Apple M4 Ultra GPU (80-core) vs M3 Max GPU (40-core)

We compared two integrated laptop professional GPUs: the Apple M4 Ultra GPU (80-core) with 1280 pipelines and 10240 shaders against the 1 year and 6 months older M3 Max GPU (40-core) that utilizes 640 pipelines and 5120 shaders. Here you will find complete details about specs, efficiency, performance tests, and more.

Please note that the tests on the Apple M4 Ultra GPU (80-core) are done on an engineering sample provided by our insiders. The data will be more accurate after the final version of this GPU is available.

Key differences

Key distinctions and advantages of M3 Max GPU (40-core) over M4 Ultra GPU (80-core)
Reasons to consider the Apple M4 Ultra GPU (80-core)
  • Performs significantly better (up to 2.1x) in 3DMark Steel Nomad Lite
  • 2.3x higher maximum theoretical performance (32.3 vs 14.1 TFLOPS)
  • Achieves 2.3x more points in the GeekBench 6 Compute test (218K vs 94K)
  • Has 2x more shading units (10240 vs 5120)

Benchmarks

Graphics cards’ performance in recent benchmarking apps

3D Mark

Multiplatform graphics benchmark suite that directly correlates with performance in modern games
Steel Nomad Lite Score
Solar Bay - 49656
Wild Life Extreme - 30893
Sources: 3DMark [1]

GeekBench 6 OpenCL

GPU test for computational tasks (image processing, photography, computer vision, and ML)
GB6 Compute Score
Background Blur - 157.6 img/sec
Face Detection - 100.9 img/sec
Horizon Detection - 3.99 Gpixels/sec
Edge Detection - 6.04 Gpixels/sec
Gaussian Blur - 5 Gpixels/sec
Feature Matching - 0.84 Gpixels/sec
Stereo Matching - 363.8 Gpixels/sec
Particle Physics - 12226.9 FPS
API OpenCL OpenCL
Sources: Geekbench [3]

Cinebench 2024 GPU

Hardware benchmark using Maxon's Cinema 4D rendering engine
Cinebench 2024 GPU

Blender

Rendering performance test for 3D modeling
Sources: Blender [9] – 444 samples

Artificial Intelligence Tests

Performance in machine learning and artificial intelligence tasks

GeekBench 6 ML

Tests throughput of AI operations in single, half, and quantized precision
GB6 ML Single Precision
GB6 ML Half Precision
GB6 ML Quantized
Image Classification (SP) - 8685
Image Segmentation (HP) - 20739
Image Super Resolution (Q) - 22775
Face Detection (HP) - 37676
Pose Estimation (Q) - 91609
Text Classification (SP) - 2909
Machine Translation (HP) - 5274
Object Detection (SP) - 7480
Depth Estimation (Q) - 38212
Style Transfer (SP) - 190681
Framework - Core ML
Backend - GPU
Sources: Geekbench [9]

Specifications

Technical specifications of Apple M4 Ultra GPU (80-core) and M3 Max GPU (40-core)

General

Vendor Apple Apple
Build Integrated Integrated
Released May 1, 2025 October 31, 2023
Case Laptop Laptop
Purpose Professional Professional
Segment High-end High-end
Architecture Apple M GPU Apple M GPU
GPU Codename Custom -
Rival Equivalent - GeForce RTX 5080 - GeForce RTX 4070 Laptop
Successor - - Apple M5 Max GPU (40-core)
Recommended CPU - Apple M4 Ultra or above - Apple M3 Max or above
Used in CPUs - Apple M4 Ultra - Apple M3 Max
Laptop GPU ranking (#21st place)

Graphics Processing Unit

Base Clock 500 MHz 500 MHz
Boost Clock 1578 MHz 1380 MHz
Shading Units 10240 5120
Texture Mapping Units (TMUs) 640 320
Render Output Units (ROPs) 320 160
Compute Units (Pipelines) 1280 640
Instructions Per Cycle 2 IPC 2 IPC

Raw Performance

Pixel Fill Rate 505 GPixel/s 221 GPixel/s
Texture Fill Rate 1010 GTexel/s 442 GTexel/s
FLOPS (FP32)
32.3 TFLOPS
14.1 TFLOPS

Physical

Interface Custom Custom
TGP 120 W 60 W
Manufacturing TSMC TSMC
Fabrication Process 3 nm 3 nm
Transistor Count - 56 billion
Max. Temperature - 100°C

Memory

Memory Type System Shared System Shared
Memory Clock 8533 MHz 6400 MHz
Effective Memory Speed - 12800 Mbps
Bus 1024-bit 512-bit
ECC No No
Memory Bandwidth

API

Ray Tracing Yes Yes
DLSS No No

Cast your vote

Choose between two graphics cards
0 (0%)
0 (0%)
Total votes: < 1

User opinions

You can share your opinion or ask a question in the comments below
🌐 Register your profile and become part of NanoReview community!