NVIDIA GB10 vs NVIDIA GeForce RTX 5090 D Comparison

The Verdict

The data presents two very different NVIDIA products that happen to share the same Blackwell 2.0 architecture. The NVIDIA GB10 is a server-oriented part, launched in October 2025, with an average benchmark score of 117,393 across its recorded tests. The NVIDIA GeForce RTX 5090 D, launched earlier in January 2025, posts an average benchmark score of 77,112, but this figure is pulled down by a set of older DirectX and 2D tests that are not representative of its raw compute capabilities. In the two directly comparable workloads, the GeForce RTX 5090 D wins both decisively. For users focused on the recorded OpenCL and Vulkan data, the RTX 5090 D is the clear performance leader. The GB10, however, offers a much lower power envelope and a significantly smaller physical footprint, making it suitable for dense server environments. The verdict from the data is straightforward: if maximum raw throughput in the tested compute APIs is the priority, the GeForce RTX 5090 D wins outright. If the priority is a compact, low-power integrated graphics processor for server deployment, the GB10 is the only option that fits that profile. The RTX 5090 D also carries a launch MSRP of 2,299 USD, while the GB10 has a launch MSRP of 3,999 USD.

Architecture Differences

Both parts are built on the Blackwell 2.0 architecture using a 5 nm process at TSMC, but the similarities end there. The GB10 uses the GB20B chip with a die size of 382 mm², while the RTX 5090 D uses the GB202 chip, which is significantly larger at 750 mm². The transistor count for the GB10 is listed as unknown, whereas the RTX 5090 D packs 92,200 million transistors with a transistor density of 122.9M per mm². The GB10 is classified under the Server Blackwell (Bxx) generation, while the RTX 5090 D belongs to the GeForce 50-series. In terms of shading units, the GB10 has 6,144, while the RTX 5090 D has 21,760, roughly 3.5 times as many. Texture mapping units stand at 384 for the GB10 versus 680 for the RTX 5090 D, and render output units are 48 versus 176. Ray tracing cores are 48 on the GB10 and 170 on the RTX 5090 D. Tensor cores follow a similar pattern: 384 on the GB10, 680 on the RTX 5090 D. The GB10 has no API support listed for DirectX, OpenGL, or Vulkan, whereas the RTX 5090 D supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. This indicates the GB10 is not intended for conventional graphics rendering workloads, while the RTX 5090 D is a fully featured consumer graphics card.

Where Each One Wins

The RTX 5090 D wins in every head-to-head benchmark recorded. In Geekbench OpenCL, it scores 310,674 against the GB10's 120,137, a delta of -61.3% from the GB10's perspective, meaning the RTX 5090 D is roughly 158% faster. In Geekbench Vulkan, the RTX 5090 D scores 376,915 against 114,648, a delta of -69.6%, meaning it is about 229% faster. The GB10 has no wins in the recorded head-to-head data. However, the GB10's advantages are architectural and physical. It has a TDP of 140 W versus 575 W for the RTX 5090 D, a critical difference for server power budgets. The GB10 is an integrated graphics processor with no power connectors and a suggested PSU of 300 W, while the RTX 5090 D is a dual-slot card requiring a 1x 16-pin connector and a 950 W suggested PSU. The GB10 measures 150 mm in length, 51 mm in height, and 150 mm in width, while the RTX 5090 D measures 304 mm in length, 137 mm in height, and 48 mm in width. The GB10 also offers 128 GB of LPDDR5X memory versus 32 GB of GDDR7 on the RTX 5090 D, a fourfold capacity advantage. The GB10's memory bandwidth is 273.2 GB/s, while the RTX 5090 D reaches 1.79 TB/s. For workloads that require massive memory capacity in a low-power, compact form factor, the GB10 is the appropriate choice. For raw compute throughput and graphics performance, the RTX 5090 D is the winner.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The GB10 has an average benchmark score of 117,393, while the RTX 5090 D has an average benchmark score of 77,112. However, the RTX 5090 D's average is affected by a broader set of tests, including older DirectX and 2D workloads, while the GB10 only has two recorded benchmarks.

Q: How does the GB10 compare to the RTX 5090 D in OpenCL performance?

A: In Geekbench OpenCL, the RTX 5090 D scores 310,674 versus 120,137 for the GB10, which is a 61.3% deficit for the GB10. The RTX 5090 D is the clear winner in this test.

Q: What is the memory capacity difference between the two?

A: The GB10 has 128 GB of LPDDR5X memory on a 256-bit bus, while the RTX 5090 D has 32 GB of GDDR7 memory on a 512-bit bus. The GB10 offers four times the capacity, but the RTX 5090 D has much higher bandwidth at 1.79 TB/s versus 273.2 GB/s.

Q: Are both GPUs based on the same architecture?

A: Yes, both the GB10 and the RTX 5090 D are based on the Blackwell 2.0 architecture and are manufactured on a 5 nm process at TSMC. They differ in chip size, with the GB10 using GB20B at 382 mm² and the RTX 5090 D using GB202 at 750 mm².

Q: Which GPU is more suitable for a server environment based on power and size?

A: The GB10 is designed as an integrated graphics processor with a 140 W TDP, no power connectors, and a 300 W suggested PSU. It is 150 mm long and 51 mm high. The RTX 5090 D, in contrast, has a 575 W TDP, requires a 16-pin connector, and a 950 W PSU, with dimensions of 304 mm by 137 mm. The GB10 is the more server-friendly option.

Q: Does the RTX 5090 D support modern graphics APIs?

A: Yes, the RTX 5090 D supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GB10 has no API support listed for DirectX, OpenGL, or Vulkan, indicating it is not intended for standard graphics applications.

Head-to-Head Benchmarks

The recorded head-to-head data contains only two tests, and the RTX 5090 D wins both. The first is Geekbench OpenCL. The GB10 scores 120,137, which places it in the 95th percentile of all GPUs in the database. Its nearest rivals include the NVIDIA RTX 4000 SFF Ada Generation at 117,088 (0.3% behind), the AMD Radeon PRO W7700 at 118,976 (1.3% ahead), the NVIDIA Tesla V100 SXM2 16 GB at 114,395 (2.6% behind), and the NVIDIA RTX A5500 Mobile at 113,944 (3% behind). The RTX 5090 D scores 310,674 in the same test, which is a 61.3% margin of victory over the GB10. This is not a close contest. The GB10's OpenCL score is competitive with its nearest rivals, but the RTX 5090 D is in a completely different performance tier.

The second head-to-head test is Geekbench Vulkan. The GB10 scores 114,648, while the RTX 5090 D scores 376,915. The delta is -69.6% for the GB10, meaning the RTX 5090 D is more than three times faster in this workload. The RTX 5090 D's Vulkan score is its highest recorded benchmark, surpassing even its OpenCL result. The GB10's Vulkan score is slightly lower than its OpenCL score, indicating a modest performance difference between the two APIs on that part. The RTX 5090 D also has a 3DMark Steel Nomad DX12 score of 14,326 and a Passmark G3D score of 44,065, but no comparable data exists for the GB10 in those tests. The RTX 5090 D's average benchmark score of 77,112 is significantly lower than the GB10's 117,393, but this is misleading: the RTX 5090 D's average includes Passmark DirectX 9, 10, 11, and 12 scores ranging from 219 to 434, as well as a Passmark G2D score of 1,487, which are not representative of its compute capabilities. When isolating the comparable compute tests, the RTX 5090 D is decisively ahead. The percentile data supports this: the GB10 sits in the 95th percentile of all GPUs, while the RTX 5090 D sits in the 92nd percentile, but the RTX 5090 D's percentile is dragged down by its inclusion of legacy graphics tests. In the two compute workloads that both parts share, the RTX 5090 D outperforms the GB10 by a wide margin.

Specification Differences

The clock speeds differ notably. The GB10 has a base clock of 1665 MHz and a boost of 2418 MHz, while the RTX 5090 D has a base of 2017 MHz and a boost of 2407 MHz. The RTX 5090 D has a higher base clock but a slightly lower boost clock. Memory clocks are 1067 MHz (8.5 Gbps effective) for the GB10 and 1750 MHz (28 Gbps effective) for the RTX 5090 D. Memory configuration differs substantially: the GB10 uses 128 GB of LPDDR5X on a 256-bit bus, while the RTX 5090 D uses 32 GB of GDDR7 on a 512-bit bus. Bandwidth is 273.2 GB/s versus 1.79 TB/s. Pixel rate is 116.1 GPixel/s for the GB10 and 423.6 GPixel/s for the RTX 5090 D. Texture rate is 928.5 GTexel/s versus 1,636.8 GTexel/s. FP32 and FP16 compute are 29.71 TFLOPS for the GB10 and 104.8 TFLOPS for the RTX 5090 D, both at a 1:1 ratio for FP16. TDP is 140 W for the GB10 and 575 W for the RTX 5090 D. The GB10 is an IGP with no power connectors, while the RTX 5090 D is a dual-slot card with a 1x 16-pin connector. Suggested PSU is 300 W versus 950 W. Display outputs are 1x HDMI for the GB10 and 1x HDMI 2.1b plus 3x DisplayPort 2.1b for the RTX 5090 D. Dimensions: the GB10 is 150 mm by 51 mm by 150 mm, while the RTX 5090 D is 304 mm by 137 mm by 48 mm. The GB10 has no listed API support, while the RTX 5090 D supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The GB10's predecessor is Server Hopper and successor is Server Rubin; the RTX 5090 D's predecessor is GeForce 40 and successor is GeForce 60.