Apple goes 3-nanometer with the new generation of M3 chips, and it’s the first in the industry.
It’s almost an accepted assumption that smaller processes translate to better performance and efficiency, but what are the numbers here?
The biggest advancement is in the graphics architecture, with new Dynamic Caching tech that efficiently uses only the exact amount of memory needed for the task in real-time. For the first time, Mac users can enjoy ray tracing (more accurate shadows and reflections) and mesh shading (enables more visually complex rendered scenes) for their machines.
Unified memory structure for the M3 Max.
By way of benchmarking, Apple says that the rendering speed of the M3 GPU is apparently 2.5 times faster than M1. The M3 GPU apparently delivers the same performance as the M1 with about half the power and, at peak power consumption, achieves up to 65% more performance.
For the CPU, performance cores are 30% faster, while efficiency cores are 50% faster than M1 (15% and 30%, respectively, vs M2). In addition, the Neural Engine is 60% faster. M3 also features a new media engine that supports AV1 decode support for efficient, high-quality video via streaming. Power-wise, compared to the M1, the M3 does the same job for half the power and offers 35% more performance at peak power.
In addition, the unified memory architecture now supports up to 128GB of memory, which is great for laptops, as this is not usually something it can support, while the Neural Engine has been updated as well and now performs up to 60% faster than M1 and 15% faster than M2.
In summary, here’s how the M3, M3 Pro, M3 Max stack up:
- M3: 25 billion transistors, 10-core GPU, 8-core CPU, up to 24GB unified memory.
- M3 Pro: 37 billion transistors, 18-core GPU, 12-core CPU, up to 36GB unified memory.
- M3 Max: 92 billion transistors, 40-core GPU, 16-core CPU, up to 128GB unified memory.
