NVIDIA GeForce RTX 4070 SUPER vs NVIDIA GeForce RTX 5090 SE Comparison

Head-to-Head Benchmarks

The recorded data presents an unusual situation: the RTX 4070 SUPER has a full set of benchmark results, while the RTX 5090 SE has no recorded benchmark scores at all. The database shows zero entries for the 5090 SE across every test, and its average benchmark score sits at 0. This makes a direct numerical comparison impossible in the traditional sense. The 4070 SUPER, by contrast, has measurable results across ten separate tests, giving it a percentile ranking of 83 among all GPUs.

The 4070 SUPER's strongest recorded result comes from 3DMark Steel Nomad DX12, where it scores 4627. Its Geekbench Vulkan score reaches 205624, while the OpenCL score lands at 172795. PassMark G3D shows 29995, and the GPU compute test produces 17108. These numbers place the 4070 SUPER within a tight cluster of rivals, which is revealing. The closest competitor is the Quadro M6000 24 GB with an average score of 43262, a delta of -0.1 percent. The RTX 5050 Mobile sits at 43268, also -0.1 percent, and the Quadro M6000 reaches 43301, a -0.2 percent difference. The RTX 4090 Mobile leads this group at 43667, putting it 1 percent ahead of the 4070 SUPER.

Since the 5090 SE has no benchmark data, the head-to-head comparison must rely on architectural specifications rather than measured performance. The clock speeds differ notably: the 4070 SUPER has a base clock of 1980 MHz and a boost of 2475 MHz, while the 5090 SE runs lower at 1740 MHz base and 2377 MHz boost. However, the 5090 SE compensates with far more compute resources. Its shading units number 14080 versus 7168 for the 4070 SUPER, and its texture mapping units reach 440 versus 224. The raster operations units double from 80 to 160. These are substantial differences that suggest the 5090 SE should outperform in raw throughput, but without benchmark confirmation, the data cannot verify this expectation.

The Verdict

The verdict from the data is straightforward but incomplete. The 4070 SUPER has demonstrated performance across multiple tests, earning its 83rd percentile ranking. The 5090 SE, with zero recorded benchmarks and a 50th percentile placeholder, cannot be evaluated on measured merit. Users who need verified results today should look to the 4070 SUPER, as its performance profile is fully documented. The 5090 SE's active production status and newer generation suggest it is the intended successor, but the absence of scores means any performance claims remain speculative.

The data indicates the 4070 SUPER is end-of-life, while the 5090 SE is active. This status difference matters for long-term availability and driver support. The 4070 SUPER's predecessor was the GeForce 30 series, and its successor is the GeForce 50 series, which is exactly where the 5090 SE sits. The 5090 SE's predecessor is the GeForce 40 series, confirming the generational handoff. For users prioritizing proven results, the 4070 SUPER is the only option with data. For users prioritizing the newest architecture, the 5090 SE is the choice, but they must accept the lack of recorded benchmarks.

Architecture Differences

The two GPUs come from different architectures and use different chips. The 4070 SUPER uses the AD104 chip built on Ada Lovelace architecture, while the 5090 SE uses the GB202 chip on Blackwell 2.0. Both are manufactured by TSMC on a 5 nm process, and the transistor densities are nearly identical: 121.8 million transistors per square millimeter for the 4070 SUPER and 122.9 million for the 5090 SE. The die sizes differ substantially, however. The 4070 SUPER has a die size of 294 square millimeters and 35,800 million transistors. The 5090 SE has a die size of 750 square millimeters and 92,200 million transistors. The larger die allows for nearly triple the transistor count.

Memory architecture shows clear generational progression. The 4070 SUPER uses 12 GB of GDDR6X on a 192-bit bus, delivering 504.2 GB/s bandwidth. The 5090 SE uses 24 GB of GDDR7 on a 384-bit bus, delivering 1.34 TB/s. That is more than double the memory capacity and roughly 2.7 times the bandwidth. The effective memory clock also increases from 21 Gbps to 28 Gbps. Ray tracing cores jump from 56 to 110, and tensor cores from 224 to 440. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

The process nodes are the same, but the architectures differ in compute organization. The 5090 SE has double the shading units, nearly double the TMUs, and double the ROPs. Its pixel rate is 380.3 GPixel/s versus 198.0 GPixel/s for the 4070 SUPER. The texture rate is 1,045.9 GTexel/s versus 554.4 GTexel/s. FP32 and FP16 compute both reach 66.94 TFLOPS on the 5090 SE, compared to 35.48 TFLOPS on the 4070 SUPER. These are exactly 1:1 ratios for both cards, indicating no half-rate FP16 penalty.

FAQ

Q: How much faster is the RTX 5090 SE than the RTX 4070 SUPER?

A: The database contains no benchmark scores for the RTX 5090 SE, so no measured performance comparison is possible. The 4070 SUPER has ten recorded test results, while the 5090 SE has none.

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

A: The RTX 4070 SUPER 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: Do both cards use the same manufacturing process?

A: Yes, both are built by TSMC on a 5 nm process. The transistor density is similar (121.8M per mm² for the 4070 SUPER, 122.9M per mm² for the 5090 SE), but the die sizes differ: 294 mm² versus 750 mm².

Q: What are the power requirements?

A: The RTX 4070 SUPER has a TDP of 220 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. Both use a single 16-pin power connector and are dual-slot cards.

Q: Which card has better driver and API support?

A: Both cards support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The 4070 SUPER uses PCIe 4.0 x16, while the 5090 SE uses PCIe 5.0 x16.

Q: What is the release status of each card?

A: The RTX 4070 SUPER is end-of-life and was released in January 2024. The RTX 5090 SE is active in production and has a release date of December 2025.

Where Each One Wins

The RTX 4070 SUPER wins in the only category where data exists: measured benchmarks. It has scores across 3DMark, Geekbench OpenCL, Geekbench Vulkan, and multiple PassMark tests. Its 83rd percentile ranking among all GPUs is established. Its nearest rivals are all older or lower-tier products, and it sits within 1 percent of the RTX 4090 Mobile's average score. For any workload represented by these tests, the 4070 SUPER has verified results.

The RTX 5090 SE wins in every architectural category based on the specification data. It has more than double the shading units (14080 versus 7168), double the TMUs (440 versus 224), double the ROPs (160 versus 80), and nearly double the ray tracing cores (110 versus 56). Its memory subsystem is larger and faster: 24 GB GDDR7 versus 12 GB GDDR6X, with 1.34 TB/s versus 504.2 GB/s bandwidth. Its compute throughput is 66.94 TFLOPS in both FP32 and FP16, versus 35.48 TFLOPS for the 4070 SUPER. Its pixel rate and texture rate are both roughly 1.9 times higher.

The 5090 SE also wins on interface standards. It uses PCIe 5.0 x16, while the 4070 SUPER uses PCIe 4.0 x16. Its display outputs include HDMI 2.1b and three DisplayPort 2.1b ports, whereas the 4070 SUPER has HDMI 2.1 and three DisplayPort 1.4a ports. The 5090 SE is also thinner, at 40 mm width versus 42 mm, and slightly shorter in height, at 111 mm versus 112 mm. Length is identical at 267 mm.

Specification Differences

The following fields differ between the two cards. The RTX 4070 SUPER has a base clock of 1980 MHz; the RTX 5090 SE has 1740 MHz. Boost clocks are 2475 MHz and 2377 MHz respectively. Memory clocks are 1313 MHz (21 Gbps effective) for the 4070 SUPER and 1750 MHz (28 Gbps effective) for the 5090 SE. Memory size doubles from 12 GB to 24 GB. Memory type changes from GDDR6X to GDDR7. Bus width goes from 192-bit to 384-bit. Bandwidth increases from 504.2 GB/s to 1.34 TB/s.

Shading units go from 7168 to 14080. TMUs go from 224 to 440. ROPs go from 80 to 160. Ray tracing cores go from 56 to 110. Tensor cores go from 224 to 440. Pixel rate goes from 198.0 GPixel/s to 380.3 GPixel/s. Texture rate goes from 554.4 GTexel/s to 1,045.9 GTexel/s. FP32 and FP16 both go from 35.48 TFLOPS to 66.94 TFLOPS. TDP goes from 220 W to 500 W. Suggested PSU goes from 550 W to 900 W.

Transistor count goes from 35,800 million to 92,200 million. Die size goes from 294 mm² to 750 mm². Transistor density is marginally different: 121.8M per mm² versus 122.9M per mm². The chip changes from AD104 to GB202. Architecture changes from Ada Lovelace to Blackwell 2.0. Generation changes from GeForce 40 to GeForce 50. The bus interface changes from PCIe 4.0 x16 to PCIe 5.0 x16. Display outputs change from HDMI 2.1 with three DisplayPort 1.4a to HDMI 2.1b with three DisplayPort 2.1b.

Dimensions differ slightly. Height changes from 112 mm to 111 mm. Width changes from 42 mm to 40 mm. Length stays at 267 mm. Production status changes from end-of-life to active. Release date shifts from January 2024 to December 2025. Predecessor changes from GeForce 30 to GeForce 40. Successor changes from GeForce 50 to GeForce 60. Launch MSRP is 599 USD for the 4070 SUPER and 1,499 USD for the 5090 SE.