Apple M4 10-core GPU vs Apple M3 Pro 18-Core GPU vs Apple M3 Pro 14-Core GPU

Apple M4 10-core GPU

The 10-core Apple M4 GPU is a graphics adapter built into Apple M4 SoCs that features 10 cores and uses the unified memory architecture (LPDDR5X-7500, 120 GB/s bandwidth). The Apple-designed iGPU unfortunately does not have any noteworthy new features that could help it stand out from its M3 series predecessors.

Much like the 10-core GPU built into the M3 SoC, the M4 GPU (10 cores) has hardware support for ray tracing as well as mesh shading and other modern technologies. It supports external displays with resolutions as high as 6K and it can hardware-decode the popular AV1, HEVC and AVC video codecs (encoding is not yet supported for AV1).

While early benchmark comparisons between the M4 GPU and the M3 GPU hint at performance gains as high as 15%, it is important to state that the M4 chip's performance sustainability is set to be very poor because of the new iPads' thin chassis that lacks active cooling.

• M4 GPU Geekbench Metal - early results: ~53,800 points
• M3 GPU Geekbench Metal - Notebookcheck database: ~47,000 points

Just like the rest of the Apple M4 chip, the graphics adapter is manufactured on a "second generation 3 nm" TSMC process for high, as of H1 2024, energy efficiency.

Apple M3 Pro 18-Core GPU

The Apple M3 Pro 18-Core GPU is a self-designed graphics card in the Apple M3 Pro (with 12 CPU cores) and offers all eighteen cores available on the chip. The graphics card uses a new architecture and now offers dynamic caching, mesh shading and ray tracing acceleration.

The GPU can access up to 36 GB LPDDR5-6400 unified memory via a 192-bit memory bus (max 150 GBit/s). Thanks to dynamic caching, the GPU only reserves as much memory as is required.

The entire chip is manufactured in the current 3nm process (probably N3B) at TSMC.

Apple M3 Pro 14-Core GPU

The Apple M3 Pro 14-Core GPU is a self-designed graphics card in the Apple M3 Pro (with 11 CPU cores) and offers fourteen of the eighteen cores available on the chip. The graphics card uses a new architecture and now offers dynamic caching, mesh shading and ray tracing acceleration.

The GPU can access up to 24 GB LPDDR5-6400 unified memory via a 192-bit memory bus (max 150 GBit/s). Thanks to dynamic caching, the GPU only reserves as much memory as is required.

The entire chip is manufactured in the current 3nm process (probably N3B) at TSMC.