NVIDIA GeForce RTX 5090 D vs NVIDIA Rubin GPU Comparison

Head-to-Head Benchmarks

The recorded data shows no direct head-to-head benchmark comparisons between the NVIDIA GeForce RTX 5090 D and the NVIDIA Rubin GPU. The database lists zero shared tests, zero wins for either side, and no delta percentages between the two. This is primarily due to the Rubin GPU being a server-oriented part with no benchmark entries at all, while the RTX 5090 D has a full suite of ten recorded tests.

For the RTX 5090 D, the strongest result comes from Geekbench Vulkan, where it scores 376,915 points. Its OpenCL result is 310,674. In 3DMark Steel Nomad DX12, the card records 14,326 points. Passmark results show a G3D score of 44,065, a GPU compute score of 28,396, a G2D score of 1,487, and DirectX 9, 10, 11, and 12 scores of 434, 231, 371, and 219 respectively. The average benchmark score across all tests is 77,712, placing it in the 92nd percentile among all GPUs.

The Rubin GPU has no recorded benchmarks, an average score of 0, and sits in the 50th percentile. Its nearest rivals list is empty. Therefore, any performance comparison must rely on the raw specification data and the RTX 5090 D's peer comparisons, not on direct measurements.

The RTX 5090 D's nearest rivals in the database include the AMD Radeon RX 6650M XT, which averages 76,904 points and trails by 1.1%, and the AMD Radeon RX 6850M XT, which averages 78,940 and leads by 1.6%. NVIDIA's own Tesla P100 PCIe 12 GB and 16 GB variants score 79,396 and 79,605, leading by 2.1% and 2.4% respectively. This indicates the RTX 5090 D sits in a tight performance band around these mobile and data-center parts, with differences under 3% in either direction. The Rubin GPU, by contrast, has no comparative data to anchor its performance.

FAQ

Q: Why does the Rubin GPU show no benchmark scores?

A: The database records no benchmark entries for the Rubin GPU. Its average benchmark score is 0, and its percentile versus all GPUs is 50, which reflects the absence of measured data rather than a performance evaluation.

Q: What is the RTX 5090 D's strongest recorded benchmark?

A: The highest single score is 376,915 in Geekbench Vulkan. The OpenCL score of 310,674 is the second highest, followed by the Passmark G3D score of 44,065.

Q: How does the RTX 5090 D compare to its nearest rivals?

A: It trails the AMD Radeon RX 6850M XT by 1.6%, the Tesla P100 PCIe 12 GB by 2.1%, and the Tesla P100 PCIe 16 GB by 2.4%, while leading the Radeon RX 6650M XT by 1.1%. All deltas are within 1.1% to 2.4%.

Q: Does the Rubin GPU support standard graphics APIs like DirectX or Vulkan?

A: No. The Rubin GPU lists DirectX, OpenGL, and Vulkan as N/A. The RTX 5090 D supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What are the memory configurations of each card?

A: The RTX 5090 D has 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The Rubin GPU has 288 GB of HBM4 on a 16,384-bit bus with 22.1 TB/s bandwidth.

Q: Which card has higher FP32 compute throughput?

A: The Rubin GPU records 130.0 TFLOPS FP32 versus 104.8 TFLOPS for the RTX 5090 D. In FP16, the Rubin GPU reaches 260.0 TFLOPS (2:1 ratio), while the RTX 5090 D delivers 104.8 TFLOPS (1:1 ratio).

Where Each One Wins

The RTX 509 D wins in pixel throughput. Its pixel rate is 423.6 GPixel/s, compared to 54.41 GPixel/s for the Rubin GPU. This reflects the RTX 5090 D's 176 render output units versus only 24 on the Rubin GPU. For rasterization-focused workloads, particularly at high resolutions with heavy fill-rate demands, the RTX 5090 D has a decisive edge.

The Rubin GPU wins in raw compute and memory bandwidth. Its FP32 output of 130.0 TFLOPS is 24% higher than the RTX 5090 D's 104.8 TFLOPS. Its FP16 output of 260.0 TFLOPS is 2.5 times the RTX 5090 D's 104.8 TFLOPS, though the Rubin GPU achieves this at a 2:1 ratio, meaning it processes half-rate FP16. The Rubin GPU's 22.1 TB/s memory bandwidth is over 12 times the RTX 5090 D's 1.79 TB/s, and its 288 GB capacity is 9 times larger. Texture rate also favors the Rubin GPU: 2,031.2 GTexel/s versus 1,636.8 GTexel/s.

The RTX 5090 D wins on compatibility and display support. It has 1x HDMI 2.1b and 3x DisplayPort 2.1b outputs, while the Rubin GPU has no display outputs. The RTX 5090 D also supports PCIe 5.0 x16, whereas the Rubin GPU uses PCIe 6.0 x16, which is newer but not yet widely adopted in consumer systems.

The Rubin GPU wins in transistor count and density. It packs 336,000 million transistors on a 1,456 mm² die at 230.8 million transistors per mm², versus 92,200 million on 750 mm² at 122.9M per mm² for the RTX 5090 D. This indicates a more complex, larger-scale silicon design.

Specification Differences

The process nodes differ: the RTX 5090 D uses a 5 nm TSMC node, while the Rubin GPU uses a 3 nm TSMC node. Both come from TSMC, but the smaller node allows the Rubin GPU to reach higher transistor density.

Die size is drastically different. The RTX 5090 D measures 750 mm², while the Rubin GPU measures 1,456 mm², nearly double. Transistor counts follow: 92,200 million versus 336,000 million.

Clock speeds favor the RTX 5090 D in base clock (2017 MHz versus 700 MHz) but the Rubin GPU has a higher boost clock (2267 MHz versus 2407 MHz). Memory clocks also differ: the RTX 5090 D runs at 1750 MHz with 28 Gbps effective, while the Rubin GPU runs at 2695 MHz with 10.8 Gbps effective. The effective rate is misleading here because the Rubin GPU's HBM4 uses a far wider bus.

Shading units: 21,760 for the RTX 5090 D versus 28,672 for the Rubin GPU. Texture mapping units: 680 versus 896. Render output units: 176 versus 24. Tensor cores: 680 versus 896. The RTX 5090 D has 170 RT cores; the Rubin GPU lists none.

Power specifications show a stark contrast. The RTX 5090 D has a TDP of 575 W with a suggested PSU of 950 W and a single 16-pin connector. The Rubin GPU has a TDP of 2300 W with a suggested PSU of 2700 W and no listed power connectors, fitting an SXM module form factor instead of a dual-slot card.

Physical dimensions exist only for the RTX 5090 D: 304 mm length, 137 mm height, and 48 mm width. The Rubin GPU has no recorded dimensions.

Release dates differ: the RTX 5090 D launched on 2025-01-29, while the Rubin GPU is dated 2025-12-31. The RTX 5090 D has a launch MSRP of 2,299 USD; the Rubin GPU has no launch MSRP recorded.

Architecture Differences

The RTX 5090 D uses the GB202 chip with Blackwell 2.0 architecture, belonging to the GeForce 50-series. The Rubin GPU uses the GR100 chip with Rubin architecture, belonging to the Server Rubin (Rxx) generation. These are separate architectural lines.

The RTX 5090 D's predecessor is the GeForce 40 series, and its successor is the GeForce 60 series. The Rubin GPU's predecessor is Server Blackwell, with no successor listed. This indicates the Rubin GPU is a new server-class architecture, not a direct consumer successor.

Memory types differ fundamentally. The RTX 5090 D uses GDDR7 on a 512-bit bus, while the Rubin GPU uses HBM4 on a 16,384-bit bus. This explains the massive bandwidth gap: 1.79 TB/s versus 22.1 TB/s.

The FP16 implementation differs. The RTX 5090 D runs FP16 at a 1:1 ratio with FP32, meaning both are 104.8 TFLOPS. The Rubin GPU runs FP16 at a 2:1 ratio, giving 260.0 TFLOPS versus 130.0 TFLOPS FP32. This suggests the Rubin GPU is optimized for mixed-precision workloads common in AI training, where FP16 throughput matters more than FP32.

The RTX 5090 D has dedicated RT cores (170) and full graphics API support. The Rubin GPU has no RT cores listed and no graphics API support, confirming its server-oriented role. Its display outputs are absent, and its slot width is an SXM module, indicating a data-center form factor.

The process node difference (5 nm versus 3 nm) combined with the transistor count difference (92.2 billion versus 336 billion) points to the Rubin GPU being a much larger, more advanced chip. The transistor density of 230.8M per mm² on the Rubin GPU versus 122.9M per mm² on the RTX 5090 D shows the 3 nm node packs more transistors per area.

The power difference is notable: 575 W versus 2300 W. This is not a linear scaling of performance, as the Rubin GPU's FP32 is only 24% higher, but its memory bandwidth and capacity are far larger, which drives the power requirement.

The Verdict

The data shows two products with fundamentally different purposes. The NVIDIA GeForce RTX 5090 D is a consumer graphics card with display outputs, full DirectX 12 Ultimate support, and a 92nd percentile ranking among all GPUs. It delivers 104.8 TFLOPS FP32, 32 GB of memory, and a pixel rate of 423.6 GPixel/s, making it suitable for real-time rendering and gaming workloads.

The NVIDIA Rubin GPU is a server compute module with no display outputs, no graphics API support, and no benchmark scores. Its strengths lie in memory capacity (288 GB), bandwidth (22.1 TB/s), and FP16 throughput (260.0 TFLOPS), which are typical priorities for large-scale compute and AI training tasks.

For users seeking a desktop graphics card, the RTX 5090 D is the only viable choice from this pair. It offers standard connectivity, a dual-slot form factor, and a 575 W power profile with a 950 W suggested PSU. Its release date of 2025-01-29 is earlier than the Rubin GPU's 2025-12-31 date, and it has an established benchmark record.

For users needing extreme memory bandwidth and capacity for server workloads, the Rubin GPU is the data-driven pick. Its 22.1 TB/s bandwidth and 288 GB capacity dwarf the RTX 5090 D's 1.79 TB/s and 32 GB. However, its 2300 W TDP and 2700 W suggested PSU require substantial infrastructure, and its lack of display outputs and graphics APIs limits it to headless compute.

The RTX 5090 D's nearest rivals are all within 2.4% of its average score, so it does not dominate its immediate competition. The Rubin GPU has no rivals in the database, leaving its relative position undefined. The choice between these two comes down to workload: the RTX 5090 D for graphics and rendering, the Rubin GPU for memory-bound compute. The recorded data does not support using either as a substitute for the other.