Apple’s M3 Pro Chip Has Reduced Memory Bandwidth vs M1/M2
Apple’s latest M3 Pro chip, featured in the new 14-inch and 16-inch MacBook Pro models, has 25% less memory bandwidth than its M1 Pro and M2 Pro counterparts used in previous generations.
Despite being based on the latest 3-nanometer technology and boasting a new GPU architecture, the M3 Pro system on a chip (SoC) offers 150GB/s memory bandwidth, compared to 200GB/s in the M1 Pro and M2 Pro chips, according to an examination of the specs sheet by MacRumors.
According to Apple, the M3 Pro chip is up to 40% faster than the 16-inch model with the M1 Pro. However, the company has altered the core ratios in the M3 Pro, featuring 6 performance cores and 6 efficiency cores, as opposed to 8 performance cores and 4 efficiency cores in the 12-core M2 Pro. The GPU in the M3 Pro has 18 cores, one less than the 19 cores in the equivalent M2 Pro chip.
The M3 chip’s 16-core Neural Engine, while having the same number of cores as the 3nm-based A17 Pro chip in the iPhone 15 Pro series, shows a lower maximum throughput. The M3 Neural Engine is capable of 18 trillion operations per second (TOPS), compared to 35 TOPS for the A17 Pro Neural Engine.
Apple suggests that the lower throughput may be compensated by the M3’s additional GPU cores for tasks like machine learning.
The real-world impact of these changes on the M3 Pro’s performance remains unclear, especially given the introduction of new Dynamic Caching memory allocation technology. Apple’s comparisons primarily focus on the older M1 variants, making performance gains against the more recent M2 chips appear modest. Further insights are expected once comprehensive third-party benchmarks become available.
We’ll wait until real-life benchmarks come out, but from the surface, this looks like Apple slowly milking profits instead of offering pure performance from one generation to the next of Apple Silicon.
You can see a similar breakdown of Apple’s new M3 chips versus M1/M2 chips below by Max Tech: