NVIDIA GeForce RTX 4080 SUPER vs NVIDIA GeForce RTX 5090 Comparison

The NVIDIA GeForce RTX 5090 and the NVIDIA GeForce RTX 4080 SUPER represent two distinct generations of NVIDIA's flagship and upper-midrange graphics cards. The data in the database shows a decisive performance gap, with the RTX 5090 winning all ten head-to-head benchmark comparisons against the RTX 4080 SUPER. This analysis breaks down the recorded measurements, architectural differences, and use-case implications without venturing beyond the provided facts.

Head-to-Head Benchmarks

The most striking result in the database is the 3DMark Steel Nomad DX12 test, where the RTX 5090 scores 18,355 compared to the RTX 4080 SUPER's 6,600. This translates to a 178.1% delta, making it the largest single-benchmark advantage recorded between the two cards. This test, which typically stresses modern rendering pipelines, indicates a massive leap in raw DX12 performance for the newer architecture.

Beyond that flagship result, the RTX 5090 maintains a commanding lead across compute-oriented benchmarks. In Geekbench OpenCL, the RTX 5090 scores 334,370 versus 219,065 for the RTX 4080 SUPER, a 52.6% advantage. The Vulkan result is similarly lopsided: 376,728 against 260,075, a 44.9% lead. These figures suggest that the newer card’s compute capabilities scale far beyond what a simple clock-speed or core-count increase would imply.

The PassMark suite shows a more moderate but still consistent pattern of wins. In PassMark G3D, the RTX 5090 posts 39,650 compared to the RTX 4080 SUPER’s 34,245, a 15.8% difference. PassMark GPU Compute shows a larger gap: 26,756 versus 19,822, which is a 35% improvement. The delta in compute is more than double the delta in the G3D test, underscoring that the RTX 5090’s advantage grows when the workload is more parallel and math-intensive.

Looking at legacy DirectX tests, the RTX 5090 wins in DirectX 12 by 38.1% (185 vs. 134), DirectX 11 by 13.3% (341 vs. 301), DirectX 10 by 17.1% (226 vs. 193), and DirectX 9 by a much narrower 3.7% (395 vs. 381). The shrinking margin in older APIs is notable. It implies that for legacy titles or workloads that rely on older instruction paths, the performance difference is less dramatic, though the RTX 5090 still holds the win.

The 2D performance test, PassMark G2D, also goes to the RTX5090 with a 1,413 score versus 1,270, an 11.3% lead. While this is not a gaming metric, it reflects memory and pixel throughput differences that can affect desktop composition and certain productivity tasks.

In total, the RTX5090 wins 10 of 10 head-to-head benchmarks, with the RTX4080 SUPER failing to claim victory in any recorded test. The average benchmark score for the RTX5090 sits at 79,842, while the RTX4080 SUPER averages 54,209, a difference of roughly 47%. The RTX5090 also places in the 92nd percentile of all GPUs in the database, compared to the 86th percentile for the RTX4080 SUPER. That percentile gap is modest relative to the raw score difference, which indicates the performance field becomes extremely dense at the top.

Architecture Differences

The foundational architectural split is significant. The RTX5090 uses the GB202 chip based on the Blackwell 2.0 architecture, whereas the RTX4080 SUPER uses the AD103 chip based on Ada Lovelace. Both are manufactured by TSMC on a 5 nm process, so the node is identical. The transistor counts diverge sharply: the GB202 packs 92,200 million transistors, while AD103 has 45,900 million. The die sizes are 750 mm² for the RTX5090 and 379 mm² for the RTX4080 SUPER, giving the newer chip a transistor density of 122.9 million per square millimeter versus 121.1 million per square millimeter for the older one.

The memory subsystem is another major divergence. The RTX5090 comes with 32 GB of GDDR7 across a 512-bit bus, delivering a bandwidth of 1.79 TB/s. The RTX4080 SUPER has 16 GB of GDDR6X on a 256-bit bus, with a bandwidth of 736.3 GB/s. The 32 GB capacity and the higher bandwidth are not just about raw speed; they affect resolution scaling and the ability to hold larger datasets in VRAM.

The core counts follow the same trend. The RTX5090 has 21,760 shading units, 680 texture mapping units, and 176 raster operations pipelines. The RTX4080 SUPER has 10,240 shading units, 320 TMUs, and 112 ROPs. The RTX5090 also doubles the ray tracing cores (170 vs 80) and tensor cores (680 vs 320). This doubling of dedicated acceleration hardware is a direct contributor to the compute benchmark deltas.

Clock speeds are the one area where the RTX4080 SUPER appears higher on paper. The RTX4080 SUPER has a base clock of 2295 MHz and a boost clock of 2550 MHz, while the RTX5090 has a base of 2017 MHz and a boost of 2407 MHz. Despite the lower clocks, the RTX5090’s much larger core count and memory bandwidth allow it to outproduce the RTX4080 SUPER in raw throughput. The pixel rate for the RTX5090 is 423.6 GPixel/s versus 285.6 GPixel/s for the RTX4080 SUPER. The texture rate is 1,636.8 GTexel/s versus 816.0 GTexel/s. The FP32 compute throughput is practically double: 104.8 TFLOPS versus 52.22 TFLOPS, with the same 1:1 ratio for FP16.

The bus interface differs as well. The RTX5090 uses PCIe 5.0 x16, while the RTX4080 SUPER uses PCIe 4.0 x16. The RTX5090 also carries a higher TDP of 575 W, compared to 320 W for the RTX4080 SUPER, which correlates with the larger die and more transistors. The physical dimensions reflect this: the RTX5090 is 304 mm long, 137 mm high, and 40 mm wide, and is a dual-slot card. The RTX4080 SUPER is 310 mm long, 140 mm high, and 61 mm wide, and takes up a triple-slot design. The power connector is the same 16-pin on both, but the recommended PSU is 950 W for the RTX5090 and 700 W for the RTX4080 SUPER.

Finally, the release and production statuses are not identical. The RTX5090 was released on 2025-01-29 and is listed as an active product. The RTX4080 SUPER was released on 2024-01-30 and is marked as end-of-life. The RTX5090’s predecessor is the GeForce 40 series, which aligns with the RTX4080 SUPER being part of that generation.

Where Each One Wins

Given that the RTX5090 wins every head-to-head benchmark, there is no recorded test where the RTX4080 SUPER comes out ahead. However, the margin of victory varies meaningfully by workload, which gives some insight into where the older card is not as far behind.

The RTX5090 wins by the largest margins in modern, heavy workloads. The 3DMark Steel Nomad test (178.1% lead) and the Geekbench Vulkan test (44.9% lead) are prime examples. These are likely to represent the kind of load that a high-end gaming or workstation GPU would face in recent titles or compute tasks. The PassMark GPU Compute test (35% lead) also falls into this category, indicating a strong advantage in general-purpose compute.

The smallest leads for the RTX5090 are in legacy DirectX 9 (3.7%) and the older DirectX 11 (13.3%). These tests are likely bound by factors that do not scale as well with the added cores and bandwidth of the newer card. For someone running older software that uses these APIs, the RTX4080 SUPER is still a competent performer, and the data shows that the gap narrows substantially. The PassMark G2D test also shows a moderate 11.3% lead, which is less than the 3D or compute tests.

In terms of the overall percentile, the RTX5090 sits at the 92nd percentile of all GPUs, while the RTX4080 SUPER is at the 86th percentile. This is a smaller relative gap than the average scores suggest, which means that the RTX4080 SUPER is still a strong performer against the wider field. The nearest rivals in the database for the RTX5090 are the NVIDIA Tesla P100 PCIe 16 GB (0.3% higher average score), the Tesla P100 PCIe 12 GB (0.6% higher), the AMD Radeon RX 6850M XT (1.1% higher), and the AMD Radeon Pro Vega 64X (1.4% lower). For the RTX4080 SUPER, the nearest rivals are the RTX4080 (0.1% lower), the AMD Radeon Pro W5700X (1.1% lower), the AMD Radeon RX 6750 GRE 12 GB (2.7% lower), and the AMD Radeon 8060S (2.8% lower). This suggests that the RTX4080 SUPER’s performance is closely matched with its direct predecessor, while the RTX5090’s closest rivals are a mix of workstation and mobile parts.

FAQ

Q: Is the RTX 5090 faster than the RTX 4080 SUPER in 3DMark Steel Nomad?

A: Yes, the RTX 5090 scores 18,355 versus 6,600 for the RTX 4080 SUPER, a 178.1% lead.

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

A: The RTX 5090 has 32 GB of GDDR7, while the RTX 4080 SUPER has 16 GB of GDDR6X. The RTX 5090 also has a 512-bit bus versus a 256-bit bus.

Q: Which card has a higher boost clock speed?

A: The RTX 4080 SUPER has a higher boost clock at 2550 MHz, while the RTX 5090 boosts up to 2407 MHz.

Q: How do their compute scores compare?

A: In Geekbench OpenCL, the RTX 5090 scores 334,370 against 219,065 for the RTX 4080 SUPER (52.6% faster). In PassMark GPU Compute, the RTX 5090 scores 26,756 versus 19,822 (35% faster).

Q: Are both cards the same width?

A: No. The RTX 5090 is a dual-slot card at 40 mm wide, while the RTX 4080 SUPER is a triple-slot card at 61 mm wide.

Q: Which card has a higher average benchmark score?

A: The RTX 5090 has an average benchmark score of 79,842, compared to 54,209 for the RTX 4080 SUPER, placing them at the 92nd and 86th percentiles of all GPUs, respectively.

Specification Differences

The following specifications differ between the two cards based on the recorded data:

• Chip: GB202 (RTX 5090) vs AD103 (RTX 4080 SUPER)
• Architecture: Blackwell 2.0 (RTX 5090) vs Ada Lovelace (RTX 4080 SUPER)
• Generation: GeForce 50 (RTX 5090) vs GeForce 40 (RTX 4080 SUPER)
• Transistors: 92,200 million (RTX 5090) vs 45,900 million (RTX 4080 SUPER)
• Die Size: 750 mm² (RTX 5090) vs 379 mm² (RTX 4080 SUPER)
• Base Clock: 2017 MHz (RTX 5090) vs 2295 MHz (RTX 4080 SUPER)
• Boost Clock: 2407 MHz (RTX 5090) vs 2550 MHz (RTX 4080 SUPER)
• Memory Clock: 1750 MHz, 28 Gbps effective (RTX 5090) vs 1438 MHz, 23 Gbps effective (RTX 4080 SUPER)
• Memory Size: 32 GB (RTX 5090) vs 16 GB (RTX 4080 SUPER)
• Memory Type: GDDR7 (RTX 5090) vs GDDR6X (RTX 4080 SUPER)
• Memory Bus Width: 512 bit (RTX 5090) vs 256 bit (RTX 4080 SUPER)
• Memory Bandwidth: 736.3 GB/s (RTX 4080 SUPER) vs 1.79 TB/s (RTX 5090)
• Shading Units: 21,760 (RTX 5090) vs 10,240 (RTX 4080 SUPER)
• TMUs: 680 (RTX 5090) vs 320 (RTX 4080 SUPER)
• ROPs: 176 (RTX 5090) vs 112 (RTX 4080 SUPER)
• RT Cores: 170 (RTX 5090) vs 80 (RTX 4080 SUPER)
• Tensor Cores: 680 (RTX 5090) vs 320 (RTX 4080 SUPER)
• Pixel Rate: 423.6 GPixel/s (RTX 5090) vs 285.6 GPixel/s (RTX 4080 SUPER)
• Texture Rate: 1,636.8 GTexel/s (RTX 5090) vs 816.0 GTexel/s (RTX 4080 SUPER)
• FP32/FP16: 104.8 TFLOPS (RTX 5090) vs 52.22 TFLOPS (RTX 4080 SUPER)
• TDP: 575 W (RTX 5090) vs 320 W (RTX 4080 SUPER)
• Slot Width: Dual-slot (RTX 5090) vs Triple-slot (RTX 4080 SUPER)
• Suggested PSU: 950 W (RTX 5090) vs 700 W (RTX 4080 SUPER)
• Bus Interface: PCIe 5.0 x16 (RTX 5090) vs PCIe 4.0 x16 (RTX 4080 SUPER)
• Display Outputs: 1x HDMI 2.1b, 3x DisplayPort 2.1b (RTX 5090) vs 1x HDMI 2.1, 3x DisplayPort 1.4a (RTX 4080 SUPER)
• Dimensions: 304 mm x 137 mm x 40 mm (RTX 5090) vs 310 mm x 140 mm x 61 mm (RTX 4080 SUPER)
• Production Status: Active (RTX 5090) vs End-of-life (RTX 4080 SUPER)
• Release Date: 2025-01-29 (RTX 5090) vs 2024-01-30 (RTX 4080 SUPER)
• Launch MSRP: 1,999 USD (RTX 5090) vs 999 USD (RTX 4080 SUPER)

The Verdict

The recorded benchmark data leaves no ambiguity: the NVIDIA GeForce RTX 5090 is the dominant performer. It wins all ten head-to-head tests, with its most decisive victories in modern 3D workloads like 3DMark Steel Nomad (178.1% ahead) and compute-centric tests like Geekbench OpenCL (52.6% ahead) and PassMark GPU Compute (35% ahead). The architecture shift from Ada Lovelace to Blackwell 2.0, combined with a larger chip, more memory, and more processing cores, drives this performance gap.

The RTX 4080 SUPER is not a weak card by any means. It still ranks in the 86th percentile of all GPUs, and its nearest rivals include the RTX4080, which it edges by 0.1%. However, against the RTX5090, the only place it shows relative strength is in legacy DirectX 9 performance, where the delta is just 3.7%. For users with older applications, the RTX4080 SUPER remains viable, but it does not match the newer card in any recorded test.

The choice is therefore straightforward based on the data. The RTX5090 is for anyone who requires the highest recorded performance in modern benchmarks, with its 32 GB memory and 104.8 TFLOPS of FP32 compute. The RTX4080 SUPER, with its lower TDP of 320 W and smaller physical footprint (despite being triple-slot), is a more power-efficient option, but its performance ceiling is lower. The RTX5090 also carries a higher launch MSRP and a higher suggested PSU rating, which the database reflects. For a user building a system around the absolute peak of current performance, the RTX5090 is the clear choice. For a user who does not need the extreme compute or memory capacity, the RTX4080 SUPER is a competent, lower-power alternative that still performs near the top of the overall percentile.