NVIDIA GeForce RTX 4090 vs NVIDIA GeForce RTX 5090 Comparison

The NVIDIA GeForce RTX 5090 and RTX 4090 represent two successive generations of flagship graphics cards, and the benchmark data in our database shows a clear, though not uniform, shift in performance. The RTX 5090 wins 9 out of 10 head-to-head tests, but the margins vary dramatically from a near-tie to a near-doubling of performance. This analysis breaks down those results, the architectural changes behind them, and what the recorded data says about which card is the right choice for different workloads.

Head-to-Head Benchmarks

The largest single victory for the RTX 5090 comes in the 3DMark Steel Nomad DX12 test, where it scores 18,355 points against the RTX 4090's 9,223 points. That is a 99% difference, meaning the newer card is nearly twice as fast in this specific, modern DirectX 12 workload. This is the most dramatic divergence in the entire dataset and highlights the generational leap in raw rendering capability.

The compute-oriented benchmarks also show substantial gains for the RTX 5090. In Geekbench Vulkan, the RTX 5090 posts 376,728 points versus 271,631 for the RTX 4090, a 38.7% advantage. The Geekbench OpenCL result is similarly lopsided: 334,370 points for the RTX 5090 against 255,416 for the RTX 4090, a 30.9% lead. These figures suggest that the RTX 5090's architectural changes deliver significant improvements in general-purpose compute and API-specific rendering paths.

The Passmark suite provides a more nuanced picture. The most striking result is in Passmark DirectX 12, where the RTX 5090 scores 185 versus the RTX 4090's 150, a 23.3% improvement. This is a solid win for the newer card, though not on the scale of the 3DMark result. In Passmark DirectX 11, the RTX 5090 leads with 341 points against 326, a 4.6% margin. The gap narrows further in Passmark G3D, where the RTX 5090 scores 39,650 and the RTX 4090 scores 38,194, a 3.8% difference.

The RTX 5090 also edges out the RTX 4090 in Passmark G2D (1,413 vs 1,299, an 8.8% lead) and in Passmark DirectX 10 (226 vs 224, a 0.9% margin). The compute gap nearly vanishes in Passmark GPU Compute, with the RTX 5090 scoring 26,756 and the RTX 4090 scoring 26,613, a mere 0.5% difference. The only test the RTX 4090 wins is Passmark DirectX 9, where it scores 397 against the RTX 5090's 395, a 0.5% margin for the older card. This is a narrow victory, but it shows the RTX 4090 is not entirely obsolete in legacy API workloads.

Looking at the broader database context, the RTX 5090 sits at the 92nd percentile of all GPUs, with an average benchmark score of 79,842. Its nearest rivals in the database are the NVIDIA Tesla P100 PCIe 16 GB (79,605, 0.3% slower) and the Tesla P100 PCIe 12 GB (79,396, 0.6% slower), while the AMD Radeon Pro Vega 64X is 1.4% faster. The RTX 4090, by contrast, is at the 88th percentile with an average score of 60,347, and its nearest rival is the Intel Arc Pro A60 with a 0% delta. The gap in average score between the two cards is substantial: 79,842 versus 60,347. This means the RTX 5090's average performance across all recorded tests is roughly 32% higher, a figure that aligns with the large wins seen in 3DMark and Geekbench.

Architecture Differences

The two cards are built on different architectures and different chips, which explains much of the performance divergence. The RTX 5090 uses the GB202 chip based on the Blackwell 2.0 architecture, while the RTX 4090 uses the AD102 chip based on Ada Lovelace. Both are manufactured by TSMC on a 5 nm process, so the process node is identical. The transistor counts, however, differ significantly: the RTX 5090 has 92,200 million transistors on a 750 mm² die, while the RTX 4090 has 76,300 million transistors on a 609 mm² die. Interestingly, the transistor density is slightly lower on the newer card: 122.9M per mm² for the RTX 5090 versus 125.3M per mm² for the RTX 4090, indicating the larger die is not purely a density play but rather a matter of adding more silicon.

The execution resources are heavily expanded on the RTX 5090. It has 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The RTX 4090, by comparison, has 16,384 shading units, 512 TMUs, 176 ROPs, 128 RT cores, and 512 tensor cores. The ROP count is identical at 176, meaning the RTX 5090's gains must come from the additional shader and texture hardware rather than pixel output. Indeed, the pixel rate is slightly lower on the RTX 5090 at 423.6 GPixel/s versus 443.5 GPixel/s for the RTX 4090, a consequence of the newer card's lower clock speeds.

The memory subsystem is another major differentiator. The RTX 5090 features 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The RTX 4090 has 24 GB of GDDR6X on a 384-bit bus, with 1.01 TB/s. This is a 78% increase in bandwidth, which is critical for the high-resolution and compute workloads that stress memory throughput. The RTX 5090's memory clock is listed as 1750 MHz with 28 Gbps effective, while the RTX 4090 runs at 1313 MHz with 21 Gbps effective.

Clock speeds tell an interesting story. The RTX 5090 has a base clock of 2017 MHz and a boost clock of 2407 MHz. The RTX 4090 runs higher: 2235 MHz base and 2520 MHz boost. Despite the lower clocks, the RTX 5090 achieves higher overall performance, demonstrating that the architectural and resource advantages outweigh the frequency deficit. The FP32 compute figures confirm this: the RTX 5090 delivers 104.8 TFLOPS versus 82.58 TFLOPS for the RTX 4090. The texture rate is also higher on the newer card at 1,636.8 GTexel/s versus 1,290.2 GTexel/s.

The power and physical specifications differ as well. The RTX 5090 has a TDP of 575 W and a suggested PSU of 950 W, while the RTX 4090 has a TDP of 450 W and a suggested PSU of 850 W. Both use a single 16-pin power connector. The RTX 5090 is a dual-slot card, while the RTX 4090 is triple-slot. Both are 304 mm long and 137 mm tall, but the RTX 5090 is 40 mm wide compared to the RTX 4090's 61 mm, making the newer card notably thinner. The RTX 5090 is also on PCIe 5.0 x16, while the RTX 4090 uses PCIe 4.0 x16.

Connectivity has been updated on the RTX 5090, which offers 1x HDMI 2.1b and 3x DisplayPort 2.1b, while the RTX 4090 offers 1x HDMI 2.1 and 3x DisplayPort 1.4a. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The production status also differs: the RTX 5090 is listed as Active, while the RTX 4090 is End-of-life.

The Verdict

The data points to a clear hierarchy. The RTX 5090 is the faster card in nearly every recorded benchmark, with its most decisive wins coming in modern DirectX 12 and Vulkan workloads. The 99% lead in 3DMark Steel Nomad and the 38.7% lead in Geekbench Vulkan indicate that the Blackwell architecture delivers substantial gains in current-generation rendering paths. For users running the latest games or GPU-accelerated compute tasks, the RTX 5090 is the stronger choice based on the recorded scores.

The RTX 4090, however, retains relevance in one specific area: legacy DirectX 9 performance. Its 0.5% win in that test is narrow, but it shows the older architecture still handles older APIs at parity. The RTX 4090 also has a lower TDP of 450 W versus 575 W, which may be relevant for system builders with power constraints, and its triple-slot design with a 61 mm width suggests it may have a larger cooler, though the database does not specify cooling performance.

The RTX 5090's higher average benchmark score (79,842 versus 60,347) and its higher percentile ranking (92nd versus 88th) reinforce its position as the more capable card overall. The RTX 4090, with an average score of 60,347, is more closely matched with the Intel Arc Pro A60 (a 0% delta) than with the RTX 5090, which sits alongside the Tesla P100 variants. The launch MSRP for the RTX 5090 is 1,999 USD, and for the RTX 4090 it is 1,599 USD.

Specification Differences

The following table summarizes the key differences between the two cards, based solely on the fields where they diverge:

| Specification | NVIDIA GeForce RTX 5090 | NVIDIA GeForce RTX 4090 |

|---|---|---|

| Architecture | Blackwell 2.0 | Ada Lovelace |

| Chip | GB202 | AD102 |

| Transistors | 92,200 million | 76,300 million |

| Die Size | 750 mm² | 609 mm² |

| Transistor Density | 122.9M / mm² | 125.3M / mm² |

| Base Clock | 2017 MHz | 2235 MHz |

| Boost Clock | 2407 MHz | 2520 MHz |

| Memory Clock | 1750 MHz, 28 Gbps effective | 1313 MHz, 21 Gbps effective |

| Memory Size | 32 GB | 24 GB |

| Memory Type | GDDR7 | GDDR6X |

| Memory Bus Width | 512 bit | 384 bit |

| Memory Bandwidth | 1.79 TB/s | 1.01 TB/s |

| Shading Units | 21760 | 16384 |

| TMUs | 680 | 512 |

| RT Cores | 170 | 128 |

| Tensor Cores | 680 | 512 |

| Pixel Rate | 423.6 GPixel/s | 443.5 GPixel/s |

| Texture Rate | 1,636.8 GTexel/s | 1,290.2 GTexel/s |

| FP32 | 104.8 TFLOPS | 82.58 TFLOPS |

| FP16 | 104.8 TFLOPS (1:1) | 82.58 TFLOPS (1:1) |

| TDP | 575 W | 450 W |

| Slot Width | Dual-slot | Triple-slot |

| Suggested PSU | 950 W | 850 W |

| Bus Interface | PCIe 5.0 x16 | PCIe 4.0 x16 |

| Display Outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1b | 1x HDMI 2.1, 3x DisplayPort 1.4a |

| Width | 40 mm | 61 mm |

| Production Status | Active | End-of-life |

| Release Date | 2025-01-29 | 2022-09-19 |

| Predecessor | GeForce 40 | GeForce 30 |

| Successor | GeForce 60 | GeForce 50 |

| Launch MSRP | 1,999 USD | 1,599 USD |

| Average Benchmark Score | 79842 | 60347 |

| Percentile vs All GPUs | 92 | 88 |

FAQ

Q: Which card is faster in 3DMark Steel Nomad DX12?

A: The NVIDIA GeForce RTX 5090, with a score of 18,355 versus 9,223 for the RTX 4090, a 99% difference.

Q: Does the RTX 4090 win any benchmark tests?

A: Yes, the RTX 4090 wins Passmark DirectX 9 with a score of 397 versus 395 for the RTX 5090, a 0.5% margin.

Q: How much memory bandwidth does each card have?

A: The RTX 5090 has 1.79 TB/s from 32 GB of GDDR7 on a 512-bit bus. The RTX 4090 has 1.01 TB/s from 24 GB of GDDR6X on a 384-bit bus.

Q: What are the power requirements for each card?

A: The RTX 5090 has a TDP of 575 W and a suggested PSU of 950 W. The RTX 4090 has a TDP of 450 W and a suggested PSU of 850 W.

Q: Which card has a higher boost clock?

A: The RTX 4090, with a boost clock of 2520 MHz versus 2407 MHz for the RTX 5090.

Q: What is the average benchmark score difference between the two?

A: The RTX 5090 has an average score of 79,842, while the RTX 4090 has an average score of 60,347, making the RTX 5090 about 32% higher on average.