NVIDIA GeForce RTX 5090 vs NVIDIA GeForce RTX 5090 SE Comparison

The Verdict

The RTX 5090 and RTX 5090 SE share the same GB202 chip, Blackwell 2.0 architecture, and 5 nm TSMC process, but they are positioned very differently in the database. The RTX 5090 is the fully enabled part, with 21,760 shading units, 32 GB of GDDR7 on a 512-bit bus, and a 575 W TDP. The RTX 5090 SE is a cut-down variant, with 14,080 shading units, 24 GB of GDDR7 on a 384-bit bus, and a 500 W TDP.

The recorded data shows the RTX 5090 delivers massive compute advantages on paper: 104.8 TFLOPS FP32 versus 66.94 TFLOPS, a 56.6% lead. Its memory bandwidth of 1.79 TB/s versus 1.34 TB/s gives it a 33.6% edge. The RTX 5090 also has more ray tracing cores (170 vs 110), tensor cores (680 vs 440), and TMUs (680 vs 440). The RTX 5090 SE is shorter (267 mm vs 304 mm) and lighter in power draw, but it also lacks any benchmark entries in the database.

For buyers, the choice is straightforward from the data. If the workload demands maximum rasterization, ray tracing, or compute throughput, the RTX 5090 is the only option with recorded benchmark scores. The RTX 5090 SE, with an average benchmark score of zero and no rivals listed, has no verified performance data yet. The RTX 5090 sits at the 92nd percentile of all GPUs, while the SE sits at the 50th percentile, which reflects the lack of measured results rather than actual performance. The RTX 5090 SE could be a practical pick for smaller cases or lower power budgets, but the database cannot confirm its real-world standing.

FAQ

Q: What is the core count difference between the RTX 5090 and RTX 5090 SE?

A: The RTX 5090 has 21,760 shading units, 680 TMUs, and 176 ROPs. The RTX 5090 SE has 14,080 shading units, 440 TMUs, and 160 ROPs. The RTX 5090 also carries 170 RT cores and 680 tensor cores versus 110 RT cores and 440 tensor cores on the SE.

Q: How much memory does each card have and what is the bus width?

A: The RTX 5090 has 32 GB of GDDR7 on a 512-bit bus, yielding 1.79 TB/s bandwidth. The RTX 5090 SE has 24 GB of GDDR7 on a 384-bit bus, yielding 1.34 TB/s bandwidth. Both run the same 1750 MHz memory clock (28 Gbps effective).

Q: Are the physical dimensions different?

A: Yes. The RTX 5090 measures 304 mm in length, 137 mm in height, and 40 mm in width. The RTX 5090 SE measures 267 mm in length, 111 mm in height, and 40 mm in width. Both are dual-slot cards.

Q: What are the power requirements for each card?

A: The RTX 5090 has a 575 W TDP and suggests a 950 W PSU. The RTX 5090 SE has a 500 W TDP and suggests a 900 W PSU. Both use a single 16-pin power connector.

Q: What is the difference in clock speeds?

A: The RTX 5090 runs a 2017 MHz base clock and 2407 MHz boost clock. The RTX 5090 SE runs a 1740 MHz base clock and 2377 MHz boost clock. The SE has a lower base by 277 MHz and a lower boost by 30 MHz.

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

A: No. The RTX 5090 SE has an empty benchmark list, an average score of zero, and no nearest rivals. The RTX 5090 has ten recorded benchmark scores, an average of 79,842, and four nearest rivals.

Architecture Differences

Both cards are built on the GB202 chip, which is the largest die in the GeForce 50-series. The architecture is Blackwell 2.0, fabricated on TSMC's 5 nm process. The die measures 750 mm² and contains 92,200 million transistors, with a density of 122.9 million transistors per square millimeter. These physical attributes are identical across both cards.

The functional differences come from how many of the chip's resources are enabled. The RTX 5090 activates the full complement of shaders, TMUs, ROPs, RT cores, and tensor cores. The RTX 5090 SE disables a significant portion of those units. Specifically, the SE loses 7,680 shading units, 240 TMUs, 16 ROPs, 60 RT cores, and 240 tensor cores relative to the RTX 5090. That is a 35.3% reduction in shading units and a 35.3% reduction in tensor cores.

The memory subsystem also differs. The RTX 5090 uses a 512-bit interface with 32 GB of GDDR7. The RTX 5090 SE uses a 384-bit interface with 24 GB of GDDR7. The memory clock is identical at 1750 MHz, but the narrower bus on the SE reduces peak bandwidth by 0.45 TB/s. The SE also has fewer ROPs, which directly impacts pixel fill rate: 380.3 GPixel/s versus 423.6 GPixel/s on the RTX 5090.

The architecture feature set is otherwise the same. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both use PCIe 5.0 x16 and have identical display outputs: 1x HDMI 2.1b and 3x DisplayPort 2.1b. The power delivery is also identical in connector type, a single 16-pin, though the TDP differs by 75 W.

Specification Differences

The following fields differ between the two cards:

• Shading Units: 21,760 (RTX 5090) vs 14,080 (RTX 5090 SE)
• TMUs: 680 vs 440
• ROPs: 176 vs 160
• RT Cores: 170 vs 110
• Tensor Cores: 680 vs 440
• Base Clock: 2017 MHz vs 1740 MHz
• Boost Clock: 2407 MHz vs 2377 MHz
• Memory Size: 32 GB vs 24 GB
• Memory Bus Width: 512 bit vs 384 bit
• Memory Bandwidth: 1.79 TB/s vs 1.34 TB/s
• Pixel Rate: 423.6 GPixel/s vs 380.3 GPixel/s
• Texture Rate: 1,636.8 GTexel/s vs 1,045.9 GTexel/s
• FP32 Compute: 104.8 TFLOPS vs 66.94 TFLOPS
• FP16 Compute: 104.8 TFLOPS vs 66.94 TFLOPS (both 1:1 ratio)
• TDP: 575 W vs 500 W
• Suggested PSU: 950 W vs 900 W
• Length: 304 mm vs 267 mm
• Height: 137 mm vs 111 mm
• Release Date: 2025-01-29 vs 2025-12-31
• Launch MSRP: 1,999 USD vs 1,499 USD

Both cards share the same process node, foundry, die size, transistor count, transistor density, memory type, memory clock, slot width, power connector type, bus interface, display outputs, and API support. The production status for both is Active.

Head-to-Head Benchmarks

The database contains no head-to-head benchmark entries between the RTX 5090 and RTX 5090 SE. The head-to-head array is empty, and the wins counter for each card is zero. This means no direct comparison of measured performance exists in the recorded data.

However, the RTX 5090 has its own benchmark results, which provide a reference point. In 3DMark Steel Nomad (DX12), it scores 18,355. In Geekbench OpenCL, it scores 334,370. In Geekbench Vulkan, it scores 376,728. Passmark results include 39,650 in G3D, 26,756 in GPU Compute, 1,413 in G2D, 395 in DirectX 9, 341 in DirectX 11, 226 in DirectX 10, and 185 in DirectX 12. The average benchmark score across all entries is 79,842.

The RTX 5090 SE has no recorded scores, so there is no measured basis to compare its performance in any specific test. The only quantitative comparison possible is derived from the specification differences, which show a 56.6% FP32 compute advantage for the RTX 5090 and a 33.6% bandwidth advantage.

The nearest rivals for the RTX 5090 give context to its standing. The NVIDIA Tesla P100 PCIe 16 GB scores 79,605, which is 0.3% lower. The Tesla P100 PCIe 12 GB scores 79,396, 0.6% lower. The AMD Radeon RX 6850M XT scores 78,940, 1.1% lower. The AMD Radeon Pro Vega 64X scores 80,959, which is 1.4% higher. These deltas are all within a narrow band, meaning the RTX 5090's average score sits near those cards. The RTX 5090's percentile rank is 92, meaning it outperforms 92% of all GPUs in the database.

Where Each One Wins

The RTX 5090 wins in every measurable specification category. It has more shading units, more TMUs, more ROPs, more RT cores, more tensor cores, higher clocks, more memory, a wider bus, more bandwidth, higher pixel rate, higher texture rate, and higher FP32 and FP16 compute. It also has a higher average benchmark score (79,842 vs 0) and a higher percentile rank (92 vs 50).

The RTX 5090 SE wins in the categories that matter for physical integration and power draw. It is shorter by 37 mm and lower by 26 mm, which allows it to fit in smaller chassis. Its TDP is 75 W lower, and its suggested PSU is 50 W lower, which reduces system power requirements. Its launch MSRP is 500 USD lower, though the database does not analyze pricing. The SE also has a later release date, but that carries no performance implication.

For a builder focused on maximum throughput, the RTX 5090 is the only card with verified results. The data shows it delivers 1,636.8 GTexel/s texture rate versus 1,045.9 GTexel/s on the SE, and 423.6 GPixel/s versus 380.3 GPixel/s. Those differences translate directly to higher frame rates in texture-bound and pixel-bound workloads, assuming the rest of the system is not the bottleneck.

For a builder constrained by case size or power budget, the RTX 5090 SE is the more practical choice, but the database offers no performance confirmation. Its 500 W TDP and 267 mm length are clear advantages in compact builds. The identical memory clock and API support mean the SE will not lose any feature compatibility, but its narrower bus and fewer cores will reduce peak output in heavy workloads.

The RTX 5090 SE also has the same transistor count and die size as the RTX 5090, which means it uses the same physical chip with disabled units. This could lead to lower effective thermals per enabled unit, but the database does not include thermal or efficiency measurements.

In summary, the RTX 5090 wins on every measured and specified performance metric. The RTX 5090 SE wins on physical footprint, power draw, and launch MSRP. Without benchmark data for the SE, any claim of real-world performance parity would be unsupported. The RTX 5090 is the card with proven results.