Quick/Comparisons
14 | Gaming | 29 |
24 | Workstation | 35 |
| — | AI/ML |
Compare with something else?
Compare Apple M4 GPU with another →97 | Energy Efficiency | 56 |
32 | Quick Comparison Final Score | 32 |
| Custom | Graphics Processor | BMG-G21 |
| 1024 | Cores | 2304 |
| 64 / 32 | TMUs / ROPs | 144 / 72 |
| 1060.26 | Blender GPU | 1677.35 |
| — | Forza Horizon 5 | 66 |
| — | The Witcher 3 | 110 |
| — | Counter-Strike 2 | 108 |
| — | Far Cry 6 | 88 |
| — | Hogwarts Legacy | 50 |
| — | Call of Duty | 67 |
| — | Ghost of Tsushima | 53 |
| — | Cyberpunk 2077 | 68 |
| — | Tomb Raider | 94 |
| — | Average FPS | 78 |
| — | 1080p High | 96 |
| — | 1080p Ultra | 78 |
| — | 1440p Ultra | 58 |
| — | 4K Ultra | 31 |
| — | Margin of Error | Low |
| 30791 | GB6 Compute Score | 85640 |
| 57.4 img/sec | Background Blur | 223.7 img/sec |
| 41.3 img/sec | Face Detection | 115.2 img/sec |
| 1.21 Gpixels/sec | Horizon Detection | 2.14 Gpixels/sec |
| 1.74 Gpixels/sec | Edge Detection | 3.95 Gpixels/sec |
| 1.12 Gpixels/sec | Gaussian Blur | 4.81 Gpixels/sec |
| 0.43 Gpixels/sec | Feature Matching | 0.93 Gpixels/sec |
| 94.9 Gpixels/sec | Stereo Matching | 294.6 Gpixels/sec |
| 3902.5 FPS | Particle Physics | 11270.6 FPS |
| OpenCL | API | OpenCL |
| — | G3D Mark Score | 14062 |
| — | G2D Mark | 627 |
| — | DirectX 11 | 112 FPS |
| — | DirectX 12 | 70 FPS |
| — | GPU Compute | 7306 Ops/s |
| — | GB6 ML Single Precision | 20296 |
| — | GB6 ML Half Precision | 33136 |
| — | GB6 ML Quantized | 12966 |
| — | Image Classification (SP) | 9924 |
| — | Image Segmentation (HP) | 20640 |
| — | Image Super Resolution (Q) | 14215 |
| — | Face Detection (HP) | 39045 |
| — | Pose Estimation (Q) | 56787 |
| — | Text Classification (SP) | 3028 |
| — | Machine Translation (HP) | 3740 |
| — | Object Detection (SP) | 13018 |
| — | Depth Estimation (Q) | 31035 |
| — | Style Transfer (SP) | 176206 |
| — | Framework | ONNX |
| — | Backend | DirectML |
| 47 GPixel/s | Pixel Fill Rate | 180 GPixel/s |
| 94 GTexel/s | Texture Fill Rate | 360 GTexel/s |
| 3 TFLOPS | FLOPS (FP32) | 11.5 TFLOPS |
| System Shared | Memory Type | GDDR6 |
| — | Memory Size | 10 GB |
| 7500 MHz | Memory Clock | 2375 MHz |
| 15000 Mbps | Effective Memory Speed | 19000 Mbps |
| 128-bit | Bus | 160-bit |
| No | ECC | No |
| 120 GB/s | Memory Bandwidth | 380 GB/s |
| Custom | Interface | PCIe 4.0 x8 |
| 15 W | TGP | 150 W |
| TSMC | Manufacturing | TSMC |
| 3 nm | Fabrication Process | 5 nm |
| — | Die Size | 272 mm² |
| — | Transistor Count | 19.6 billion |
| — | Transistor Density | 72.06 MTr/mm² |
| — | DirectX | 12 |
| — | Vulkan | 1.3 |
| — | OpenGL | 4.6 |
| — | OpenCL | 3 |
| — | CUDA | No |
| Yes | Ray Tracing | Yes |
| No | DLSS | No |
| — | DisplayPort | 2.1 |
| Apple | Vendor | Intel |
| Integrated | Build | Discrete |
| October 29, 2024 | Released | December 3, 2024 |
| Laptop | Case | Desktop |
| Professional | Purpose | Gaming |
| Mid-range | Segment | Mid-range |
| Apple M GPU | Architecture | Generation 12.7 |
| Custom | GPU Codename | BMG-G21 |
| -Apple M4 (8-Core)or above | Recommended CPU | -Intel Core Ultra 5 245Kor above |
| 3450 | Steel Nomad Lite Score | 9733 |
| — | Time Spy | 12517 |
| 13908 | Solar Bay | 48642 |
| — | Port Royal | 6728 |
| — | Fire Strike | 31226 |
| 7928 | Wild Life Extreme | 24464 |
| — | Night Raid | 112674 |
| 500 MHz | Base Clock | 1700 MHz |
| 1470 MHz | Boost Clock | 2500 MHz |
| 1024 | Shading Units | 2304 |
| 64 | Texture Mapping Units (TMUs) | 144 |
| 32 | Render Output Units (ROPs) | 72 |
| 128 | Compute Units (Pipelines) | 18 |
| — | Tensor Cores | 144 |
| — | Ray-tracing Cores | 18 |
| — | L1 Cache | 256KB per cluster |
| — | L2 Cache | 10MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |