Quick/Comparisons
14 | Gaming | 7 |
23 | Workstation | 16 |
10 | AI/ML |
Compare with something else?
Compare Apple M3 GPU with another →8 |
95 | Energy Efficiency | 54 |
30 | Quick Comparison Final Score | 11 |
| — | Graphics Processor | Polaris 21 |
| 1280 | Cores | 1024 |
| 80 / 40 | TMUs / ROPs | 64 / 16 |
| 155.27 | Blender GPU | 137.81 |
| — | Forza Horizon 5 | 15 |
| — | The Witcher 3 | 15 |
| — | Counter-Strike 2 | 22 |
| — | Far Cry 6 | 19 |
| — | Hogwarts Legacy | 10 |
| — | Call of Duty | 20 |
| — | Ghost of Tsushima | 11 |
| — | Cyberpunk 2077 | 10 |
| — | Tomb Raider | 18 |
| — | Average FPS | 16 |
| — | 1080p High | 22 |
| — | 1080p Ultra | 16 |
| — | 1440p Ultra | 12 |
| — | 4K Ultra | - |
| — | Margin of Error | Low |
| 30027 | GB6 Compute Score | 19139 |
| 60.3 img/sec | Background Blur | 34 img/sec |
| 40.8 img/sec | Face Detection | 17.6 img/sec |
| 1.21 Gpixels/sec | Horizon Detection | 0.69 Gpixels/sec |
| 1.64 Gpixels/sec | Edge Detection | 1.43 Gpixels/sec |
| 1.2 Gpixels/sec | Gaussian Blur | 1.15 Gpixels/sec |
| 0.37 Gpixels/sec | Feature Matching | 0.2 Gpixels/sec |
| 102.6 Gpixels/sec | Stereo Matching | 65.5 Gpixels/sec |
| 3293.3 FPS | Particle Physics | 2232.5 FPS |
| OpenCL | API | OpenCL |
| — | G3D Mark Score | 3682 |
| — | G2D Mark | 486 |
| — | DirectX 11 | 25 FPS |
| — | DirectX 12 | 21 FPS |
| — | GPU Compute | 1417 Ops/s |
| 4120 | GB6 ML Single Precision | 4551 |
| 6992 | GB6 ML Half Precision | 5280 |
| 5638 | GB6 ML Quantized | 3415 |
| 1579 | Image Classification (SP) | 1731 |
| 2944 | Image Segmentation (HP) | 2205 |
| 4425 | Image Super Resolution (Q) | 4705 |
| 6661 | Face Detection (HP) | 6423 |
| 15020 | Pose Estimation (Q) | 19379 |
| 2984 | Text Classification (SP) | 1172 |
| 5876 | Machine Translation (HP) | 1724 |
| 1826 | Object Detection (SP) | 1937 |
| 9339 | Depth Estimation (Q) | 8173 |
| 22291 | Style Transfer (SP) | 36502 |
| Core ML | Framework | ONNX |
| CPU | Backend | DirectML |
| 55 GPixel/s | Pixel Fill Rate | 20 GPixel/s |
| 110 GTexel/s | Texture Fill Rate | 82 GTexel/s |
| 3.5 TFLOPS | FLOPS (FP32) | 2.6 TFLOPS |
| System Shared | Memory Type | GDDR5 |
| — | Memory Size | 4 GB |
| 6400 MHz | Memory Clock | 1750 MHz |
| 12800 Mbps | Effective Memory Speed | 7000 Mbps |
| 128-bit | Bus | 128-bit |
| No | ECC | No |
| 102.4 GB/s | Memory Bandwidth | 112 GB/s |
| Custom | Interface | PCIe 3.0 x8 |
| 15 W | TGP | 75 W |
| TSMC | Manufacturing | GlobalFoundries |
| 3 nm | Fabrication Process | 14 nm |
| — | Die Size | 123 mm² |
| 13.8 billion | Transistor Count | 3 billion |
| — | Transistor Density | 24.39 MTr/mm² |
| — | DirectX | 12 |
| — | Vulkan | 1.3 |
| — | OpenGL | 4.6 |
| — | OpenCL | 2.1 |
| — | CUDA | — |
| Yes | Ray Tracing | No |
| No | DLSS | No |
| — | DisplayPort | 1.4a |
| Apple | Vendor | Amd |
| Integrated | Build | Discrete |
| October 31, 2023 | Released | April 18, 2017 |
| Laptop | Case | Desktop |
| Professional | Purpose | Gaming |
| Mid-range | Segment | Mid-range |
| Apple M GPU | Architecture | GCN 4.0 |
| — | GPU Codename | Polaris 21 |
| -Apple M3or above | Recommended CPU | -Intel Core i5 8400or above |
| 3478 | Steel Nomad Lite Score | 1768 |
| — | Time Spy | 1854 |
| 13529 | Solar Bay | — |
| — | Port Royal | — |
| — | Fire Strike | 6492 |
| 8229 | Wild Life Extreme | 3605 |
| — | Night Raid | 25413 |
| 500 MHz | Base Clock | 1175 MHz |
| 1380 MHz | Boost Clock | 1275 MHz |
| 1280 | Shading Units | 1024 |
| 80 | Texture Mapping Units (TMUs) | 64 |
| 40 | Render Output Units (ROPs) | 16 |
| 160 | Compute Units (Pipelines) | — |
| — | Tensor Cores | No |
| — | Ray-tracing Cores | No |
| — | L1 Cache | 16KB per cluster |
| — | L2 Cache | 1024KB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |