NVIDIA GeForce RTX 4090 D vs NVIDIA GeForce RTX 5090 D Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive sweep for the NVIDIA GeForce RTX 5090 D across all three directly comparable benchmark tests. The largest margin appears in 3DMark Steel Nomad DX12, where the RTX 5090 D scores 14,326 against the RTX 4090 D's 8,587, a 40.1% advantage. This is the most demanding test of the trio, and the gap confirms that the Blackwell architecture's generational leap is not incremental but substantial. A 40% delta in a modern API workload is a major performance gulf, not a minor refresh.

In Geekbench Vulkan, the RTX 5090 D posts 376,915 versus 246,941 for the RTX 4090 D, a 34.5% lead. Vulkan tends to expose raw compute and draw-call efficiency, and the newer card's advantage here suggests that its architectural improvements scale well beyond just rasterization. The RTX 4090 D's older driver stack and Ada Lovelace design simply cannot match the throughput of the GB202 chip in this API.

The smallest margin of the three comes in Geekbench OpenCL, where the RTX 5090 D scores 310,674 against 278,621, a 10.3% lead. OpenCL is often more sensitive to memory bandwidth and driver optimization than to raw shader count, so a single-digit-percentile gap here is notable. Still, the RTX 5090 D wins every head-to-head test, with zero wins for the RTX 4090 D across the measured suite.

The average benchmark score tells a more complicated story, however. The RTX 4090 D holds an average score of 178,050 across its recorded tests, placing it in the 98th percentile of all GPUs. Its nearest rivals include the NVIDIA RTX PRO 5000 Blackwell at 182,109 (-2.2%), the A100 SXM4 80 GB at 183,725 (-3.1%), and the RTX 5000 Ada Generation at 184,664 (-3.6%). This indicates that the RTX 4090 D's aggregate performance is extremely close to professional-grade compute cards, which makes sense given its 14592 shading units and 456 tensor cores.

The RTX 5090 D, by contrast, shows an average benchmark score of 77,712, which places it in the 92nd percentile. Its nearest rivals in the database are utterly different in class: the AMD Radeon RX 6650M XT at 76,904 (+1.1%), the RX 6850M XT at 78,940 (-1.6%), and the Tesla P100 PCIe 12 GB at 79,396 (-2.1%). This discrepancy is puzzling at first glance, but it results from the benchmark suite composition. The RTX 5090 D has many more low-level PassMark tests recorded, including DirectX 9, 10, 11, and 12 scores, which drag down its average. The RTX 4090 D's average is based solely on its three strongest tests. The head-to-head data is the more reliable comparison, and there the RTX 5090 D is unequivocally faster.

FAQ

Q: Which card wins in 3DMark Steel Nomad DX12?

A: The RTX 5090 D wins decisively with a score of 14,326 versus 8,587 for the RTX 4090 D, a 40.1% delta in favor of the newer card.

Q: How much faster is the RTX 5090 D in Vulkan compute?

A: The RTX 5090 D scores 376,915 in Geekbench Vulkan, while the RTX 4090 D scores 246,941, a 34.5% advantage for the RTX 5090 D.

Q: Is the RTX 4090 D closer to its rivals than the RTX 5090 D is to its own?

A: Yes. The RTX 4090 D's average score of 178,050 puts it within 2.2% to 4.9% of its four nearest rivals. The RTX 5090 D's average of 77,712 sits within 1.1% to 2.4% of its nearest rivals, but those rivals are mobile and older data-center parts, reflecting a different benchmark mix.

Q: What is the gap in OpenCL performance?

A: The RTX 5090 D leads with 310,674 against 278,621, a 10.3% margin. This is the smallest of the three head-to-head deltas.

Q: Does the RTX 4090 D win any benchmark in the direct comparison?

A: No. Across 3dmark_steel_nomad_dx12, geekbench_opencl, and geekbench_vulkan, the RTX 5090 D wins all three. The win count is 3 for the RTX 5090 D and 0 for the RTX 4090 D.

Q: How do the two cards compare in their overall percentile rankings?

A: The RTX 4090 D sits in the 98th percentile of all GPUs, while the RTX 5090 D sits in the 92nd percentile. The lower percentile for the RTX 5090 D reflects its broader test set, not slower hardware; the direct benchmarks show it is faster.

Where Each One Wins

The RTX 5090 D is the clear choice for any workload that stresses modern APIs. Its 40.1% lead in 3DMark Steel Nomad DX12 and 34.5% lead in Vulkan make it the superior card for current-generation gaming, ray tracing, and compute-heavy tasks that leverage those interfaces. The 10.3% OpenCL advantage, while smaller, still means the RTX 5090 D is the better option for OpenCL-based rendering or scientific workloads where that API is common. With 32 GB of GDDR7 memory and 1.79 TB/s of bandwidth, the RTX 5090 D also provides more headroom for large datasets and high-resolution textures.

The RTX 4090 D, however, should not be dismissed. Its average benchmark score of 178,050 places it in the 98th percentile, and its nearest rivals are all professional or data-center cards within a few percent. This indicates that the RTX 4090 D remains an extremely capable GPU for tasks that are not covered in the head-to-head suite, such as legacy DirectX titles or specific compute kernels where its Ada Lovelace architecture and 456 tensor cores perform efficiently. Its 24 GB of GDDR6X memory is substantial, and its 73.54 TFLOPS of FP32 throughput is no small figure. For users who do not need the absolute peak in modern APIs, the RTX 4090 D still delivers elite-level performance.

The data splits cleanly: the RTX 5090 D wins every measured benchmark directly, but the RTX 4090 D's aggregate profile shows it is closer to its own peer group than the RTX 5090 D is to its oddly matched rivals. The RTX 4090 D is the better choice for a legacy-heavy software stack or for users who prioritize the 98th-percentile standing across a mixed suite. The RTX 5090 D is the better choice for anyone who wants the fastest recorded results in DX12 and Vulkan, the two most relevant APIs for future software.

Specification Differences

The two cards diverge sharply on nearly every physical and electrical specification. The RTX 4090 D uses a 5 nm process with 76,300 million transistors on a 609 mm² die, while the RTX 5090 D also uses 5 nm but packs 92,200 million transistors on a larger 750 mm² die. The transistor density is nearly identical: 125.3M per mm² for the RTX 4090 D versus 122.9M per mm² for the RTX 5090 D, confirming that the newer card's advantage comes from die size, not process refinement.

Clock speeds differ in favor of the older card. The RTX 4090 D has a base clock of 2280 MHz and a boost clock of 2520 MHz, while the RTX 5090 D runs at 2017 MHz base and 2407 MHz boost. Despite lower clocks, the RTX 5090 D achieves higher throughput due to its larger core count: 21760 shading units versus 14592, 680 TMUs versus 456, and 680 tensor cores versus 456. The RTX 5090 D also has 170 RT cores versus 114. Both cards have 176 ROPs, a rare point of parity.

Memory is a major differentiator. The RTX 4090 D has 24 GB of GDDR6X on a 384-bit bus, yielding 1.01 TB/s of bandwidth. The RTX 5090 D has 32 GB of GDDR7 on a 512-bit bus, yielding 1.79 TB/s. The effective memory clock jumps from 21 Gbps to 28 Gbps. Pixel rate actually favors the RTX 4090 D slightly at 443.5 GPixel/s versus 423.6 GPixel/s, but texture rate reverses that: 1,149.1 GTexel/s for the RTX 4090 D versus 1,636.8 GTexel/s for the RTX 5090 D. FP32 throughput is 73.54 TFLOPS for the RTX 4090 D and 104.8 TFLOPS for the RTX 5090 D, a 42.5% increase.

Power and cooling also differ. The RTX 4090 D has a TDP of 425 W and requires a suggested 800 W PSU, while the RTX 5090 D draws 575 W and asks for a 950 W PSU. Both use a single 16-pin connector, but the RTX 4090 D is triple-slot while the RTX 5090 D is dual-slot. The RTX 5090 D is also thinner: 48 mm versus 61 mm, though both are 304 mm long and 137 mm tall. The RTX 4090 D uses PCIe 4.0 x16, while the RTX 5090 D uses PCIe 5.0 x16. Display outputs differ as well: the RTX 4090 D offers HDMI 2.1 and DisplayPort 1.4a, while the RTX 5090 D offers HDMI 2.1b and DisplayPort 2.1b.

Architecture Differences

The RTX 4090 D is built on the AD102 chip using the Ada Lovelace architecture, released on 2023-12-27. The RTX 5090 D uses the GB202 chip with the Blackwell 2.0 architecture, released on 2025-01-29. This generational shift explains the performance gap despite similar process nodes. Ada Lovelace introduced the 40-series feature set, including improved RT cores and tensor cores for DLSS 3. Blackwell 2.0 doubles down on those areas: the RTX 5090 D has 170 RT cores versus 114, and 680 tensor cores versus 456.

The transistor count difference is stark: 92,200 million versus 76,300 million, a 21% increase. The die size grows from 609 mm² to 750 mm², a 23% increase. Since both are on TSMC 5 nm, the density is nearly identical, meaning the RTX 5090 D's advantage comes purely from silicon area. Blackwell 2.0 also brings a memory architecture shift from GDDR6X to GDDR7, enabling the 512-bit bus and 1.79 TB/s bandwidth, a 77% increase over the RTX 4090 D's 1.01 TB/s.

The API support is identical: both cards list DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The production status differs: the RTX 4090 D is end-of-life and its successor is the GeForce 50 series, while the RTX 5090 D is active with a successor in the GeForce 60 series. The RTX 4090 D's predecessor is the GeForce 30 series; the RTX 5090 D's predecessor is the GeForce 40 series. This places the RTX 5090 D as the direct successor to the RTX 4090 D in the product stack, confirmed by the release dates and the head-to-head data showing a 40.1% improvement in the most demanding test.

The launch MSRP for the RTX 4090 D was 1,599 USD, while the RTX 5090 D launched at 2,299 USD. The higher price aligns with the larger die, more memory, and higher TDP. The RTX 5090 D's dual-slot form factor is a notable engineering change, allowing it to fit in more chassis despite the higher power draw. The PCIe 5.0 interface doubles the bus bandwidth available to the card, which matters for data transfer in compute workloads even if gaming performance is less sensitive to it. The DisplayPort 2.1b outputs on the RTX 5090 D support higher refresh rates at high resolutions compared to the RTX 4090 D's DisplayPort 1.4a. All of these differences compound into a clear architectural upgrade, with the RTX 5090 D winning every direct benchmark by margins that range from 10.3% to 40.1%.