AMD Radeon Pro W6900X vs NVIDIA GeForce RTX 5090 D Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive advantage for the NVIDIA GeForce RTX 5090 D in the two directly comparable benchmark tests. In Geekbench OpenCL, the RTX 5090 D scores 310,674 against the AMD Radeon Pro W6900X’s 130,035, a delta of 58.1% in NVIDIA’s favor. The gap widens further in Geekbench Vulkan, where NVIDIA posts 376,915 versus AMD’s 148,865, a 60.5% lead. These are not marginal differences; they represent a complete generational overhaul in raw compute throughput.

However, the average benchmark scores tell a more nuanced story. The AMD card holds a 97th percentile ranking across all GPUs in the database, with an average score of 168,574. The NVIDIA card, despite its massive wins in the head-to-head tests, sits at the 92nd percentile with an average score of 77,712. This discrepancy arises because the RTX 5090 D’s benchmark set includes several Passmark tests (DirectX 9, 10, 11, 12, G2D, G3D, and GPU Compute) with scores ranging from 219 to 44,065, which drag down its aggregate average relative to the AMD card’s more focused Geekbench-only profile.

Looking at the nearest rivals in the database, the AMD W6900X slots in just ahead of the NVIDIA RTX 4500 Ada Generation (1.5% delta), the RTX A5500 (2% delta), and the AMD Radeon PRO W7800 (2.2% delta). It also edges out the NVIDIA A100 PCIe 40 GB by 3.7%. These are tight margins, indicating that the W6900X sits in a competitive mid-to-high tier for professional workloads. The RTX 5090 D, by contrast, finds its nearest rivals among mobile GPUs: the AMD Radeon RX 6650M XT (1.1% higher), the RX 6850M XT (1.6% lower), and two NVIDIA Tesla P100 variants (2.1% and 2.4% lower). This suggests that the RTX 5090 D’s average score is heavily influenced by its diverse test suite rather than its true compute ceiling.

In the two overlapping benchmarks, the NVIDIA card is not just ahead; it is ahead by a margin that dwarfs any delta seen among the AMD card’s rivals. The 58.1% and 60.5% deltas in OpenCL and Vulkan respectively are over 15 times larger than the W6900X’s closest rival gap. This indicates a fundamental performance tier separation, not a subtle optimization difference.

Architecture Differences

The two cards come from different eras and design philosophies. The AMD Radeon Pro W6900X uses the Navi 21 chip on the RDNA 2.0 architecture, built on a 7 nm TSMC process. It packs 26,800 million transistors into a 520 mm² die, yielding a transistor density of 51.5 million per mm². The NVIDIA GeForce RTX 5090 D uses the GB202 chip on the Blackwell 2.0 architecture, manufactured on a 5 nm TSMC process. It contains 92,200 million transistors across a 750 mm² die, achieving 122.9 million transistors per mm². The NVIDIA chip is more than twice as dense and nearly three times the transistor count.

Memory configurations differ sharply. The AMD card features 32 GB of GDDR6 on a 256-bit bus, delivering 512.0 GB/s bandwidth at 16 Gbps effective. The NVIDIA card also has 32 GB, but uses GDDR7 on a 512-bit bus, achieving 1.79 TB/s at 28 Gbps effective. That is roughly 3.5 times the memory bandwidth, a critical factor for large datasets and high-resolution textures.

Compute resources are heavily skewed toward NVIDIA. The RTX 5090 D has 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The W6900X has 5,120 shading units, 320 TMUs, 128 ROPs, and 80 RT cores, with no tensor core equivalent. Pixel rate stands at 423.6 GPixel/s for NVIDIA versus 277.9 GPixel/s for AMD. Texture rate is 1,636.8 GTexel/s versus 694.7 GTexel/s. FP32 compute is 104.8 TFLOPS for NVIDIA versus 22.23 TFLOPS for AMD. FP16 is also 104.8 TFLOPS for NVIDIA (1:1 ratio) versus 44.46 TFLOPS for AMD (2:1 ratio).

Clock speeds favor NVIDIA as well, with a base of 2017 MHz and boost of 2407 MHz versus AMD’s 1825 MHz base and 2171 MHz boost. Power draw reflects this: the NVIDIA card is rated at 575 W TDP with a 950 W suggested PSU, while the AMD card draws 300 W with a 700 W suggested PSU. The NVIDIA card uses a single 16-pin connector and occupies dual slots, while the AMD card uses the Apple MPX bus interface with no slot width listed. Display outputs differ: NVIDIA offers 1x HDMI 2.1b and 3x DisplayPort 2.1b, while AMD offers 1x HDMI 2.1 and 4x Thunderbolt.

The NVIDIA card is built on PCIe 5.0 x16, whereas AMD uses the proprietary Apple MPX interface. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API compatibility is equivalent on paper.

Where Each One Wins

The RTX 5090 D wins decisively in every directly comparable compute benchmark. Its Geekbench OpenCL and Vulkan scores are roughly 2.4 and 2.5 times higher, respectively. For workloads that stress raw GPU compute, such as rendering, physics simulation, or machine learning inference, the data points exclusively to NVIDIA. The 104.8 TFLOPS FP32 figure is nearly five times the AMD card’s 22.23 TFLOPS, and the 1.79 TB/s memory bandwidth is a massive advantage for memory-bound tasks.

The W6900X wins in power efficiency and integration context. At 300 W versus 575 W, it delivers its 97th percentile performance with nearly half the power draw. Its Apple MPX interface and 4x Thunderbolt outputs make it purpose-built for Mac Pro systems, where the RTX 5090 D’s PCIe 5.0 x16 interface and dual-slot design may not fit. The AMD card’s 267 mm length and 120 mm height are smaller than NVIDIA’s 304 mm length, 137 mm height, and 48 mm width, which matters in constrained chassis.

The average benchmark percentile also favors AMD in aggregate standing, though this is an artifact of test coverage. The W6900X’s 97th percentile versus the RTX 5090 D’s 92nd percentile reflects that the AMD card’s benchmarks are all compute-heavy and high-scoring, while the NVIDIA card’s Passmark suite includes lower scores that pull its average down.

The Verdict

The data is unambiguous: the NVIDIA GeForce RTX 5090 D is the superior performer in every head-to-head test recorded. Its 58.1% lead in OpenCL and 60.5% lead in Vulkan, combined with 104.8 TFLOPS FP32 and 1.79 TB/s bandwidth, place it in a different performance class. Users prioritizing raw compute throughput should choose the RTX 5090 D without hesitation.

The AMD Radeon Pro W6900X remains relevant for a specific niche. Its 97th percentile ranking, 300 W power draw, and Apple MPX interface make it the only option for Mac Pro environments that require NVIDIA-level compute without the power envelope or PCIe compatibility. For anyone building a standard PCIe system, the RTX 5090 D is the clear choice based on benchmark deltas alone.

Neither card is a value pick based on the data available. The AMD card is end-of-life, released in August 2021, while the NVIDIA card is active, released in January 2025. The RTX 5090 D has a successor in the GeForce 60 series, whereas the W6900X has none listed. For longevity and performance headroom, the RTX 5090 D is the rational selection.

FAQ

Q: Which card has higher raw compute performance?

A: The NVIDIA GeForce RTX 5090 D. It scores 310,674 in Geekbench OpenCL and 376,915 in Geekbench Vulkan, versus 130,035 and 148,865 for the AMD Radeon Pro W6900X. Its FP32 rating is 104.8 TFLOPS compared to 22.23 TFLOPS.

Q: How do their memory subsystems compare?

A: Both have 32 GB, but the NVIDIA card uses GDDR7 on a 512-bit bus for 1.79 TB/s bandwidth, while the AMD card uses GDDR6 on a 256-bit bus for 512.0 GB/s. The NVIDIA card has roughly 3.5 times the bandwidth.

Q: Which card is more power-efficient?

A: The AMD Radeon Pro W6900X, with a 300 W TDP and 700 W suggested PSU, versus 575 W and 950 W for the NVIDIA card. The AMD card achieves a 97th percentile ranking with less than half the power draw.

Q: Are they compatible with the same systems?

A: No. The AMD card uses the Apple MPX bus interface with 4x Thunderbolt outputs, while the NVIDIA card uses PCIe 5.0 x16 with 1x HDMI 2.1b and 3x DisplayPort 2.1b. The AMD card is 267 mm long and 120 mm high; the NVIDIA card is 304 mm long, 137 mm high, and 48 mm wide.

Q: What are their production statuses?

A: The AMD Radeon Pro W6900X is end-of-life, released on August 2, 2021. The NVIDIA GeForce RTX 5090 D is active, released on January 29, 2025, with a successor in the GeForce 60 series.

Q: How do their average benchmark scores rank them?

A: The AMD card has an average benchmark score of 168,574 and sits at the 97th percentile. The NVIDIA card has an average of 77,712 at the 92nd percentile, though this is skewed by its broader Passmark test set.