AMD Radeon RX 6950 XT vs NVIDIA GeForce RTX 5090 D Comparison

Head-to-Head Benchmarks

The benchmark database records a decisive sweep: the NVIDIA GeForce RTX 5090 D wins all 10 head-to-head comparisons against the AMD Radeon RX 6950 XT. The largest margin appears in 3DMark Steel Nomad DX12, where the RTX 5090 D scores 14,326 versus 4,235 for the RX 6950 XT, a 238.3% advantage. That is not a marginal gap; it places the NVIDIA card in a different performance tier entirely.

Vulkan results tell a similar story. The RTX 5090 D posts 376,915 points in Geekbench Vulkan against 165,212 for the RX 6950 XT, a 128.1% lead. OpenCL shows a narrower but still substantial gap: 310,674 versus 205,998, a 50.8% difference. The PassMark suite reinforces the pattern. In DirectX 12, the RTX 5090 D scores 219 against 114, a 92.1% advantage. DirectX 11 sees 371 versus 300, a 23.7% lead, while DirectX 10 results are 231 versus 164, a 40.9% margin. Even legacy DirectX 9 favors the newer card: 434 versus 303, a 43.2% difference.

Compute workloads show the most striking doubling. The RTX 5090 D records 28,396 in PassMark GPU Compute, exactly 100% ahead of the RX 6950 XT's 14,199. That is a clean 2x result, not a rounding artifact. The 3DMark Steel Nomad result is the outlier in magnitude, but the compute score is the cleanest evidence of architectural scaling.

The 2D and 3D graphics split also favors NVIDIA. PassMark G3D gives the RTX 5090 D 44,065 against 28,070, a 57% lead. PassMark G2D shows 1,487 versus 1,063, a 39.9% advantage. Even the least flattering comparison for NVIDIA, DirectX 11 at 23.7%, remains a comfortable win. Across every recorded workload, the RTX 5090 D holds between roughly a quarter and nearly two and a half times the RX 6950 XT's score.

The average benchmark score reinforces this. The RTX 5090 D averages 77,712 across its recorded tests, while the RX 6950 XT averages 58,392. The NVIDIA card sits at the 92nd percentile among all GPUs in the database; the AMD card sits at the 88th. The nearest rivals for the RTX 5090 D include the AMD Radeon RX 6850M XT at 78,940 (1.6% behind the NVIDIA card) and the NVIDIA Tesla P100 PCIe 16 GB at 79,605 (2.4% behind). The RX 6950 XT, by contrast, trades blows with the NVIDIA P102-100 at 58,528 (0.2% behind) and the Intel Arc A570M at 58,239 (0.3% ahead of the AMD card). That comparison range shows how far apart these two products are in the database's ranking hierarchy.

Architecture Differences

The two cards come from different manufacturing generations and design philosophies. The RTX 5090 D uses the GB202 chip built on Blackwell 2.0 architecture, fabricated on a 5 nm process at TSMC. It packs 92,200 million transistors on a 750 mm² die, yielding a transistor density of 122.9 million per square millimeter. The RX 6950 XT uses the Navi 21 chip on RDNA 2.0, built on a 7 nm process, also at TSMC. It contains 26,800 million transistors on a 520 mm² die, for a density of 51.5 million per square millimeter.

The transistor count difference is enormous: the NVIDIA chip has roughly 3.4 times the transistor count of the AMD chip. That scale shows up directly in the shading unit counts. The RTX 5090 D carries 21,760 shading units, 680 texture mapping units, and 176 render output units. The RX 6950 XT has 5,120 shading units, 320 TMUs, and 128 ROPs. The NVIDIA card also features 170 ray tracing cores and 680 tensor cores, while the AMD card has 80 ray tracing cores and no dedicated tensor core array.

Memory subsystems diverge sharply. The RTX 5090 D uses 32 GB of GDDR7 on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The RX 6950 XT uses 16 GB of GDDR6 on a 256-bit bus, providing 576.0 GB/s. That is a 3.1x bandwidth advantage for NVIDIA, which matters in memory-heavy workloads like 4K texture streaming and compute. The RX 6950 XT's memory clock runs at 2250 MHz with 18 Gbps effective speed; the RTX 5090 D runs its memory at 1750 MHz with 28 Gbps effective speed, the higher data rate compensating for a lower base clock.

Clock speeds favor the AMD card in raw frequency. The RX 6950 XT has a base clock of 1860 MHz and a boost of 2310 MHz, with a game clock of 2100 MHz. The RTX 5090 D has a base of 2017 MHz and a boost of 2407 MHz. The NVIDIA card boosts higher, but the AMD card's game clock is a notable reference point. The real separation comes from throughput metrics. The RTX 5090 D reaches 104.8 TFLOPS FP32 and 104.8 TFLOPS FP16 (1:1 ratio). The RX 6950 XT reaches 23.65 TFLOPS FP32 and 47.31 TFLOPS FP16 (2:1 ratio). The NVIDIA card offers 4.4x the FP32 throughput and 2.2x the FP16 throughput.

Pixel and texture rates follow the same pattern. The RTX 5090 D delivers 423.6 GPixel/s and 1,636.8 GTexel/s. The RX 6950 XT manages 295.7 GPixel/s and 739.2 GTexel/s. Power draw differs accordingly: the RTX 5090 D has a TDP of 575 W with a suggested 950 W PSU, while the RX 6950 XT draws 335 W with a suggested 700 W PSU. The NVIDIA card uses a single 16-pin connector and is dual-slot; the AMD card uses two 8-pin connectors and is triple-slot. The RTX 5090 D is longer at 304 mm versus 267 mm, and taller at 137 mm versus 120 mm, though the AMD card is slightly thicker at 50 mm versus 48 mm.

Interface support favors the newer card. The RTX 5090 D runs PCIe 5.0 x16 with one HDMI 2.1b and three DisplayPort 2.1b outputs. The RX 6950 XT runs PCIe 4.0 x16 with one HDMI 2.1 and two DisplayPort 1.4a outputs. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API compatibility is not a differentiator.

Where Each One Wins

The RTX 5090 D wins every recorded benchmark, so the "where each one wins" split is a question of relative margin rather than absolute victory. The biggest wins for NVIDIA come in 3DMark Steel Nomad DX12 (238.3% ahead), Vulkan (128.1%), and GPU compute (100%). These are workloads that stress the full pipeline: geometry throughput, ray tracing, and parallel compute. The 3DMark result in particular suggests the RTX 5090 D excels in modern DX12 titles with heavy effects and high geometry counts.

The compute result at exactly 100% ahead points to the tensor core advantage. The RX 6950 XT has no tensor cores at all, while the RTX 5090 D has 680. That architectural feature shows up most clearly in GPU compute benchmarks, where the NVIDIA card doubles the AMD card's score. For users running AI inference, physics simulation, or scientific workloads, the RTX 5090 D is not just faster; it is in a different class.

The narrower NVIDIA wins, such as DirectX 11 at 23.7% and OpenCL at 50.8%, still represent solid advantages but indicate workloads where the RX 6950 XT's architecture remains competitive. DirectX 11 is legacy API territory, and the RX 6950 XT's high boost clock of 2310 MHz helps it keep pace better there than in modern APIs. The G2D result at 39.9% shows the NVIDIA card also leads in 2D operations, though this is rarely a deciding factor for gaming.

For the RX 6950 XT, the best case is not a win but damage control. Its smallest deficits come in DirectX 11 and G2D, where it trails by under 40%. Its nearest rival cluster includes the NVIDIA P102-100, Intel Arc A570M, and AMD Radeon PRO V710, all within 0.5% of its average score. That places the RX 6950 XT in a mid-range bracket, far below the RTX 5090 D's neighborhood of Tesla P100 cards and the RX 6850M XT.

The use-case split is therefore clear. The RTX 5090 D is the choice for anyone prioritizing the highest possible frame rates in modern DirectX 12 and Vulkan titles, plus any compute-heavy work. The RX 6950 XT remains viable for older DirectX 11 games and for users who need a capable card without the RTX 5090 D's power draw or physical footprint. The AMD card is end-of-life in production status, while the NVIDIA card is active, which matters for long-term driver support and availability.

The Verdict

The data is unambiguous: the NVIDIA GeForce RTX 5090 D outperforms the AMD Radeon RX 6950 XT in every recorded benchmark, with margins ranging from 23.7% to 238.3%. The average benchmark score difference (77,712 versus 58,392) places the RTX 5090 D at the 92nd percentile and the RX 6950 XT at the 88th. For users who want the highest performance in the database, the RTX 5090 D is the clear pick.

The RX 6950 XT's launch MSRP was 1,099 USD, and the RTX 5090 D's launch MSRP was 2,299 USD. That price differential is substantial, but the performance gap is even larger in the most demanding workloads. The RTX 5090 D offers 4.4x the FP32 throughput, 3.1x the memory bandwidth, and double the compute score. In 3DMark Steel Nomad, it is more than three times faster.

Who should pick the RX 6950 XT? Users who prioritize lower power consumption (335 W versus 575 W), a smaller physical footprint in height (120 mm versus 137 mm), and compatibility with standard 8-pin power connectors. The RX 6950 XT also uses a triple-slot cooler while the RTX 5090 D is dual-slot, so case compatibility varies. For legacy DirectX 11 workloads, the RX 6950 XT's deficit narrows to under a quarter, making it a reasonable choice for older game libraries.

Who should pick the RTX 5090 D? Anyone running modern DX12 or Vulkan titles, compute tasks, or ray-traced content. The 170 ray tracing cores and 680 tensor cores give it capabilities the RX 6950 XT simply does not have. The 32 GB of memory doubles the AMD card's 16 GB, and the 1.79 TB/s bandwidth is unmatched in the database's comparison set. For users building a top-tier system with a 950 W PSU and PCIe 5.0 support, the RTX 5090 D is the superior choice on every metric recorded.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA GeForce RTX 5090 D averages 77,712 points, while the AMD Radeon RX 6950 XT averages 58,392 points.

Q: What is the largest performance gap between the two cards?

A: In 3DMark Steel Nomad DX12, the RTX 5090 D scores 14,326 against 4,235 for the RX 6950 XT, a 238.3% difference.

Q: Does the RX 6950 XT win any benchmark?

A: No. The database records 10 head-to-head benchmarks, and the RTX 5090 D wins all 10. The closest margin is DirectX 11, where NVIDIA leads by 23.7%.

Q: How do the memory configurations compare?

A: The RTX 5090 D has 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The RX 6950 XT has 16 GB of GDDR6 on a 256-bit bus with 576.0 GB/s bandwidth.

Q: What are the transistor counts for each chip?

A: The RTX 5090 D's GB202 chip contains 92,200 million transistors on a 750 mm² die. The RX 6950 XT's Navi 21 chip contains 26,800 million transistors on a 520 mm² die.

Q: Which card has dedicated tensor cores?

A: Only the RTX 5090 D, which features 680 tensor cores. The RX 6950 XT has no tensor core array, which contributes to the 100% gap in PassMark GPU Compute.