Obtaining estimates for individual P and E core performance of processes run mainly in an ALU and those using floating-point and SIMD gives further insight and confirms the cores haven’t changed from M1 to M1 Pro.
How does macOS load processes onto the cores in M1 series processors? Are its policies similar between the original M1 and the M1 Pro?
Geekbench 5 scores for the M1 Pro are around 2800 single- and 12500 multi-core. Do they represent maximum performance, though?
Explore running test code of different types on the different cores in your M1 Mac. Provides insight into how your Mac performs, rather than just how fast it is.
Benchmarking 32-bit Float vector dot-product calculations using Swift, NEON assembly, and Apple’s SIMD libraries, on Intel and M1 Macs.
How to measure read speed on storage using 1600 files totalling over 320 GB, enabling comparison between internal and external SSDs.
Progress benchmarking performance of internal SSDs in M1 Macs: tests of 400 files totalling over 130 GB written and read in less than 50 seconds. And results.
Fast compact external SSDs are prone to thermal throttling. This is demonstrated during tests writing 96 files, when transfer rates fell from 2 GB/s to around 700 MB/s.
It started with wanting to benchmark SSDs, passed through robust linear regression, and is now diving deep into linear programming and matrices.
Improved tests, statistical analysis, and interface, together with a 21 page Help book detailing its range of tests in 5,000 words. Ready to use for real.