NVIDIA GeForce RTX 5090 D V2 vs NVIDIA Rubin GPU Comparison

FAQ

Q: What is the process node difference between the two GPUs?

A: The NVIDIA GeForce RTX 5090 D V2 uses a 5 nm process from TSMC, while the NVIDIA Rubin GPU uses a 3 nm process from TSMC. The smaller node contributes to the Rubin GPU's higher transistor density.

Q: How do the memory configurations compare?

A: The RTX 5090 D V2 has 24 GB of GDDR7 memory on a 384-bit bus with 1.34 TB/s bandwidth. The Rubin GPU has 288 GB of HBM4 memory on a 16384-bit bus with 22.1 TB/s bandwidth, which is substantially larger in both capacity and bandwidth.

Q: What is the difference in FP32 compute performance?

A: The RTX 5090 D V2 delivers 104.8 TFLOPS of FP32 performance. The Rubin GPU delivers 130.0 TFLOPS, which is approximately 24% higher in raw single-precision throughput.

Q: Are both GPUs suitable for the same use case?

A: No. The RTX 5090 D V2 is a dual-slot, PCIe 5.0 x16 card with display outputs and a 575 W TDP, designed for client systems. The Rubin GPU is an SXM module with no display outputs, a 2300 W TDP, and a PCIe 6.0 x16 interface, targeting server environments.

Q: What is the release timing for each product?

A: The RTX 5090 D V2 was released on 2025-08-14. The Rubin GPU was released on 2025-12-31, making it a later product in the same year.

Q: How does the shading unit count compare?

A: The RTX 5090 D V2 has 21,760 shading units. The Rubin GPU has 28,672 shading units, which is a higher count by 6,912 units.

Architecture Differences

The two GPUs represent distinct architectural generations with different design priorities. The GeForce RTX 5090 D V2 is built on the Blackwell 2.0 architecture using the GB202 chip, while the Rubin GPU uses the Rubin architecture with the GR100 chip. The process node difference is clear: the RTX 5090 D V2 uses a 5 nm process, while the Rubin GPU uses a 3 nm process, both from TSMC.

Transistor counts diverge sharply. The RTX 5090 D V2 contains 92,200 million transistors on a 750 mm² die, giving a density of 122.9 million transistors per mm². The Rubin GPU contains 336,000 million transistors on a 1456 mm² die, with a density of 230.8 million per mm². The Rubin GPU's die is nearly double the area and has over three times the transistor count.

Memory architecture reflects different target markets. The RTX 5090 D V2 uses 24 GB of GDDR7 on a 384-bit bus with 1.34 TB/s bandwidth. The Rubin GPU uses 288 GB of HBM4 on a 16384-bit bus with 22.1 TB/s bandwidth. The bandwidth advantage for the Rubin GPU is substantial, at over sixteen times that of the RTX 5090 D V2.

Compute resources differ in composition. The RTX 5090 D V2 has 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The Rubin GPU has 28,672 shading units, 896 TMUs, 24 ROPs, no listed RT cores, and 896 tensor cores. The Rubin GPU's ROP count is notably low at 24, which is consistent with a compute-focused accelerator rather than a graphics-oriented product.

The feature set for graphics APIs is absent on the Rubin GPU. The RTX 5090 D V2 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Rubin GPU lists N/A for all three APIs, reinforcing its server-only positioning. The RTX 5090 D V2 provides display outputs (1x HDMI 2.1b and 3x DisplayPort 2.1b), while the Rubin GPU has no outputs.

Clock behavior also differs. The RTX 5090 D V2 has a base clock of 2017 MHz and a boost clock of 2407 MHz. The Rubin GPU has a base clock of 700 MHz and a boost clock of 2267 MHz. The lower base clock on the Rubin GPU suggests a power-conscious idle state, while the boost clock remains competitive.

Power delivery and form factor diverge. The RTX 5090 D V2 is a dual-slot card with a 1x 16-pin power connector, a 575 W TDP, and a suggested PSU of 950 W. The Rubin GPU is an SXM module with no listed power connectors, a 2300 W TDP, and a suggested PSU of 2700 W.

Head-to-Head Benchmarks

The database contains no direct head-to-head benchmark entries for these two products. The RTX 5090 D V2 has a recorded score in the 3DMark Steel Nomad DX12 test of 16,504. The Rubin GPU has no recorded benchmark scores.

The RTX 5090 D V2's score places it at the 59th percentile among all GPUs in the database. Its nearest rivals include the NVIDIA T400 with an average score of 16,508 and a delta of 0%, the AMD Radeon PRO W7500 at 16,415 with a 0.5% delta, the NVIDIA RTX PRO 6000 Blackwell at 16,408 with a 0.6% delta, and the AMD Radeon RX 5700 XT at 16,361 with a 0.9% delta. The data indicates that the RTX 5090 D V2 is essentially tied with the T400 in this specific test, with a performance difference of less than 1% across all four rivals.

The Rubin GPU has an average benchmark score of zero and a 50th percentile ranking, which reflects the absence of recorded measurements in the database. Without any benchmark entries, the database cannot provide a comparative performance assessment for the Rubin GPU against the RTX 5090 D V2 or any other product.

Given the lack of head-to-head data, the specification differences provide the only basis for performance inference. The Rubin GPU's FP32 throughput of 130.0 TFLOPS exceeds the RTX 5090 D V2's 104.8 TFLOPS. The texture rate of the Rubin GPU is 2,031.2 GTexel/s versus 1,636.8 GTexel/s for the RTX 5090 D V2. However, the RTX 5090 D V2 has a much higher pixel rate at 423.6 GPixel/s compared to 54.41 GPixel/s for the Rubin GPU, indicating that rasterization-oriented workloads would favor the GeForce product.

The FP16 performance shows a different ratio. The RTX 5090 D V2 achieves 104.8 TFLOPS with a 1:1 FP16 to FP32 ratio. The Rubin GPU achieves 260.0 TFLOPS with a 2:1 ratio, indicating a stronger relative focus on half-precision compute, which is typical for AI training workloads.

Specification Differences

| Specification | RTX 5090 D V2 | Rubin GPU |

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

| Architecture | Blackwell 2.0 | Rubin |

| Chip | GB202 | GR100 |

| Generation | GeForce 50 | Server Rubin (Rxx) |

| Process Node | 5 nm | 3 nm |

| Transistors | 92,200 million | 336,000 million |

| Die Size | 750 mm² | 1456 mm² |

| Transistor Density | 122.9M / mm² | 230.8M / mm² |

| Base Clock | 2017 MHz | 700 MHz |

| Boost Clock | 2407 MHz | 2267 MHz |

| Memory Clock | 1750 MHz, 28 Gbps effective | 2695 MHz, 10.8 Gbps effective |

| Memory Size | 24 GB | 288 GB |

| Memory Type | GDDR7 | HBM4 |

| Memory Bus Width | 384 bit | 16384 bit |

| Memory Bandwidth | 1.34 TB/s | 22.1 TB/s |

| Shading Units | 21,760 | 28,672 |

| TMUs | 680 | 896 |

| ROPs | 176 | 24 |

| RT Cores | 170 | null |

| Tensor Cores | 680 | 896 |

| Pixel Rate | 423.6 GPixel/s | 54.41 GPixel/s |

| Texture Rate | 1,636.8 GTexel/s | 2,031.2 GTexel/s |

| FP32 | 104.8 TFLOPS | 130.0 TFLOPS |

| FP16 | 104.8 TFLOPS (1:1) | 260.0 TFLOPS (2:1) |

| TDP | 575 W | 2300 W |

| Slot Width | Dual-slot | SXM Module |

| Power Connectors | 1x 16-pin | null |

| Suggested PSU | 950 W | 2700 W |

| Bus Interface | PCIe 5.0 x16 | PCIe 6.0 x16 |

| Display Outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1b | No outputs |

| DirectX | 12 Ultimate (12_2) | N/A |

| OpenGL | 4.6 | N/A |

| Vulkan | 1.4 | N/A |

| Release Date | 2025-08-14 | 2025-12-31 |

| Predecessor | GeForce 40 | Server Blackwell |

| Successor | GeForce 60 | null |

| Launch MSRP | 2,299 USD | null |

| Dimensions | 304 mm x 137 mm x 48 mm | null |

The Verdict

The data indicates two products with separate purposes. The GeForce RTX 5090 D V2 is a client graphics card, evidenced by its dual-slot form factor, display outputs, graphics API support, and the 3DMark benchmark score. The Rubin GPU is a server accelerator, shown by its SXM module format, absence of display outputs, lack of graphics API support, and its 2300 W TDP.

For graphics rendering workloads, the RTX 5090 D V2 is the only viable choice between the two. It has a pixel rate of 423.6 GPixel/s, which is substantially higher than the Rubin GPU's 54.41 GPixel/s. It also supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, while the Rubin GPU lists none of these APIs. The RTX 5090 D V2 also carries a recorded benchmark score of 16,504 in 3DMark Steel Nomad DX12, placing it at the 59th percentile.

For compute-heavy server applications, the Rubin GPU offers larger resources. Its 288 GB of HBM4 memory with 22.1 TB/s bandwidth provides a memory subsystem that is far beyond the RTX 5090 D V2's 24 GB and 1.34 TB/s. Its FP16 performance of 260.0 TFLOPS versus 104.8 TFLOPS suggests a stronger position for half-precision workloads. The Rubin GPU also has 28,672 shading units compared to 21,760, and 896 tensor cores versus 680.

The Rubin GPU's 3 nm process and 336,000 million transistors indicate a more advanced manufacturing approach, but this does not translate to graphics capability. The RTX 5090 D V2 has 176 ROPs versus 24 for the Rubin GPU, and a higher pixel rate. The Rubin GPU's power requirements are also markedly higher, with a 2300 W TDP and a 2700 W suggested PSU, versus 575 W and 950 W for the RTX 5090 D V2.

The RTX 5090 D V2 was released earlier on 2025-08-14, with the Rubin GPU following on 2025-12-31. The RTX 5090 D V2 succeeded the GeForce 40 series and has a successor in the GeForce 60 series. The Rubin GPU succeeded Server Blackwell and has no listed successor.

In summary, the recorded data shows that the RTX 5090 D V2 is the product for graphics output and rasterization performance, while the Rubin GPU is the product for large-scale memory and compute density. There is no overlap in their intended usage domains, and the absence of benchmark data for the Rubin GPU prevents any direct performance ranking. The choice rests on whether the workload requires display output and graphics APIs, which points to the RTX 5090 D V2, or server-scale memory and compute throughput, which points to the Rubin GPU.