I was yet again spectacularly wrong in speculating that we had another eight months to wait before Apple would release the first Macs with M3 chips. Another few days and the first will be upon us, and the fortunate few will start bragging or moaning about their performance. That has suddenly grown more complex: the number of each CPU core type has diversified with the M3 Pro in particular. This article looks at some of the factors involved in comparing CPU performance across Apple’s expanded range of M-series chips.
As Apple’s line-up of processors grows, it’s becoming harder to keep track of all the details. This article does a good job of highlighting some of the changes coming inside M3-based Macs.
I agree, these modern asymmetric CPUs with E cores and P cores make things more complicated.
The benchmarks that max everything out are still legitimate, but not necessarily representative for specific workloads. I like when reviewers benchmark these CPUs using many real as well as synthetic workloads. This gives us a more rounded appreciation for what they are capable of and what to expect for bottlenecks.
Also, this is an older article about the M1, but they an insightful point about how comparing “threads” is not the same thing as comparing “cores” due to hyperthreading.
https://www.extremetech.com/computing/318020-flaw-current-measurements-x86-versus-apple-m1-performance
Cores that are optimized to run two threads are disadvantaged by only running a single threaded workload because 30% of a course’s resources might still go unused.
1T vs 1T makes hyperthreading cores look bad since unused performance is being left on the table.
1C vs 1C makes hyperthreading cores look good since a loaded hyperthreading core can run 2 threads instead of 1 on a non-hyperthreading core.
It’s just something to keep in mind when different CPU typologies using portable benchmarks.
Now that we have CPUs with more kind of variations, apples to apples comparisons are becoming more complicated!