Quick/Comparisons
12 | Gaming | 5 |
21 | Workstation | 9 |
11 | AI/ML |
Compare with something else?
Compare GeForce GTX 970 with another →3 |
10 | Energy Efficiency | 68 |
15 | Quick Comparison Final Score | 16 |
| GM204 | Graphics Processor | GP108 |
| 1664 | Cores | 384 |
| 104 / 56 | TMUs / ROPs | 24 / 16 |
| 286.89 | Blender GPU | 87.25 |
| 32 | Forza Horizon 5 | - |
| 35 | The Witcher 3 | - |
| 57 | Counter-Strike 2 | 17 |
| 39 | Far Cry 6 | 16 |
| 18 | Hogwarts Legacy | - |
| 31 | Call of Duty | 23 |
| 17 | Ghost of Tsushima | - |
| 24 | Cyberpunk 2077 | — |
| 44 | Tomb Raider | - |
| 33 | Average FPS | 10 |
| 44 | 1080p High | 14 |
| 33 | 1080p Ultra | 10 |
| 23 | 1440p Ultra | - |
| 12 | 4K Ultra | - |
| Low | Margin of Error | Medium |
| 28989 | GB6 Compute Score | 10870 |
| 77.1 img/sec | Background Blur | 27.8 img/sec |
| 38.1 img/sec | Face Detection | 13 img/sec |
| 1.28 Gpixels/sec | Horizon Detection | 0.4 Gpixels/sec |
| 2.08 Gpixels/sec | Edge Detection | 0.59 Gpixels/sec |
| 1.85 Gpixels/sec | Gaussian Blur | 0.64 Gpixels/sec |
| 0.17 Gpixels/sec | Feature Matching | 0.12 Gpixels/sec |
| 92.9 Gpixels/sec | Stereo Matching | 34 Gpixels/sec |
| 2443 FPS | Particle Physics | 960.9 FPS |
| OpenCL | API | OpenCL |
| 9639 | G3D Mark Score | 2376 |
| 763 | G2D Mark | 238 |
| 71 FPS | DirectX 11 | 18 FPS |
| 41 FPS | DirectX 12 | 12 FPS |
| 4080 Ops/s | GPU Compute | 1011 Ops/s |
| 6791 | GB6 ML Single Precision | 1840 |
| 6738 | GB6 ML Half Precision | 1885 |
| 5128 | GB6 ML Quantized | 1401 |
| 1852 | Image Classification (SP) | 405 |
| 3622 | Image Segmentation (HP) | 968 |
| 8987 | Image Super Resolution (Q) | 2273 |
| 7074 | Face Detection (HP) | 1652 |
| 43726 | Pose Estimation (Q) | 8039 |
| 1446 | Text Classification (SP) | 837 |
| 1576 | Machine Translation (HP) | 944 |
| 2425 | Object Detection (SP) | 552 |
| 12902 | Depth Estimation (Q) | 3111 |
| 86714 | Style Transfer (SP) | 17477 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| 66 GPixel/s | Pixel Fill Rate | 25 GPixel/s |
| 123 GTexel/s | Texture Fill Rate | 38 GTexel/s |
| 3.9 TFLOPS | FLOPS (FP32) | 1.2 TFLOPS |
| GDDR5 | Memory Type | GDDR5 |
| 4 GB | Memory Size | 2 GB |
| 1753 MHz | Memory Clock | 1502 MHz |
| 7000 Mbps | Effective Memory Speed | 6000 Mbps |
| 256-bit | Bus | 64-bit |
| No | ECC | — |
| 224.4 GB/s | Memory Bandwidth | 48 GB/s |
| PCIe 3.0 x16 | Interface | PCIe 3.0 x4 |
| 148 W | TGP | 25 W |
| TSMC | Manufacturing | Samsung |
| 28 nm | Fabrication Process | 14 nm |
| 398 mm² | Die Size | 74 mm² |
| 5 billion | Transistor Count | 1.8 billion |
| 12.56 MTr/mm² | Transistor Density | 24.32 MTr/mm² |
| 12 | DirectX | 12 |
| 1.3 | Vulkan | 1.2 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 1.2 |
| 5.2 | CUDA | 6.1 |
| No | Ray Tracing | No |
| No | DLSS | No |
| 1.2 | DisplayPort | 1.4 |
| Nvidia | Vendor | Nvidia |
| Discrete | Build | Discrete |
| September 19, 2014 | Released | February 21, 2009 |
| Desktop | Case | Laptop |
| Gaming | Purpose | Gaming |
| Mid-range | Segment | Entry-level |
| Maxwell 2.0 | Architecture | Pascal |
| GM204 | GPU Codename | GP108 |
| — | Recommended CPU | -Intel Core i7 1065G7or above |
| 3577 | Steel Nomad Lite Score | 896 |
| 3666 | Time Spy | 1078 |
| — | Solar Bay | — |
| 4353 | Port Royal | — |
| 12102 | Fire Strike | 3619 |
| 6972 | Wild Life Extreme | 2227 |
| 47591 | Night Raid | 14485 |
| 1050 MHz | Base Clock | 1519 MHz |
| 1178 MHz | Boost Clock | 1582 MHz |
| 1664 | Shading Units | 384 |
| 104 | Texture Mapping Units (TMUs) | 24 |
| 56 | Render Output Units (ROPs) | 16 |
| 13 | Compute Units (Pipelines) | 3 |
| No | Tensor Cores | No |
| No | Ray-tracing Cores | No |
| 48KB per cluster | L1 Cache | 48KB per cluster |
| 2MB shared | L2 Cache | 512KB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |