Quick/Comparisons
The GeForce RTX 4090's superior gaming performance, with a 22-point lead, makes it the clear winner in this use case. Its 92 score is significantly higher than the GeForce RTX 4070 Ti SUPER's 70.
Although the GeForce RTX 4070 Ti SUPER has a 19-point lead in workstation performance, the GeForce RTX 4090's strong workstation performance and 19-point lead in this use case make it the winner. Its 91 score is significantly higher than the GeForce RTX 4070 Ti SUPER's 72.
The GeForce RTX 4090's 17-point lead in AI/ML compute makes it the winner in this use case. Its 76 score is significantly higher than the GeForce RTX 4070 Ti SUPER's 59.
70 | Gaming | 92 |
72 | Workstation | 91 |
59 | AI/ML | 76 |
46 | Energy Efficiency | 43 |
69 | Quick Comparison Final Score | 89 |
| AD103 | Graphics Processor | AD102 |
| 8448 | Cores | 16384 |
| 264 / 96 | TMUs / ROPs | 512 / 176 |
| 7089.35 | Blender GPU | 11688.48 |
| 150 | Forza Horizon 5 | 203 |
| 198 | The Witcher 3 | 270 |
| 220 | Counter-Strike 2 | 245 |
| 152 | Far Cry 6 | 183 |
| 118 | Hogwarts Legacy | 165 |
| 188 | Call of Duty | 269 |
| 111 | Ghost of Tsushima | 150 |
| 155 | Cyberpunk 2077 | 204 |
| 249 | Tomb Raider | 283 |
| 171 | Average FPS | 219 |
| 200 | 1080p High | 256 |
| 171 | 1080p Ultra | 219 |
| 134 | 1440p Ultra | 176 |
| 79 | 4K Ultra | 107 |
| Low | Margin of Error | Low |
| 237068 | GB6 Compute Score | 316301 |
| 331.7 img/sec | Background Blur | 318.9 img/sec |
| 211.6 img/sec | Face Detection | 222.8 img/sec |
| 9.63 Gpixels/sec | Horizon Detection | 14.2 Gpixels/sec |
| 13.7 Gpixels/sec | Edge Detection | 21 Gpixels/sec |
| 13.9 Gpixels/sec | Gaussian Blur | 23.9 Gpixels/sec |
| 2.15 Gpixels/sec | Feature Matching | 2.38 Gpixels/sec |
| 1210 Gpixels/sec | Stereo Matching | 1950 Gpixels/sec |
| 33566 FPS | Particle Physics | 48090.4 FPS |
| OpenCL | API | OpenCL |
| 31798 | G3D Mark Score | 38071 |
| 1240 | G2D Mark | 1303 |
| 277 FPS | DirectX 11 | 324 FPS |
| 119 FPS | DirectX 12 | 151 FPS |
| 18264 Ops/s | GPU Compute | 26193 Ops/s |
| 32793 | GB6 ML Single Precision | 42962 |
| 47412 | GB6 ML Half Precision | 59636 |
| 24566 | GB6 ML Quantized | 31670 |
| 12970 | Image Classification (SP) | 15600 |
| 26364 | Image Segmentation (HP) | 36657 |
| 37217 | Image Super Resolution (Q) | 48114 |
| 58720 | Face Detection (HP) | 78111 |
| 177604 | Pose Estimation (Q) | 289639 |
| 3815 | Text Classification (SP) | 4074 |
| 5829 | Machine Translation (HP) | 6349 |
| 15675 | Object Detection (SP) | 21186 |
| 62447 | Depth Estimation (Q) | 75599 |
| 357918 | Style Transfer (SP) | 601428 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| 251 GPixel/s | Pixel Fill Rate | 444 GPixel/s |
| 689 GTexel/s | Texture Fill Rate | 1290 GTexel/s |
| 44.1 TFLOPS | FLOPS (FP32) | 82.6 TFLOPS |
| GDDR6X | Memory Type | GDDR6X |
| 16 GB | Memory Size | 24 GB |
| 1313 MHz | Memory Clock | 2625 MHz |
| 21000 Mbps | Effective Memory Speed | 21000 Mbps |
| 256-bit | Bus | 384-bit |
| No | ECC | No |
| 672.3 GB/s | Memory Bandwidth | 1010 GB/s |
| PCIe 4.0 x16 | Interface | PCIe 4.0 x16 |
| 285 W | TGP | 450 W |
| TSMC | Manufacturing | TSMC |
| 5 nm | Fabrication Process | 5 nm |
| 379 mm² | Die Size | 609 mm² |
| 45 billion | Transistor Count | 76.3 billion |
| 118.73 MTr/mm² | Transistor Density | 125.29 MTr/mm² |
| 12 | DirectX | 12 |
| 1.3 | Vulkan | 1.3 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 3 |
| 8.9 | CUDA | 8.9 |
| Yes | Ray Tracing | Yes |
| DLSS 3 | DLSS | DLSS 3 |
| 1.4a | DisplayPort | 1.4a |
| Nvidia | Vendor | Nvidia |
| Discrete | Build | Discrete |
| January 24, 2024 | Released | September 20, 2022 |
| Desktop | Case | Desktop |
| Gaming | Purpose | Gaming |
| High-end | Segment | High-end |
| Ada Lovelace | Architecture | Ada Lovelace |
| AD103 | GPU Codename | AD102 |
| -Intel Core i7 14700Kor above | Recommended CPU | -Intel Core i9 14900Kor above |
| 25741 | Steel Nomad Lite Score | 42169 |
| 24222 | Time Spy | 36311 |
| 116542 | Solar Bay | 187386 |
| 15802 | Port Royal | 26128 |
| 56340 | Fire Strike | 72387 |
| 49901 | Wild Life Extreme | 85230 |
| 176609 | Night Raid | 195880 |
| 2340 MHz | Base Clock | 2235 MHz |
| 2610 MHz | Boost Clock | 2520 MHz |
| 8448 | Shading Units | 16384 |
| 264 | Texture Mapping Units (TMUs) | 512 |
| 96 | Render Output Units (ROPs) | 176 |
| 66 | Compute Units (Pipelines) | 128 |
| 264 | Tensor Cores | 512 |
| 66 | Ray-tracing Cores | 128 |
| 128KB per cluster | L1 Cache | 128KB per cluster |
| 48MB shared | L2 Cache | 72MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |
Compare with something else?
Compare GeForce RTX 4070 with another →