AMD Radeon RX 6900 XT vs NVIDIA GeForce RTX 5090 D Comparison

FAQ

Q: How much faster is the NVIDIA GeForce RTX 5090 D than the AMD Radeon RX 6900 XT in the primary DirectX 12 benchmark?

A: In 3DMark Steel Nomad DX12, the RTX 5090 D scores 14,326 versus 4,079 for the RX 6900 XT, a 251.2% advantage for the NVIDIA card.

Q: Which card has a higher overall benchmark percentile ranking?

A: The RTX 5090 D sits in the 92nd percentile of all GPUs, while the RX 6900 XT is in the 86th percentile. The NVIDIA card's average benchmark score is 77,712, compared to 50,951 for the AMD card.

Q: What is the difference in memory capacity and type between the two cards?

A: The RTX 5090 D features 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The RX 6900 XT has 16 GB of GDDR6 memory on a 256-bit bus, yielding 512.0 GB/s of bandwidth.

Q: How do the two cards compare in compute performance?

A: In Passmark GPU Compute, the RTX 5090 D scores 28,396 versus 14,547 for the RX 6900 XT, a 95.2% difference. The FP32 throughput figures are 104.8 TFLOPS for NVIDIA and 23.04 TFLOPS for AMD.

Q: What are the production statuses of these two cards?

A: The RTX 5090 D is listed as Active, with a release date of 2025-01-29. The RX 6900 XT is End-of-life, having been released on 2020-10-27.

Q: Which card wins in the Vulkan benchmark, and by how much?

A: The RTX 5090 D wins in Geekbench Vulkan with a score of 376,915, versus 148,525 for the RX 6900 XT. This represents a 153.8% advantage for the NVIDIA card.

Architecture Differences

The two GPUs come from fundamentally different architectural generations. The NVIDIA GeForce RTX 5090 D is built on the Blackwell 2.0 architecture using the GB202 chip, while the AMD Radeon RX 6900 XT uses the RDNA 2.0 architecture with the Navi 21 chip.

The manufacturing processes differ significantly. The RTX 5090 D is fabricated on a 5 nm process at TSMC, whereas the RX 6900 XT uses a 7 nm process, also from TSMC. This process advantage contributes to a substantial difference in transistor counts: the NVIDIA chip contains 92,200 million transistors on a 750 mm² die, while the AMD chip has 26,800 million transistors on a 520 mm² die. The resulting transistor density is 122.9M per mm² for NVIDIA versus 51.5M per mm² for AMD.

Compute resources show a wide gap. The RTX 5090 D has 21,760 shading units, 680 texture mapping units, and 176 ROPs. The RX 6900 XT has 5,120 shading units, 320 TMUs, and 128 ROPs. Ray tracing hardware differs as well: NVIDIA includes 170 RT cores, while AMD has 80. The NVIDIA card also carries 680 tensor cores, a feature entirely absent from the AMD specification.

Memory architecture represents another major divergence. The RTX 5090 D uses 32 GB of GDDR7 across a 512-bit interface, while the RX 6900 XT uses 16 GB of GDDR6 on a 256-bit bus. The resulting bandwidth figures are 1.79 TB/s versus 512.0 GB/s, a 3.5x difference in favor of NVIDIA.

Clock behavior differs between the two. The RTX 5090 D has a base clock of 2017 MHz and a boost clock of 2407 MHz. The RX 6900 XT has a lower base clock at 1825 MHz, a boost clock of 2250 MHz, and a game clock of 2015 MHz. Despite lower clocks, the AMD card's memory runs at 2000 MHz (16 Gbps effective), while the NVIDIA memory runs at 1750 MHz (28 Gbps effective).

Power and physical specifications also differ. The RTX 5090 D has a TDP of 575 W and requires a 950 W suggested PSU, using a single 16-pin connector in a dual-slot form factor. The RX 6900 XT has a 300 W TDP, a 700 W suggested PSU, two 8-pin connectors, and a triple-slot design. The NVIDIA card is longer at 304 mm versus 267 mm, taller at 137 mm versus 120 mm, but slightly narrower at 48 mm versus 50 mm.

Interface support also differs. The RTX 5090 D uses PCIe 5.0 x16, while the RX 6900 XT uses PCIe 4.0 x16. Display outputs vary as well: NVIDIA provides 1x HDMI 2.1b and 3x DisplayPort 2.1b, while AMD offers 1x HDMI 2.1, 2x DisplayPort 1.4a, and 1x USB Type-C.

The Verdict

The benchmark data presents a clear hierarchy. The RTX 5090 D wins all ten head-to-head benchmark comparisons, with zero wins for the RX 6900 XT. The smallest margin is 33% in Passmark DirectX 11, while the largest is 251.2% in 3DMark Steel Nomad DX12.

For users prioritizing raw DirectX 12 performance, the RTX 5090 D is the definitive choice. Its 3DMark Steel Nomad score of 14,326 versus 4,079 places it in a different performance class. The 95.2% advantage in GPU compute further reinforces this position for compute-heavy workloads.

The RX 6900 XT, despite being end-of-life, still holds a respectable 86th percentile ranking. Its nearest rivals in the database include the NVIDIA GeForce RTX 5070 Ti, from which it trails by 2%, and the AMD Radeon RX Vega 64, which it leads by 1.9%. This indicates the RX 6900 XT remains competitive within its own generation, but it cannot match the newer NVIDIA architecture.

The RTX 5090 D's nearest rivals are notably different in character. It leads the AMD Radeon RX 6850M XT by 1.6%, trails the NVIDIA Tesla P100 PCIe 16 GB by 2.4%, and sits 1.1% ahead of the AMD Radeon RX 6650M XT. These are relatively small margins, suggesting the RTX 5090 D is closely grouped with high-end mobile and data center parts in overall average scoring, despite its commanding lead over the RX 6900 XT.

For users on a strict power budget, the RX 6900 XT's 300 W TDP versus 575 W for the RTX 5090 D may be a consideration. However, the performance differential across every recorded benchmark favors NVIDIA overwhelmingly. The data indicates that the RTX 5090 D is the appropriate choice for users who need maximum performance in DX12, Vulkan, and compute workloads, while the RX 6900 XT remains a viable option for those within an older ecosystem or with more limited power delivery.

Specification Differences

The following fields differ between the two cards:

• Process Node: 5 nm (NVIDIA) versus 7 nm (AMD)
• Transistors: 92,200 million versus 26,800 million
• Die Size: 750 mm² versus 520 mm²
• Transistor Density: 122.9M / mm² versus 51.5M / mm²
• Base Clock: 2017 MHz versus 1825 MHz
• Boost Clock: 2407 MHz versus 2250 MHz
• Memory Clock: 1750 MHz 28 Gbps effective versus 2000 MHz 16 Gbps effective
• Memory Size: 32 GB versus 16 GB
• Memory Type: GDDR7 versus GDDR6
• Memory Bus Width: 512 bit versus 256 bit
• Memory Bandwidth: 1.79 TB/s versus 512.0 GB/s
• Shading Units: 21,760 versus 5,120
• TMUs: 680 versus 320
• ROPs: 176 versus 128
• RT Cores: 170 versus 80
• Tensor Cores: 680 versus null
• Pixel Rate: 423.6 GPixel/s versus 288.0 GPixel/s
• Texture Rate: 1,636.8 GTexel/s versus 720.0 GTexel/s
• FP32: 104.8 TFLOPS versus 23.04 TFLOPS
• FP16: 104.8 TFLOPS (1:1) versus 46.08 TFLOPS (2:1)
• TDP: 575 W versus 300 W
• Slot Width: Dual-slot versus Triple-slot
• Power Connectors: 1x 16-pin versus 2x 8-pin
• Suggested PSU: 950 W versus 700 W
• Bus Interface: PCIe 5.0 x16 versus PCIe 4.0 x16
• Display Outputs: 1x HDMI 2.1b, 3x DisplayPort 2.1b versus 1x HDMI 2.1, 2x DisplayPort 1.4a, 1x USB Type-C
• Dimensions: 304 mm x 137 mm x 48 mm versus 267 mm x 120 mm x 50 mm
• Production Status: Active versus End-of-life
• Release Date: 2025-01-29 versus 2020-10-27
• Predecessor: GeForce 40 versus Navi
• Successor: GeForce 60 versus Navi III

Head-to-Head Benchmarks

The RTX 5090 D dominates every recorded benchmark comparison, but the margins vary considerably by workload.

The largest victory comes in 3DMark Steel Nomad DX12, where the NVIDIA card scores 14,326 against 4,079 for AMD, a 251.2% delta. This benchmark highlights the generational gap in DirectX 12 performance, where the Blackwell architecture's feature set and raw throughput produce a massive lead.

Geekbench Vulkan shows a 153.8% advantage, with scores of 376,915 versus 148,525. This indicates the NVIDIA card's Vulkan driver and hardware implementation are substantially more efficient than the RDNA 2.0 implementation.

Passmark GPU Compute delivers a 95.2% margin, with 28,396 versus 14,547. This aligns with the FP32 throughput difference, where NVIDIA's 104.8 TFLOPS dwarfs AMD's 23.04 TFLOPS. The compute lead is consistent across both synthetic and application-style workloads.

Passmark DirectX 12 shows a 93.8% advantage, with 219 versus 113. This is a smaller absolute gap than the 3DMark result, but still a decisive NVIDIA win. The DirectX 11 benchmark shows a 33% lead, with 371 versus 279, representing the narrowest margin of all tests.

Geekbench OpenCL shows a 65.5% advantage, with 310,674 versus 187,673. This reinforces the compute narrative, though the OpenCL gap is smaller than the Vulkan gap, suggesting AMD's OpenCL implementation is relatively stronger than its Vulkan path.

Passmark G3D shows a 64.8% lead, with 44,065 versus 26,732. This aggregate 3D score reflects the overall gaming performance picture. DirectX 9 shows a 61.9% margin, with 434 versus 268, indicating the NVIDIA card also excels in legacy API workloads.

The 2D benchmark shows a 40.9% advantage, with 1,487 versus 1,055. While 2D performance is rarely a bottleneck, the margin shows NVIDIA's advantage extends beyond 3D rendering. DirectX 10 shows a 38.3% lead, with 231 versus 167, another legacy API win for NVIDIA.

Across all ten benchmarks, the average delta is approximately 89.9%, with the RTX 5090 D winning every test. The RX 6900 XT's best relative performance comes in DirectX 11, where it trails by only 33%, while its worst relative performance is in 3DMark Steel Nomad, where it trails by 251.2%. The data indicates that the NVIDIA card is not merely faster, but categorically superior in every measured workload, with the largest advantages appearing in modern DirectX 12 and Vulkan scenarios.