NVIDIA GeForce RTX 4070 vs NVIDIA GeForce RTX 5090 SE Comparison

NVIDIA GeForce RTX 4070 and NVIDIA GeForce RTX 5090 SE represent two distinct generations of NVIDIA graphics hardware, with the former built on the Ada Lovelace architecture and the latter on Blackwell 2.0. The RTX 4070 is an end-of-life product from the GeForce 40-series, while the RTX 5090 SE is an active GeForce 50-series part with a substantially different specification sheet. The database contains a full benchmark profile for the RTX 4070, including an average benchmark score of 37,648 and a percentile rank of 81 among all GPUs, while the RTX 5090 SE has no recorded benchmark scores and an average score of 0, making direct numerical comparison impossible. This analysis relies on the architectural and specification differences recorded in the database to interpret what the data implies about relative performance.

Where Each One Wins

The RTX 4070 wins in the only category where measured data exists: actual benchmark results. Its recorded scores include 38,54 in 3DMark Steel Nomad DX12, 154,858 in Geekbench OpenCL, 174,152 in Geekbench Vulkan, and 26,927 in Passmark G3D. These results place it in the 81st percentile of all GPUs, with an average benchmark score of 37,648. The nearest rivals in the database are the NVIDIA Tesla P4 with an average score of 37,628 (0.1% delta), the AMD Radeon RX Vega 56 at 37,507 (0.4% delta), the NVIDIA GeForce RTX 4080 Mobile at 38,135 (-1.3% delta), and the AMD Radeon PRO W6400 at 37,157 (1.3% delta). This indicates that the RTX 4070 sits in a tightly clustered performance band, within roughly 1.3% of its nearest competitors, and its average score is essentially at parity with those cards.

The RTX 5090 SE wins in raw specification comparisons, which is the only field where it has data. It carries 14,080 shading units versus 5,888 for the RTX 4070, a 2.39x increase. Its FP32 compute is 66.94 TFLOPS compared to 29.15 TFLOPS, more than double. The RTX 5090 SE also has 110 RT cores versus 46, 440 tensor cores versus 184, 440 TMUs versus 184, and 160 ROPs versus 64. These figures suggest that in compute-heavy workloads, the RTX 5090 SE would likely dominate, but the database contains no benchmark results to confirm this. The RTX 4070 has a production status of end-of-life, while the RTX 5090 SE is active, so the newer card represents the current generation's intended performance tier.

The use-case split is therefore: the RTX 4070 has verified performance data across multiple tests, while the RTX 5090 SE has none. For workloads like DirectX 12 gaming, OpenCL compute, Vulkan rendering, and DirectX 10/11/12 legacy tests, the RTX 4070 has concrete scores. For workloads that stress raw throughput, memory bandwidth, or ray tracing capability, the RTX 5090 SE's specifications indicate a substantial theoretical advantage, but no measured confirmation exists in the database.

Architecture Differences

The two cards use different chips and architectures. The RTX 4070 uses the AD104 chip on the Ada Lovelace architecture, fabricated by TSMC on a 5 nm process. The RTX 5090 SE uses the GB202 chip on the Blackwell 2.0 architecture, also on a 5 nm TSMC process. Both have the same transistor density at 121.8M per mm² for the RTX 4070 and 122.9M per mm² for the RTX 5090 SE, a negligible difference. However, the actual transistor counts differ dramatically: the RTX 4070 has 35,800 million transistors on a 294 mm² die, while the RTX 5090 SE has 92,200 million transistors on a 750 mm² die. This means the RTX 5090 SE has 2.58x more transistors and a die that is 2.55x larger.

The memory subsystems are entirely different. The RTX 4070 uses 12 GB of GDDR6X on a 192-bit bus, yielding 504.2 GB/s bandwidth. The RTX 5090 SE uses 24 GB of GDDR7 on a 384-bit bus, yielding 1.34 TB/s bandwidth. That is 2.66x the bandwidth and 2x the memory capacity. The memory clock rates also differ: the RTX 4070 runs at 1313 MHz with 21 Gbps effective, while the RTX 5090 SE runs at 1750 MHz with 28 Gbps effective.

Compute resources scale substantially. The RTX 5090 SE has 14,080 shading units versus 5,888, a 2.39x increase. TMUs go from 184 to 440 (2.39x), ROPs from 64 to 160 (2.5x), RT cores from 46 to 110 (2.39x), and tensor cores from 184 to 440 (2.39x). Pixel rate rises from 158.4 GPixel/s to 380.3 GPixel/s, and texture rate from 455.4 GTexel/s to 1,045.9 GTexel/s. FP32 and FP16 both scale from 29.15 TFLOPS to 66.94 TFLOPS, a 2.3x improvement.

Clock speeds tell a different story. The RTX 4070 has a base clock of 1920 MHz and a boost clock of 2475 MHz. The RTX 5090 SE has a lower base clock of 1740 MHz and a boost clock of 2377 MHz. So the RTX 4070 actually runs at higher clocks, by 180 MHz at base and 98 MHz at boost. The performance advantage of the RTX 5090 SE comes from having far more compute units, not from higher clocks.

The bus interface also differs: the RTX 4070 uses PCIe 4.0 x16, while the RTX 5090 SE uses PCIe 5.0 x16. Display outputs are similar in count but different in version: the RTX 4070 has 1x HDMI 2.1 and 3x DisplayPort 1.4a, while the RTX 5090 SE has 1x HDMI 2.1b and 3x DisplayPort 2.1b. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Power specifications are notably different. The RTX 4070 has a TDP of 200 W and a suggested PSU of 550 W, while the RTX 5090 SE has a TDP of 500 W and a suggested PSU of 900 W. Both use a single 16-pin power connector and are dual-slot cards. Physical dimensions are close: the RTX 4070 is 240 mm long, 110 mm tall, and 40 mm wide, while the RTX 5090 SE is 267 mm long, 111 mm tall, and 40 mm wide.

Head-to-Head Benchmarks

The database contains no head-to-head benchmark entries for these two cards. The winsA and winsB fields are both 0, and the headToHeadBenchmarks array is empty. This means there is no direct measured comparison available. The only performance data belongs to the RTX 4070, which has ten benchmark scores recorded. The RTX 5090 SE has an empty benchmarks array, so its performance cannot be quantified from the database.

What the data does show is the RTX 4070's absolute scores. Its strongest result is in Geekbench Vulkan at 174,152, followed by Geekbench OpenCL at 154,858. In Passmark tests, it scores 26,927 in G3D, 14,720 in GPU Compute, 1,164 in G2D, 320 in DirectX 9, 244 in DirectX 11, 139 in DirectX 10, and 103 in DirectX 12. The 3DMark Steel Nomad DX12 result is 3,854. These numbers show a pattern: the RTX 4070 performs better in compute-oriented APIs like Vulkan and OpenCL than in legacy DirectX tests, and its DirectX 12 score in Passmark is notably low at 103 compared to its DirectX 11 score of 244.

The RTX 5090 SE's specifications imply it would outperform the RTX 4070 in every measured category if benchmarks existed, based on the 2.3x to 2.66x scaling in compute units, memory bandwidth, and pixel/texture rates. However, the database does not confirm this, and the lower boost clock of the RTX 5090 SE could affect real-world results. The RTX 4070's higher boost clock of 2475 MHz versus 2377 MHz means that in lightly threaded or clock-bound scenarios, the older card could hold an advantage, but the RTX 5090 SE's massive unit count would likely overwhelm that in parallel workloads.

The Verdict

From the recorded data, the RTX 4070 is the only card with verified performance. It sits in the 81st percentile of all GPUs, with an average score of 37,648, and its nearest rivals are all within 1.3% of that figure. This indicates that the RTX 4070 delivers consistent, mid-to-high-range performance across the benchmarks in the database, with a slight edge in Vulkan and OpenCL compute tasks.

The RTX 5090 SE has no benchmark scores, so no performance verdict can be drawn from measurements. Its specifications, however, show a card with 2.58x more transistors, 2.39x more shading units, 2.66x more memory bandwidth, and 2.3x more FP32 throughput. The database records its average benchmark score as 0, which means it has not been tested or has not produced results. Its percentile rank of 50 is the default value for untested cards, not a performance indicator.

For users who need proven performance data, the RTX 4070 is the choice, because its scores are documented. For users who prioritize the architectural and specification advantages of the Blackwell 2.0 generation, the RTX 5090 SE offers a larger memory pool (24 GB vs 12 GB), a wider 384-bit bus, GDDR7 memory, and more than double the compute resources. The RTX 5090 SE also has a newer bus interface (PCIe 5.0 vs 4.0) and newer display outputs (DisplayPort 2.1b vs 1.4a). The data cannot say which is faster in practice, so the verdict is strictly that the RTX 4070 has measured results and the RTX 5090 SE has superior specifications on paper.

FAQ

Q: What is the average benchmark score for the RTX 4070?

A: The RTX 4070 has an average benchmark score of 37,648 across its recorded benchmarks, placing it in the 81st percentile of all GPUs.

Q: Does the RTX 5090 SE have any benchmark scores in the database?

A: No, the RTX 5090 SE has an empty benchmarks array, and its average benchmark score is recorded as 0.

Q: How do the FP32 compute figures compare between the two cards?

A: The RTX 4070 has 29.15 TFLOPS of FP32 compute, while the RTX 5090 SE has 66.94 TFLOPS, which is a 2.3x difference.

Q: What are the memory capacities and bandwidths?

A: The RTX 4070 has 12 GB of GDDR6X on a 192-bit bus with 504.2 GB/s bandwidth. The RTX 5090 SE has 24 GB of GDDR7 on a 384-bit bus with 1.34 TB/s bandwidth.

Q: Which card has a higher boost clock?

A: The RTX 4070 has a boost clock of 2475 MHz, which is higher than the RTX 5090 SE's boost clock of 2377 MHz.

Q: What are the TDP and suggested PSU requirements?

A: The RTX 4070 has a TDP of 200 W and a suggested PSU of 550 W. The RTX 5090 SE has a TDP of 500 W and a suggested PSU of 900 W.

Specification Differences

The two cards differ in the following recorded fields:

• Chip: AD104 (RTX 4070) vs GB202 (RTX 5090 SE)
• Architecture: Ada Lovelace vs Blackwell 2.0
• Transistors: 35,800 million vs 92,200 million
• Die Size: 294 mm² vs 750 mm²
• Base Clock: 1920 MHz vs 1740 MHz
• Boost Clock: 2475 MHz vs 2377 MHz
• Memory Clock: 1313 MHz (21 Gbps effective) vs 1750 MHz (28 Gbps effective)
• Memory Size: 12 GB vs 24 GB
• Memory Type: GDDR6X vs GDDR7
• Memory Bus Width: 192 bit vs 384 bit
• Memory Bandwidth: 504.2 GB/s vs 1.34 TB/s
• Shading Units: 5,888 vs 14,080
• TMUs: 184 vs 440
• ROPs: 64 vs 160
• RT Cores: 46 vs 110
• Tensor Cores: 184 vs 440
• Pixel Rate: 158.4 GPixel/s vs 380.3 GPixel/s
• Texture Rate: 455.4 GTexel/s vs 1,045.9 GTexel/s
• FP32: 29.15 TFLOPS vs 66.94 TFLOPS
• FP16: 29.15 TFLOPS vs 66.94 TFLOPS
• TDP: 200 W vs 500 W
• Suggested PSU: 550 W vs 900 W
• Bus Interface: PCIe 4.0 x16 vs PCIe 5.0 x16
• Display Outputs: 1x HDMI 2.1, 3x DisplayPort 1.4a vs 1x HDMI 2.1b, 3x DisplayPort 2.1b
• Dimensions: 240 mm length, 110 mm height, 40 mm width vs 267 mm length, 111 mm height, 40 mm width
• Production Status: End-of-life vs Active
• Release Date: 2023-04-11 vs 2025-12-31
• Launch MSRP: 599 USD vs 1,499 USD
• Benchmarks: 10 recorded scores vs none
• Average Benchmark Score: 37,648 vs 0
• Percentile vs All GPUs: 81 vs 50