Quick/Comparisons
41 | Gaming | 56 |
40 | Workstation | 67 |
38 | AI/ML |
Compare with something else?
Compare GeForce RTX 4070 with another →53 |
56 | Energy Efficiency | 49 |
43 | Quick Comparison Final Score | 57 |
| AD106 (GN21-X6) | Graphics Processor | GB203-300-A1 |
| 4608 | Cores | 7680 |
| 144 / 48 | TMUs / ROPs | 240 / 96 |
| 3587.08 | Blender GPU | 5699.95 |
| 90 | Forza Horizon 5 | 114 |
| 132 | The Witcher 3 | 157 |
| 178 | Counter-Strike 2 | 204 |
| 112 | Far Cry 6 | 132 |
| 75 | Hogwarts Legacy | 96 |
| 104 | Call of Duty | 143 |
| 68 | Ghost of Tsushima | 87 |
| 87 | Cyberpunk 2077 | 118 |
| 139 | Tomb Raider | 197 |
| 109 | Average FPS | 139 |
| 131 | 1080p High | 166 |
| 109 | 1080p Ultra | 139 |
| 81 | 1440p Ultra | 106 |
| 46 | 4K Ultra | 63 |
| Low | Margin of Error | High |
| 115081 | GB6 Compute Score | 197909 |
| 158.5 img/sec | Background Blur | 267.1 img/sec |
| 116.5 img/sec | Face Detection | 165.5 img/sec |
| 4.35 Gpixels/sec | Horizon Detection | 8.72 Gpixels/sec |
| 5.67 Gpixels/sec | Edge Detection | 12.3 Gpixels/sec |
| 6.44 Gpixels/sec | Gaussian Blur | 12.6 Gpixels/sec |
| 1.36 Gpixels/sec | Feature Matching | 1.83 Gpixels/sec |
| 528.5 Gpixels/sec | Stereo Matching | 942.6 Gpixels/sec |
| 16372.3 FPS | Particle Physics | 25637.1 FPS |
| OpenCL | API | OpenCL |
| 19527 | G3D Mark Score | 23401 |
| 767 | G2D Mark | 874 |
| 178 FPS | DirectX 11 | 232 FPS |
| 85 FPS | DirectX 12 | 109 FPS |
| 8389 Ops/s | GPU Compute | 8154 Ops/s |
| 21186 | GB6 ML Single Precision | 27844 |
| 34404 | GB6 ML Half Precision | 48544 |
| 16559 | GB6 ML Quantized | 21478 |
| 8642 | Image Classification (SP) | 11723 |
| 20635 | Image Segmentation (HP) | 32142 |
| 26378 | Image Super Resolution (Q) | 30735 |
| 39776 | Face Detection (HP) | 58722 |
| 105983 | Pose Estimation (Q) | 129551 |
| 2995 | Text Classification (SP) | 3539 |
| 4507 | Machine Translation (HP) | 6242 |
| 10198 | Object Detection (SP) | 14495 |
| 41078 | Depth Estimation (Q) | 50690 |
| 190315 | Style Transfer (SP) | 231154 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| 59-104 GPixel/s | Pixel Fill Rate | 163-180 GPixel/s |
| 177-313 GTexel/s | Texture Fill Rate | 408-450 GTexel/s |
| 11.3-20 TFLOPS | FLOPS (FP32) | 26.1-28.8 TFLOPS |
| GDDR6 | Memory Type | GDDR7 |
| 8 GB | Memory Size | 16 GB |
| 2000 MHz | Memory Clock | 1750 MHz |
| 16000 Mbps | Effective Memory Speed | 28000 Mbps |
| 128-bit | Bus | 256-bit |
| No | ECC | No |
| 256 GB/s | Memory Bandwidth | 896 GB/s |
| PCIe 4.0 x16 | Interface | PCIe 5.0 x16 |
| 35-140 W (configurable) | TGP | 80-175 W (configurable) |
| TSMC | Manufacturing | TSMC |
| 5 nm | Fabrication Process | 5 nm |
| 186 mm² | Die Size | 378 mm² |
| 22.9 billion | Transistor Count | 45.6 billion |
| 123.12 MTr/mm² | Transistor Density | 120.63 MTr/mm² |
| 12 | DirectX | 12.2 |
| 1.3 | Vulkan | 1.4 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 3 |
| 8.9 | CUDA | 12 |
| Yes | Ray Tracing | Yes |
| DLSS 3 | DLSS | DLSS 4 |
| 1.4a | DisplayPort | 2.1b |
| Nvidia | Vendor | Nvidia |
| Discrete | Build | Discrete |
| February 22, 2023 | Released | March 18, 2025 |
| Laptop | Case | Laptop |
| Gaming | Purpose | Professional |
| High-end | Segment | High-end |
| Ada Lovelace | Architecture | Blackwell 2.0 |
| AD106 (GN21-X6) | GPU Codename | GB203-300-A1 |
| -Intel Core Ultra 7 155Hor above | Recommended CPU | -Intel Core Ultra 9 275HXor above |
| 12729 | Steel Nomad Lite Score | 22035 |
| 12518 | Time Spy | 13961 |
| 54826 | Solar Bay | — |
| 7403 | Port Royal | 12094 |
| 30044 | Fire Strike | 38402 |
| 23244 | Wild Life Extreme | — |
| 113720 | Night Raid | — |
| 735-2070 MHz (configurable) | Base Clock | 975-1702 MHz (configurable) |
| 1230-2175 MHz (configurable) | Boost Clock | 1702-1875 MHz (configurable) |
| 4608 | Shading Units | 7680 |
| 144 | Texture Mapping Units (TMUs) | 240 |
| 48 | Render Output Units (ROPs) | 96 |
| 36 | Compute Units (Pipelines) | 60 |
| 144 | Tensor Cores | 240 |
| 36 | Ray-tracing Cores | 60 |
| 128KB per cluster | L1 Cache | 128KB per cluster |
| 32MB shared | L2 Cache | 64MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |