX PRO 5000 wins in areas that are not captured by the recorded benchmarks. Its 48 GB memory is 50% larger than the 32 GB on the RTX 5090 D. Its power draw of 300 W is considerably lower than 575 W. Its suggested PSU of 700 W is 250 W lower. Its dimensions are smaller: 267 mm length versus 304 mm, 111 mm height versus 137 mm, and 40 mm width versus 48 mm. The RTX PRO 5000 was released on 2025-03-17, while the RTX 5090 D launched on 2025-01-29. The RTX PRO 5000 is in the "Blackwell PRO W (x000)" generation, while the RTX 5090 D is in the "GeForce 50" generation.

The data implies the RTX 5090 D is for users who prioritize raw throughput and frame rates. The RTX PRO 5000 is for users who need large memory capacity and lower power consumption in a workstation form factor. The RTX 5090 D successor is listed as "GeForce 60", while the RTX PRO 5000 has no successor listed. The RTX PRO 5000 predecessor is "Workstation Ada", while the RTX 5090 D predecessor is "GeForce 40".

Architecture Differences

Both GPUs use the same GB202 chip, built on TSMC's 5 nm process, with 92,200 million transistors on a 750 mm² die. The transistor density is identical at 122.9M per mm². The architecture name is the same: Blackwell 2.0. The API support is identical: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The memory type is GDDR7 for both, with the same memory clock of 1750 MHz, which is 28 Gbps effective.

The first major divergence is in the execution units. The RTX 5090 D has 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The RTX PRO 5000 has 14,080 shading units, 440 TMUs, 160 ROPs, 110 RT cores, and 440 tensor cores. The RTX 5090 D has 54.5% more shading units, 54.5% more TMUs, 10% more ROPs, 54.5% more RT cores, and 54.5% more tensor cores. This explains the raw performance advantage.

The memory subsystem also differs. The RTX 5090 D has a 512-bit bus with 1.79 TB/s bandwidth. The RTX PRO 5000 has a 384-bit bus with 1.34 TB/s bandwidth. The bandwidth is 33.6% higher on the RTX 5090 D. However, capacity reverses the order: 32 GB on the RTX 5090 D versus 48 GB on the RTX PRO 5000.

Clock speeds are higher on the RTX 5090 D. Its base clock is 2017 MHz versus 1740 MHz, a 16% increase. The boost clock is 2407 MHz versus 2377 MHz, a smaller 1.3% gap. The pixel rate is 423.6 GPixel/s versus 380.3 GPixel/s. The texture rate is 1,636.8 GTexel/s versus 1,045.9 GTexel/s. The FP32 and FP16 throughput is 104.8 TFLOPS on the RTX 5090 D versus 66.94 TFLOPS on the RTX PRO 5000, both at 1:1 ratio.

Power and physical design differ significantly. The RTX 5090 D has a TDP of 575 W, the RTX PRO 5000 has a TDP of 300 W. The RTX 5090 D requires a 950 W PSU, the RTX PRO 5000 a 700 W PSU. Both are dual-slot and use a single 16-pin connector. Both use PCIe 5.0 x16. Display outputs differ: the RTX PRO 5000 has 4x DisplayPort 2.1b, while the RTX 5090 D has 1x HDMI 2.1b and 3x DisplayPort 2.1b.

FAQ

Q: Which GPU has more memory?

A: The RTX PRO 5000 has 48 GB of GDDR7 memory, while the RTX 5090 D has 32 GB. The RTX PRO 5000 uses a 384-bit bus, the RTX 5090 D uses a 512-bit bus.

Q: Which GPU is faster in the recorded benchmarks?

A: The RTX 5090 D wins all three head-to-head tests: 3DMark Steel Nomad DX12 (14,326 vs 9,579.5), Geekbench OpenCL (310,674 vs 254,116), and Geekbench Vulkan (376,915 vs 282,631).

Q: Why does the RTX PRO 5000 have a higher average benchmark score?

A: The RTX PRO 5000 has an average score of 182,109 because its three recorded tests are all high-end synthetic workloads. The RTX 5090 D's average of 77,712 is pulled down by several PassMark scores (like DirectX 9 at 434 and G2D at 1,487) that are much lower than its primary 3D results.

Q: What is the power consumption difference?

A: The RTX PRO 5000 has a TDP of 300 W and a suggested PSU of 700 W. The RTX 5090 D has a TDP of 575 W and a suggested PSU of 950 W.

Q: Do they have the same transistor count?

A: Yes, both use the GB202 chip with 92,200 million transistors on a 750 mm² die, fabricated on TSMC's 5 nm process.

Q: Which GPU has a higher percentile ranking?

A: The RTX PRO 5000 ranks in the 98th percentile of all GPUs, while the RTX 5090 D ranks in the 92nd percentile. This is based on average benchmark scores across the respective recorded suites.

Head-to-Head Benchmarks

The 3DMark Steel Nomad DX12 test provides the clearest performance gap. The RTX 5090 D scores 14,326, which is 33.1% higher than the RTX PRO 5000's 9,579.5. This is a DX12 rendering workload that stresses ray tracing and rasterization. The margin is substantial, and it aligns with the RTX 5090 D's higher shading unit count (21,760 vs 14,080) and higher boost clock (2407 MHz vs 2377 MHz).

In Geekbench Vulkan, the RTX 5090 D scores 376,915 against 282,631 for the RTX PRO 5000, a delta of -25% when viewed from the RTX PRO side. The Geekbench Vulkan test benefits from the RTX 5090 D's 176 ROPs and 170 RT cores, compared to the RTX PRO 5000's 160 ROPs and 110 RT cores. The RTX 5090 D's 512-bit memory bus also helps in memory-bound scenarios.

In Geekbench OpenCL, the RTX 5090 D scores 310,674 against 254,116, a delta of -18.2% for the RTX PRO 5000. This is the smallest margin among the three. OpenCL workloads often scale with compute unit count and memory bandwidth. The RTX 5090 D has a 33.6% bandwidth advantage (1.79 TB/s vs 1.34 TB/s) and a 54.5% advantage in shading units, but the gap is smaller than in Vulkan or 3DMark, suggesting other bottlenecks.

The RTX 5090 D wins all three recorded comparisons, with deltas of -33.1%, -18.2%, and -25%. The RTX PRO 5000 has no recorded wins. However, the RTX PRO 5000's nearest rivals in the database show it is competitive with other workstation parts. Its average score of 182,109 is within 0.9% of the A100 SXM4 80 GB, within 1.4% of the RTX 5000 Ada Generation, and 2.3% ahead of the RTX 4090 D. The RTX 5090 D's nearest rivals are the RX 6650 M XT (1.1% slower), RX 6850 M XT (1.6% faster), Tesla P100 12 GB (2.1% faster), and Tesla P100 16 GB (2.4% faster). This suggests the RTX 5090 D's average is skewed by its low PassMark scores.

Specification Differences

| Specification | RTX PRO 5000 Blackwell | GeForce RTX 5090 D |

|----------------|----------------------|---------------------|

| Generation | Blackwell PRO W (x000) | GeForce 50 |

| Release Date | 2025-03-17 | 2025-01-29 |

| Shading Units | 14,080 | 21,760 |

| TMUs | 440 | 680 |

| ROPs | 160 | 176 |

| RT Cores | 110 | 170 |

| Tensor Cores | 440 | 680 |

| Base Clock | 1740 MHz | 2017 MHz |

| Boost Clock | 2377 MHz | 2407 MHz |

| Memory Size | 48 GB | 32 GB |

| Memory Bus | 384 bit | 512 bit |

| Memory Bandwidth | 1.34 TB/s | 1.79 TB/s |

| Pixel Rate | 380.3 GPixel/s | 423.6 GPixel/s |

| Texture Rate | 1,045.9 GTexel/s | 1,636.8 GTexel/s |

| FP32 / FP16 | 66.94 TFLOPS | 104.8 TFLOPS |

| TDP | 300 W | 575 W |

| Suggested PSU | 700 W | 950 W |

| Display Outputs | 4x DP 2.1b | 1x HDMI 2.1b, 3x DP 2.1b |

| Dimensions | 267 mm x 111 mm x 44 mm | 310 mm x 137 mm x 48 mm |

| Launch MSRP | 5,099 USD | 2,299 USD |

The only fields that are identical are the chip (GB202), architecture (Blackwell 2.0), process node (5 nm), foundry (TSMC), transistor count (92,200 million), die size (750 mm²), memory type (GDDR7), memory clock (1750 MHz), bus interface (PCIe 5.0 x16), DirectX version (12 Ultimate), OpenGL version (4.6), Vulkan version (1.4), slot width (Dual-slot), power connector (1x 16-pin), and production status (Active).