How has the CPU core of the Apple M3 series changed from M1?



Apple's Arm-based SoC '

Apple Silicon ' has been released up to the third generation M3 series at the time of writing, and Apple emphasizes that the performance of Apple silicon improves with each generation. I am. The Eclectic Light Company, a blog that deals with Mac-related information, is examining how specifically the performance of M3 has changed compared to M1, focusing on the CPU core.

What has changed in CPU cores in M3 chips? – The Eclectic Light Company
https://eclecticlight.co/2023/11/22/what-has-changed-in-cpu-cores-in-m3-chips/

Apple silicon CPUs are equipped with two types: P cores (high performance cores) and E cores (high efficiency cores).



The biggest difference between the M1/M2 CPU core and the M3 CPU core is the size of the core cluster.In the M1/M2 series, the CPU cores are grouped in groups of two or four, and the L2 cache is shared among them. and operate at the same frequency. However, the M3 has four CPU cores each, and the M3 Pro and M3 Max have six CPU cores each.



macOS tries to allocate threads that run at higher priorities to P cores whenever possible. Although processing by P cores consumes more power, it is designed to ensure that running applications can achieve the best performance. However, when the processing on the P core is full, it may be executed on the E core. In this case, threads may run more slowly.

The maximum frequency of the E core is 2064MHz for M1, but reaches 2748MHz for M3 chip. However, when processing low priority threads, the E-core frequency on the M1 Pro is set to 972MHz, while on the M3 Pro it is set to 744MHz. In other words, the background thread execution speed is slightly slower on M3 than on M1.

The graph below measures the time taken to run a thread with a low processing priority using The Eclectic Light Company's benchmark tool 'AsmAttic' on the M1 (red) and M3 (black) E cores. The vertical axis is the time required for processing, and the horizontal axis is the number of threads. From this graph, you can see that M3 takes longer to process.



There is also a big difference between M1 and M3 regarding the P core. M1's P core has a maximum frequency of 3228MHz, while M3's P core has a maximum frequency of 4056MHz. The M3's P core is approximately 30% higher in integer arithmetic performance, floating point arithmetic performance, and vector arithmetic performance, and this result is consistent with Apple's official values. The Eclectic Light Company speculates that Apple may have improved the P-core design with the M3.



The graph below is the result of measuring the time required for 200 million floating point operations using AsmAttic on the P core on M1 (red) and M3 (black). The vertical axis shows the cumulative time and the horizontal axis shows the number of threads. I am. You can see that M3 has faster processing speed than M1 when it comes to P cores.



Changing cluster size has a big impact on your Mac's performance, and it also impacts power consumption. At full capacity, the M3 Pro's P-core cluster consumes approximately 5.5W, while the M1 Pro's P-core cluster consumes approximately 5.8W.

The Eclectic Light Company says, ``If you already have an Apple silicon-powered Mac and are wondering whether to upgrade to an M3-powered model, you can use your current Mac to run the apps you normally use for work. , look at its usage in the Activity Monitor's CPU history window.If you have enough free space on P cores, you probably won't see much benefit in upgrading to an M3-equipped model with more P cores. states.

in Hardware, Posted by log1i_yk