The RTX 5090 lists 21,760 CUDA cores, 32GB of GDDR7 and a 512-bit interface, against 10,752 cores, 16GB and a 256-bit interface on the RTX 5080. In Tom’s Hardware’s 4K ray-tracing tests with upscaling the 5090 reached 48 fps against 29 fps in Alan Wake 2, but those are not native-performance results.
| Specification | NVIDIA GeForce RTX 5090 | NVIDIA GeForce RTX 5080 Founders Edition |
|---|---|---|
| GPU architecture | NVIDIA Blackwell | NVIDIA Blackwell |
| CUDA cores | 21,760 (NVIDIA specification) | 10,752 (NVIDIA specification) |
| Memory | 32GB GDDR7 | 16GB GDDR7 |
| Memory interface | 512-bit | 256-bit |
| Boost clock | 2.41GHz (NVIDIA specification) | 2.62GHz (NVIDIA specification) |
What the specifications show
Both cards use the Blackwell architecture. The RTX 5090 lists 21,760 CUDA cores, 32GB of GDDR7 and a 512-bit memory interface, while the RTX 5080 lists 10,752 CUDA cores, 16GB of GDDR7 and a 256-bit interface. That is roughly double the shader count, memory capacity and interface width.
The RTX 5090’s listed boost clock (2.41GHz) is lower than the RTX 5080’s (2.62GHz), so the difference comes from the extra hardware rather than clock speed. NVIDIA lists a 575W total graphics power for the RTX 5090, with an 850W minimum power supply and four 8-pin connectors or one 600W PCIe Gen 5 cable; check NVIDIA’s RTX 5080 page for that card’s requirements.
What independent reviews add
Against the previous generation, TechSpot measured a 27% average gain for the RTX 5090 over the RTX 4090 at 4K and Tom’s Hardware measured 25% at 4K Ultra, with much smaller gains at 1440p and below. Tom’s Hardware found the RTX 5080 Founders Edition’s conventional gains over the RTX 4080 Super modest in many tests.
Tom’s Hardware’s February 2025 test of both cards used five ray-tracing games at 4K with upscaling. In Alan Wake 2 with full ray tracing the RTX 5090 reached 48 fps against 29 fps for the RTX 5080, and in Indiana Jones and the Great Circle it reached 102 fps against 69 fps. The same article noted the RTX 5080 ran out of VRAM with frame generation at 4K in that game. These are heavy ray-tracing results with upscaling, not native-resolution comparisons, and TWW has not measured either card.
How to decide
The extra memory and shader count matter most for 4K, heavy ray tracing and memory-hungry workloads. Reviewers found the RTX 5090’s gains over the RTX 4090 shrink at lower resolutions, so a 1440p player may see less benefit from the added hardware. Also weigh power supply requirements, case clearance, the cost difference in your market, and whether 16GB meets the games and creative work you actually run.
Frequently asked questions
How much more VRAM does the RTX 5090 have than the RTX 5080?
The RTX 5090 has 32GB of GDDR7 on a 512-bit interface. The RTX 5080 has 16GB of GDDR7 on a 256-bit interface.
Is the RTX 5090 twice as fast as the RTX 5080?
It lists roughly twice the CUDA cores, but frame rates do not scale directly with core count. In Tom’s Hardware’s 4K ray-tracing tests with upscaling it reached 48 fps against 29 fps in Alan Wake 2 and 102 against 69 in Indiana Jones and the Great Circle. Those are not native-resolution results.
How much power does the RTX 5090 need?
NVIDIA lists 575W total graphics power, an 850W minimum power supply (its specification table lists 1,000W required system power), and four 8-pin connectors via the included adapter or one 600W PCIe Gen 5 cable.
Is this a TWW benchmark?
No. The page uses manufacturer specifications and attributed findings from outside reviewers. TWW has not tested either card.
This comparison uses manufacturer specifications and attributed outside review findings. TWW has not tested either card.
NVIDIA GeForce RTX 5090 specifications and sources →
NVIDIA GeForce RTX 5080 Founders Edition specifications and sources →