Apple M4 Max GPU (40-core) vs M3 Ultra GPU (80-core)
VS
Apple M4 Max GPU (40-core)
Custom
Apple M3 Ultra GPU (80-core)
Custom
We performed a head-to-head comparison of the Apple M4 Max GPU (40-core) with 640 pipelines and 5120 shaders against the 4 months newer M3 Ultra GPU (80-core) that utilizes 1280 pipelines and 10240 shaders. Here you will find complete details about specs, efficiency, performance tests, and more.
Review
Gaming
Performance in DirectX, OpenCL, and Vulkan games
Workstation
Perf. in 3D modeling, video editing and rendering apps
AI/ML
Capabilities for machine learning and AI-related tasks
Energy Efficiency
Power consumption efficiency in different scenarios
NanoReview Final Score
Overall video card score
The "Energy Efficiency" metric has less impact on the NanoReview Score for desktop GPU.
Key differences
Reasons to consider the Apple M3 Ultra GPU (80-core)
- Performs better (up to 34%) in 3DMark Steel Nomad Lite
- 75% higher maximum theoretical performance (28.3 vs 16.2 TFLOPS)
- Has 50% higher memory bandwidth: 819.3 vs 546 GB/s
- Achieves 27% more points in the GeekBench 6 Compute test (147K vs 116K)
- Has 2x more shading units (10240 vs 5120)
Benchmarks
Graphics cards’ performance in recent benchmarking apps3D Mark
Steel Nomad Lite Score
| Solar Bay | 61312 | 63597 |
| Wild Life Extreme | 36641 | 47814 |
GeekBench 6 OpenCL
GB6 Compute Score
| Background Blur | 195 img/sec | 204.5 img/sec |
| Face Detection | 125.9 img/sec | 126 img/sec |
| Horizon Detection | 4.82 Gpixels/sec | 6.53 Gpixels/sec |
| Edge Detection | 7.47 Gpixels/sec | 10.7 Gpixels/sec |
| Gaussian Blur | 5.8 Gpixels/sec | 9.5 Gpixels/sec |
| Feature Matching | 1.06 Gpixels/sec | 1.05 Gpixels/sec |
| Stereo Matching | 417.4 Gpixels/sec | 560.3 Gpixels/sec |
| Particle Physics | 17079.9 FPS | 25267.8 FPS |
| API | OpenCL | OpenCL |
Cinebench 2024 GPU
Cinebench 2024 GPU
Blender
Blender GPU
Recent User Tests
Apple M4 Max GPU (40-core)
| Date | Benchmark | Result |
|---|---|---|
| 📘 2025-07-10 (thomas) | GeekBench | 116478 |
Apple M3 Ultra GPU (80-core)
| Date | Benchmark | Result |
|---|---|---|
| 📘 2025-03-16 (Tyler) | GeekBench | 145706 |
| 📘 2025-03-16 (Tyler) | Cinebench 2024 | 19448 |
Artificial Intelligence Tests
Performance in machine learning and artificial intelligence tasksGeekBench 6 ML
GB6 ML Single Precision
GB6 ML Half Precision
GB6 ML Quantized
| Image Classification (SP) | 10302 | 8094 |
| Image Segmentation (HP) | 23763 | 24595 |
| Image Super Resolution (Q) | 29314 | 25190 |
| Face Detection (HP) | 42016 | 38337 |
| Pose Estimation (Q) | 105972 | 111021 |
| Text Classification (SP) | 2963 | 2786 |
| Machine Translation (HP) | 6594 | 4339 |
| Object Detection (SP) | 8935 | 7668 |
| Depth Estimation (Q) | 43143 | 37488 |
| Style Transfer (SP) | 221563 | 244483 |
| Framework | Core ML | Core ML |
| Backend | GPU | GPU |
Specifications
Technical specifications of Apple M4 Max GPU (40-core) and M3 Ultra GPU (80-core)General
| Vendor | Apple | Apple |
| Build | Integrated | Integrated |
| Released | October 30, 2024 | March 5, 2025 |
| Case | Laptop | Desktop |
| Purpose | Professional | Professional |
| Segment | Mid-range | High-end |
| Architecture | Apple M GPU | Apple M GPU |
| GPU Codename | Custom | Custom |
| Rival Equivalent | - GeForce RTX 4070 Laptop | - GeForce RTX 5070 Ti |
| Successor | - Apple M5 Max GPU (40-core) | - Apple M4 Ultra GPU (80-core) |
| Recommended CPU | - Apple M4 Max (16-Core) or above | - Apple M3 Ultra or above |
| Used in CPUs | - Apple M4 Max (16-Core) | - Apple M3 Ultra |
Laptop GPU ranking (#14th place)
Desktop GPU rating (#30th place)
Graphics Processing Unit
| Base Clock | 500 MHz | - |
| Boost Clock | 1578 MHz | 1380 MHz |
| Shading Units | 5120 | 10240 |
| Texture Mapping Units (TMUs) | 320 | 640 |
| Render Output Units (ROPs) | 160 | 320 |
| Compute Units (Pipelines) | 640 | 1280 |
| Instructions Per Cycle | 2 IPC | 2 IPC |
Raw Performance
| Pixel Fill Rate | 252 GPixel/s | 442 GPixel/s |
| Texture Fill Rate | 505 GTexel/s | 883 GTexel/s |
FLOPS (FP32)
Physical
| Interface | Custom | Custom |
| TGP | 50 W | 140 W |
| Manufacturing | TSMC | TSMC |
| Fabrication Process | 3 nm | 3 nm |
| Transistor Count | - | 184 billion |
| Max. Temperature | 100°C | 100°C |
Memory
| Memory Type | System Shared | System Shared |
| Memory Clock | 8533 MHz | 6400 MHz |
| Bus | 512-bit | 1024-bit |
| ECC | No | No |
Memory Bandwidth
API
| Ray Tracing | Yes | Yes |
| DLSS | No | No |
Cast your vote
Total votes: 31