M4 chip: the most advanced processor for AI

The M4 chip was introduced in a special way this year, not in a Mac, but in the iPad Pro. It will probably not be limited to just the iPad. The M4 processor may also be added to the Mac mini and other Apple devices later this year.

The M3 chip was unveiled in the autumn of 2023 and was replaced by the M4 within six months. A special choice by Apple. Just like the A17 Pro for iPhone 15 Pro (Max) and M3 processor, the M4 chip is designed using a 3nm manufacturing process and this is reflected in the performance of the chip.

Extra powerful 3-nanometer chip

The M4 series is designed using a second-generation 3nm manufacturing process. This is an advanced manufacturing technology used to make semiconductor chips. A 3nm process means that the transistor gate is only 3 nanometers wide.

The size of the transistor gate is important because it determines how many transistors can be placed on a single chip. The smaller the transistor gate, the more transistors fit on the chip.

M4 chip construction

This leads to chips that are faster, more efficient and more powerful than chips produced using larger processes. This means they perform better with the same specifications. The previous generations of Apple chips used a 5nm process.

M4 processor in short

The new CPU has up to 10 cores, and the new 10-core GPU builds on M3’s advanced GPU architecture. For example, on the iPad Pro you can benefit from dynamic caching, mesh shading and ray tracing with hardware acceleration for the first time.

M4’s new 10-core GPU builds on the advanced graphics architecture of the M3 chips. The GPU features dynamic caching, an Apple innovation that allows local memory to be allocated dynamically and in real time on the hardware.

M4 has a Neural Engine that can handle a whopping 38 trillion operations per second. That’s faster than the neural processing unit in any available AI PC. Together with faster memory bandwidth, advanced ML accelerators in the CPU and a super-powerful GPU, every device with M4 chip becomes an unprecedentedly powerful device for AI.

M4 chip specifications

Below we list the specifications of the M4 chip and what its possibilities are. The M4 chip can be found in the 11-inch iPad Pro and 13-inch iPad Pro from 2024. Later this year, the new processor is also expected in the Mac, including the Pro, Max and Ultra extensions.

M4 chip

The specifications below are for the iPad Pro, the Mac is expected to have more options, including more memory.

  • 10-core CPU with 4 performance cores and 6 efficiency cores
  • up to 10-core GPU
  • Up to 16 GB memory (standard 8 GB)
  • 16-core Neural Engine
  • Ray tracing with hardware acceleration
  • 120 GB/s memory bandwidth
  • Media engine: 8K H.264, HEVC, ProRes and ProRes RAW with hardware acceleration
  • Video and ProRes encoding and decoding engine + AV1 decoding
  • choice of up to 2 TB of storage

New 10-core CPU

M4 has a new CPU with up to 10 cores, including four performance cores and now a whopping six efficiency cores. The advanced cores have improved branch prediction, larger decoding and execution engines for the performance cores and a deeper execution engine for the efficiency cores. In addition, both types of cores feature improved machine learning accelerators.

M4 chip

Brand new GPU

M4’s new 10-core GPU builds on the advanced graphics architecture of the M3 chips. The GPU features dynamic caching, an Apple innovation that allows local memory to be allocated dynamically on the hardware in real time. As a result, the GPU can be used much better on average and the most demanding professional apps and games run a lot smoother.

With these improvements to the CPU and GPU, Apple silicon continues to lead the industry in performance per watt. M4 delivers the same performance as M2 and uses half as much power.

iPad Pro with M4 chip

The most powerful Neural Engine

M4 has a lightning-fast Neural Engine, the IP block in the chip that ensures the acceleration of AI workloads. This is the most powerful Neural Engine Apple has developed to date, with up to 38 trillion operations per second. This makes it no less than 60 times as fast as the first Neural Engine in the A11 Bionic chip. Together with the advanced ML accelerators in the CPU, the super-powerful GPU and a central memory with more bandwidth, the Neural Engine makes M4 ideally suited for AI and the processor is ready for the future.

Recent Articles

Related Stories