NVIDIA GB10 vs NVIDIA GeForce RTX 5090 Comparison

The Verdict

The NVIDIA GeForce RTX 5090 is the clear performance winner in every recorded benchmark comparison. It beats the GB10 by 64.1% in Geekbench OpenCL and by 69.6% in Geekbench Vulkan. If your workload demands maximum compute throughput and you have the power delivery and physical space for a dual-slot card, the RTX 5090 is the obvious choice. The GB10, however, is not without merit: it carries 128 GB of memory versus 32 GB, draws only 140 W versus 575 W, and fits in an integrated graphics package form factor. For memory-bound tasks or constrained builds, the GB10 makes sense. For raw speed in general-purpose compute, the data points firmly to the RTX 5090.

Architecture Differences

Both parts use the same Blackwell 2.0 architecture and are built on a 5 nm process at TSMC. The underlying chip designs diverge significantly. The GB10 uses the GB20B die, while the RTX 5090 uses the GB202 die. The GB202 is substantially larger at 750 mm² compared to 382 mm², and it packs 92,200 million transistors versus an unknown count for the GB10. Transistor density for the GB202 is recorded at 122.9M per mm²; the GB10's density is not listed.

The RTX 5090 has far more execution resources: 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The GB10 offers 6,144 shading units, 384 TMUs, 48 ROPs, 48 RT cores, and 384 tensor cores. The RTX 5090 also clocks higher, with a base of 2017 MHz and boost of 2407 MHz, while the GB10 runs at 1665 MHz base and 2418 MHz boost. Memory architecture differs completely: the GB10 uses 128 GB of LPDDR5X on a 256-bit bus with 273.2 GB/s bandwidth, while the RTX 5090 uses 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The GB10's memory clock is 1067 MHz (8.5 Gbps effective), versus 1750 MHz (28 Gbps effective) for the RTX 5090.

API support also separates them. The GB10 reports N/A for DirectX, OpenGL, and Vulkan, indicating it is not designed for conventional graphics API workloads. The RTX 5090 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GB10 has a single HDMI output, while the RTX 5090 provides one HDMI 2.1b and three DisplayPort 2.1b outputs. Power requirements differ sharply: the GB10 needs a 300 W suggested PSU and has no power connectors, while the RTX 5090 requires a 950 W suggested PSU and uses a single 16-pin connector.

Head-to-Head Benchmarks

The database records two direct comparisons between these parts. In Geekbench OpenCL, the RTX 5090 scores 334,370 against the GB10's 120,137. That is a 64.1% advantage for the RTX 5090. In Geekbench Vulkan, the gap widens: the RTX 5090 scores 376,728 versus 114,648, a 69.6% lead. The RTX 5090 wins both recorded tests, giving it a 2-0 head-to-head record.

These results align with the resource disparity. The RTX 5090 has roughly 3.5 times the shading units, 1.8 times the TMUs, and 3.7 times the ROPs of the GB10. Its FP32 throughput is 104.8 TFLOPS versus 29.71 TFLOPS, a factor of about 3.5. The texture rate is 1,636.8 GTexel/s versus 928.5 GTexel/s, and pixel rate is 423.6 GPixel/s versus 116.1 GPixel/s. The RTX 5090's memory bandwidth of 1.79 TB/s dwarfs the GB10's 273.2 GB/s by a factor of 6.5.

However, the GB10 is not without competitive standing in its own segment. Its average benchmark score of 117,393 places it at the 95th percentile among all GPUs in the database. It edges the NVIDIA RTX 4000 SFF Ada Generation by 0.3%, trails the AMD Radeon PRO W7700 by 1.3%, and beats the NVIDIA Tesla V100 SXM2 16 GB by 2.6% and the NVIDIA RTX A5500 Mobile by 3%. The RTX 5090, despite its higher raw scores, sits at the 92nd percentile with an average score of 79,842, which is only 0.3% ahead of the Tesla P100 PCIe 16 GB and 0.6% ahead of the Tesla P100 PCIe 12 GB, while trailing the AMD Radeon Pro Vega 64X by 1.4%. The percentile difference reflects that the RTX 5090's average is pulled down by its PassMark DirectX 9 and DirectX 10 results, which are low relative to its other scores.

FAQ

Q: Which GPU has more memory?

A: The GB10 has 128 GB of LPDDR5X, while the RTX 5090 has 32 GB of GDDR7.

Q: What is the power draw difference?

A: The GB10 has a TDP of 140 W with a suggested PSU of 300 W, while the RTX 5090 has a TDP of 575 W with a suggested PSU of 950 W.

Q: Does the GB10 support DirectX?

A: No. The GB10 lists N/A for DirectX, OpenGL, and Vulkan, while the RTX 5090 supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4.

Q: Which card is larger?

A: The GB10 measures 150 mm by 51 mm by 150 mm and is an integrated graphics package, while the RTX 5090 measures 304 mm by 137 mm by 40 mm and is dual-slot.

Q: How do their average benchmark scores compare?

A: The GB10 averages 117,393 across its recorded benchmarks, while the RTX 5090 averages 79,842. The GB10 sits at the 95th percentile, the RTX 5090 at the 92nd.

Q: Which card has a higher boost clock?

A: The GB10 boosts to 2418 MHz, slightly higher than the RTX 5090's 2407 MHz, though the RTX 5090 has a much higher base clock at 2017 MHz versus 1665 MHz.

Where Each One Wins

The RTX 5090 wins decisively in compute-heavy tasks. Its FP32 throughput of 104.8 TFLOPS is 3.5 times the GB10's 29.71 TFLOPS. Texture rate is 1,636.8 GTexel/s versus 928.5 GTexel/s, and pixel rate is 423.6 GPixel/s versus 116.1 GPixel/s. The memory bandwidth advantage is even larger: 1.79 TB/s versus 273.2 GB/s. In Geekbench OpenCL and Vulkan, the RTX 5090 leads by 64.1% and 69.6% respectively. For rendering, simulation, or any workload that scales with shader count and memory bandwidth, the RTX 5090 is the stronger part.

The GB10 wins on capacity and efficiency. It offers 128 GB of memory, four times the RTX 5090's 32 GB. Its 140 W TDP is less than a quarter of the 575 W TDP of the RTX 5090, and it requires no power connectors and a 300 W PSU versus a 16-pin connector and 950 W PSU. The GB10 is an integrated graphics package measuring 150 mm by 51 mm by 150 mm, while the RTX 5090 is a 304 mm by 137 mm by 40 mm dual-slot card. For memory-hungry inference workloads, large dataset processing, or builds with strict power and space limits, the GB10 is the practical choice.

Specification Differences

| Specification | NVIDIA GB10 | NVIDIA GeForce RTX 5090 |

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

| Chip | GB20B | GB202 |

| Die Size | 382 mm² | 750 mm² |

| Transistors | Unknown | 92,200 million |

| Transistor Density | N/A | 122.9M / mm² |

| Base Clock | 1665 MHz | 2017 MHz |

| Boost Clock | 2418 MHz | 2407 MHz |

| Memory Clock | 1067 MHz, 8.5 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 128 GB | 32 GB |

| Memory Type | LPDDR5X | GDDR7 |

| Memory Bus | 256 bit | 512 bit |

| Memory Bandwidth | 273.2 GB/s | 1.79 TB/s |

| Shading Units | 6,144 | 21,760 |

| TMUs | 384 | 680 |

| ROPs | 48 | 176 |

| RT Cores | 48 | 170 |

| Tensor Cores | 384 | 680 |

| Pixel Rate | 116.1 GPixel/s | 423.6 GPixel/s |

| Texture Rate | 928.5 GTexel/s | 1,636.8 GTexel/s |

| FP32 | 29.71 TFLOPS | 104.8 TFLOPS |

| FP16 | 29.71 TFLOPS (1:1) | 104.8 TFLOPS (1:1) |

| TDP | 140 W | 575 W |

| Slot Width | IGP | Dual-slot |

| Power Connectors | None | 1x 16-pin |

| Suggested PSU | 300 W | 950 W |

| Display Outputs | 1x HDMI | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

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

| OpenGL | N/A | 4.6 |

| Vulkan | N/A | 1.4 |

| Dimensions | 150 mm x 51 mm x 150 mm | 304 mm x 137 mm x 40 mm |

| Release Date | 2025-10-14 | 2025-01-29 |