Quick/Comparisons
The GeForce RTX 5090's superior gaming performance is due to its higher benchmark scores and more efficient architecture, making it the clear winner in this category.
The GeForce RTX 5090's strong workstation performance and better efficiency make it the winner in content creation, despite the GeForce RTX 3080 Ti's slight edge in workstation scores.
The GeForce RTX 5090's capable AI/ML compute and high AI/ML score make it the clear winner in this category, outperforming the GeForce RTX 3080 Ti in AI/ML tasks.
60 | Gaming | 100 |
64 | Workstation | 93 |
46 | AI/ML | 95 |
35 | Energy Efficiency | 49 |
59 | Quick Comparison Final Score | 96 |
| GA102 | Graphics Processor | GB202-300 |
| 10240 | Cores | 21760 |
| 320 / 112 | TMUs / ROPs | 680 / 176 |
| 5565.9 | Blender GPU | 15010.34 |
| 128 | Forza Horizon 5 | 244 |
| 174 | The Witcher 3 | 323 |
| 205 | Counter-Strike 2 | 265 |
| 139 | Far Cry 6 | 202 |
| 108 | Hogwarts Legacy | 207 |
| 161 | Call of Duty | 327 |
| 96 | Ghost of Tsushima | 179 |
| 128 | Cyberpunk 2077 | 229 |
| 219 | Tomb Raider | 299 |
| 151 | Average FPS | 253 |
| 180 | 1080p High | 294 |
| 151 | 1080p Ultra | 253 |
| 117 | 1440p Ultra | 207 |
| 69 | 4K Ultra | 124 |
| Low | Margin of Error | Low |
| 196756 | GB6 Compute Score | 419309 |
| 239.8 img/sec | Background Blur | 333 img/sec |
| 156.8 img/sec | Face Detection | 298.5 img/sec |
| 10.8 Gpixels/sec | Horizon Detection | 21.7 Gpixels/sec |
| 14.8 Gpixels/sec | Edge Detection | 33.7 Gpixels/sec |
| 11.9 Gpixels/sec | Gaussian Blur | 34.5 Gpixels/sec |
| 1.53 Gpixels/sec | Feature Matching | 2.53 Gpixels/sec |
| 889.6 Gpixels/sec | Stereo Matching | 2850 Gpixels/sec |
| 26054.6 FPS | Particle Physics | 59146.9 FPS |
| OpenCL | API | OpenCL |
| 26777 | G3D Mark Score | 38955 |
| 1095 | G2D Mark | 1413 |
| 222 FPS | DirectX 11 | 341 FPS |
| 109 FPS | DirectX 12 | 178 FPS |
| 15025 Ops/s | GPU Compute | 24632 Ops/s |
| 25945 | GB6 ML Single Precision | 53861 |
| 38430 | GB6 ML Half Precision | 81080 |
| 15331 | GB6 ML Quantized | 38938 |
| 9747 | Image Classification (SP) | 18948 |
| 26654 | Image Segmentation (HP) | 55109 |
| 21475 | Image Super Resolution (Q) | 58478 |
| 49813 | Face Detection (HP) | 114685 |
| 135370 | Pose Estimation (Q) | 344729 |
| 2796 | Text Classification (SP) | 4464 |
| 4757 | Machine Translation (HP) | 8184 |
| 12924 | Object Detection (SP) | 27802 |
| 36756 | Depth Estimation (Q) | 85572 |
| 274032 | Style Transfer (SP) | 687080 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| 186 GPixel/s | Pixel Fill Rate | 424 GPixel/s |
| 533 GTexel/s | Texture Fill Rate | 1637 GTexel/s |
| 34.1 TFLOPS | FLOPS (FP32) | 104.8 TFLOPS |
| GDDR6X | Memory Type | GDDR7 |
| 12 GB | Memory Size | 32 GB |
| 1188 MHz | Memory Clock | 1750 MHz |
| 19000 Mbps | Effective Memory Speed | 28000 Mbps |
| 384-bit | Bus | 512-bit |
| No | ECC | No |
| 912 GB/s | Memory Bandwidth | 1792 GB/s |
| PCIe 4.0 x16 | Interface | PCIe 5.0 x16 |
| 350 W | TGP | 575 W |
| Samsung | Manufacturing | TSMC |
| 8 nm | Fabrication Process | 4 nm |
| 628 mm² | Die Size | 750 mm² |
| 28 billion | Transistor Count | 92.2 billion |
| 44.59 MTr/mm² | Transistor Density | 122.93 MTr/mm² |
| 12 | DirectX | 12.2 |
| 1.3 | Vulkan | 1.4 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 3 |
| 8.6 | CUDA | 12 |
| Yes | Ray Tracing | Yes |
| DLSS 2 | DLSS | DLSS 4 |
| 1.4a | DisplayPort | 2.1b |
| Nvidia | Vendor | Nvidia |
| Discrete | Build | Discrete |
| January 3, 2021 | Released | January 7, 2025 |
| Desktop | Case | Desktop |
| Gaming | Purpose | Gaming |
| High-end | Segment | High-end |
| Ampere | Architecture | Blackwell 2.0 |
| GA102 | GPU Codename | GB202-300 |
| -Intel Core i9 12900Kor above | Recommended CPU | -Intel Core Ultra 9 285Kor above |
| 21408 | Steel Nomad Lite Score | 52638 |
| 19632 | Time Spy | 46941 |
| 96597 | Solar Bay | 235224 |
| 13265 | Port Royal | 38113 |
| 47867 | Fire Strike | 87433 |
| 42716 | Wild Life Extreme | 108736 |
| 144955 | Night Raid | 207058 |
| 1365 MHz | Base Clock | 2017 MHz |
| 1665 MHz | Boost Clock | 2407 MHz |
| 10240 | Shading Units | 21760 |
| 320 | Texture Mapping Units (TMUs) | 680 |
| 112 | Render Output Units (ROPs) | 176 |
| 80 | Compute Units (Pipelines) | 170 |
| 320 | Tensor Cores | 680 |
| 80 | Ray-tracing Cores | 170 |
| 128KB per cluster | L1 Cache | 128KB per cluster |
| 6MB shared | L2 Cache | 96MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |
Compare with something else?
Compare GeForce RTX 3080 with another →