NVIDIA B300 SXM6 AC vs NVIDIA GeForce RTX 5090 D Comparison

The NVIDIA B300 SXM6 AC and the NVIDIA GeForce RTX 5090 D represent two distinct extremes of NVIDIA’s Blackwell generation. One is a compute-focused accelerator for server infrastructure, while the other is a graphics card for desktop workstations and gaming. The database shows a clear divergence in their performance profiles, memory configurations, and physical requirements.

Head-to-Head Benchmarks

The only directly comparable benchmark recorded in the database is Geekbench OpenCL, and the results favor the NVIDIA B300 SXM6 AC significantly. The B300 SXM6 AC scores 369,831 points, while the RTX 5090 D scores 310,674 points. This gives the B300 SXM6 AC a 19% lead over the RTX 5090 D in this compute-oriented test.

This 19% margin is substantial, especially considering that the RTX 5090 D has a higher raw FP32 throughput. The RTX 5090 D is rated at 104.8 TFLOPS for FP32 operations, while the B300 SXM6 AC is rated at 76.99 TFLOPS. The fact that the B300 SXM6 AC still outperforms the RTX 5090 D in OpenCL by 19% suggests that the server chip’s massive memory bandwidth and capacity provide a decisive advantage in workloads that are not purely limited by shader compute.

The B300 SXM6 AC’s lead is consistent with its position in the broader database. It sits in the 100th percentile of all GPUs, meaning no other recorded GPU scores higher. Its nearest rivals include the NVIDIA B200, which it beats by 7%, and the NVIDIA H200 NVL, which it beats by 10.4%. It also outperforms the AMD Instinct MI300X by 16.3% and the NVIDIA L40S by 25%. These deltas show that the B300 SXM6 AC is not merely faster than the RTX 5090 D; it is also at the absolute top of the server GPU hierarchy.

The RTX 5090 D, by contrast, sits in the 92nd percentile of all GPUs. Its average benchmark score across all recorded tests is 77,712 points, a figure that is heavily influenced by its diverse set of benchmarks. Its nearest rivals in the database are consumer and workstation GPUs, including the AMD Radeon RX 6650M XT, which trails it by just 1.1%, and the AMD Radeon RX 6850M XT, which beats it by 1.6%. It also trades places with NVIDIA Tesla P100 variants, with the RTX 5090 D leading the 12 GB Tesla P100 by 2.1% and the 16 GB Tesla P100 by 2.4%. These narrow margins indicate that the RTX 5090 D’s average score is closely grouped with other mid-to-high-end GPUs, unlike the B300 SXM6 AC, which stands alone at the top.

The RTX 5090 D does have one notable advantage in the recorded benchmark suite: it has a much wider range of tests. It has scores for 3DMark Steel Nomad DX12 (14,326 points), Passmark DirectX 10 (231 points), DirectX 11 (371 points), DirectX 12 (219 points), DirectX 9 (434 points), Passmark G2D (1,487 points), Passmark G3D (44,065 points), Passmark GPU Compute (28,396 points), and Geekbench Vulkan (376,915 points). The B300 SXM6 AC only has a single recorded benchmark, Geekbench OpenCL. This means the RTX 5090 D is the only one of the two with verified DirectX and Vulkan performance data, which is expected given its consumer-oriented feature set.

FAQ

Q: Which GPU has the higher recorded Geekbench OpenCL score?

A: The NVIDIA B300 SXM6 AC scores 369,831 points, which is 19% higher than the RTX 5090 D’s 310,674 points in the same test.

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

A: The RTX 5090 D’s nearest rivals are the AMD Radeon RX 6650M XT, RX 6850M XT, and two Tesla P100 variants. It leads the RX 6650M XT by 1.1%, trails the RX 6850M XT by 1.6%, leads the Tesla P100 12 GB by 2.1%, and leads the Tesla P100 16 GB by 2.4%.

Q: What is the B300 SXM6 AC’s standing among all GPUs in the database?

A: The B300 SXM6 AC is in the 100th percentile of all GPUs, meaning no recorded GPU has a higher average benchmark score. It leads the NVIDIA B200 by 7%, the H200 NVL by 10.4%, the AMD Instinct MI300X by 16.3%, and the L40S by 25%.

Q: Does the RTX 5090 D have any benchmark results that the B300 SXM6 AC lacks?

A: Yes. The RTX 5090 D has recorded scores for 3DMark Steel Nomad DX12, Passmark DirectX 9, 10, 11, and 12, Passmark G2D, Passmark G3D, Passmark GPU Compute, and Geekbench Vulkan. The B300 SXM6 AC only has a Geekbench OpenCL result.

Q: Which GPU has the higher FP32 compute rating?

A: The RTX 5090 D is rated at 104.8 TFLOPS for FP32, while the B300 SXM6 AC is rated at 76.99 TFLOPS. Despite this, the B300 SXM6 AC wins the recorded OpenCL benchmark by 19%.

Q: What is the launch MSRP of the RTX 5090 D?

A: The RTX 5090 D has a launch MSRP of 2,299 USD. The B300 SXM6 AC has no recorded launch MSRP.

Architecture Differences

The two GPUs share a 5 nm process node and TSMC as the foundry, but they diverge sharply in every other architectural aspect. The B300 SXM6 AC is built on the Blackwell Ultra architecture with the GB110 chip, while the RTX 5090 D uses the Blackwell 2.0 architecture with the GB202 chip. The B300 SXM6 AC belongs to the Server Blackwell (Bxx) generation, whereas the RTX 5090 D is part of the GeForce 50 series.

The transistor counts reveal the scale of the server chip. The B300 SXM6 AC packs 208,000 million transistors on a 1628 mm² die, giving it a transistor density of 127.8 million transistors per mm². The RTX 5090 D has 92,200 million transistors on a 750 mm² die, with a density of 122.9 million per mm². The B300 SXM6 AC’s die is more than twice the size of the RTX 5090 D’s, and it holds more than twice the transistors.

Memory architecture is the most dramatic difference. The B300 SXM6 AC uses 288 GB of HBM3e memory on an 8192-bit bus, delivering 8.19 TB/s of bandwidth. The RTX 5090 D uses 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The B300 SXM6 AC has nine times the memory capacity and roughly 4.5 times the memory bandwidth of the RTX 5090 D.

The compute unit counts also differ. The RTX 5090 D has 21,760 shading units, 680 TMUs, and 176 ROPs, while the B300 SXM6 AC has 18,944 shading units, 592 TMUs, and only 24 ROPs. The RTX 5090 D also has 170 dedicated RT cores, while the B300 SXM6 AC has no recorded RT core count. Both have tensor cores, with the RTX 5090 D carrying 680 and the B300 SXM6 AC carrying 592. The B300 SXM6 AC’s low ROP count and lack of RT cores reinforce its role as a compute accelerator rather than a graphics card.

The B300 SXM6 AC uses an SXM module form factor with PCIe 6.0 x16 connectivity and no display outputs. The RTX 5090 D is a dual-slot card with PCIe 5.0 x16 connectivity and display outputs including one HDMI 2.1b and three DisplayPort 2.1b connections. The B300 SXM6 AC has no recorded API support for DirectX, OpenGL, or Vulkan, while the RTX 5090 D supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Specification Differences

The two cards differ in nearly every recorded specification field. In terms of clocks, the RTX 5090 D has a higher base clock of 2017 MHz and a higher boost clock of 2407 MHz, compared to the B300 SXM6 AC’s 1665 MHz base and 2032 MHz boost. Memory clocks also differ: the B300 SXM6 AC runs at 2000 MHz with 8 Gbps effective speed, while the RTX 5090 D runs at 1750 MHz with 28 Gbps effective speed.

Pixel and texture rates favor the RTX 5090 D. It produces 423.6 GPixel/s and 1,636.8 GTexel/s, while the B300 SXM6 AC produces 48.77 GPixel/s and 1,202.9 GTexel/s. As noted, FP32 and FP16 rates are higher on the RTX 5090 D at 104.8 TFLOPS each, versus 76.99 TFLOPS each on the B300 SXM6 AC.

Power and physical requirements are vastly different. The B300 SXM6 AC has a TDP of 1100 W and a suggested PSU of 1500 W, while the RTX 5090 D has a TDP of 575 W and a suggested PSU of 950 W. The B300 SXM6 AC has no recorded power connectors, while the RTX 5090 D uses a single 16-pin connector. The RTX 5090 D has recorded dimensions of 304 mm in length, 137 mm in height, and 48 mm in width, while the B300 SXM6 AC has no recorded dimensions.

Release dates also differ. The RTX 5090 D was released on 2025-01-29, while the B300 SXM6 AC followed on 2025-09-10. Their predecessors and successors come from different product lines: the B300 SXM6 AC’s predecessor is Server Hopper and its successor is Server Rubin, while the RTX 5090 D’s predecessor is GeForce 40 and its successor is GeForce 60.

Where Each One Wins

The B300 SXM6 AC wins in compute density, memory capacity, and memory bandwidth. Its 288 GB of HBM3e memory and 8.19 TB/s bandwidth make it suitable for large models and data-heavy server workloads that cannot fit in the RTX 5090 D’s 32 GB GDDR7 pool. Its 19% lead in the Geekbench OpenCL test, combined with its 100th percentile standing and its 7% lead over the B200, shows that it is the highest-performing GPU in the database for general compute benchmarks.

The RTX 5090 D wins in graphics-oriented performance and consumer accessibility. It has higher pixel rate, texture rate, and FP32 throughput, plus dedicated RT cores and full DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4 support. It is the only one of the two with display outputs, making it functional as a standalone graphics card. Its 92nd percentile standing and its benchmark results across DirectX 9 through 12, Vulkan, and Passmark tests show that it is a versatile GPU for rendering and gaming workloads.

The B300 SXM6 AC is the clear choice for server deployments where memory capacity and compute throughput are paramount, while the RTX 5090 D is the practical option for desktop systems requiring graphics output, driver-level API support, and a dual-slot form factor. The B300 SXM6 AC’s 1100 W TDP and SXM module design require server infrastructure, whereas the RTX 5090 D’s 575 W TDP and dual-slot design fit into conventional workstation builds. Each card dominates in its intended domain, and the recorded data shows no overlap in their strengths.