The M3 Max beats the M2 Ultra in single-task benchmarks, that measure how well various kinds of software can take advantage of the hardware. Such benchmarks don

The M3 Max beats the M2 Ultra in single-task benchmarks, that measure how well various kinds of software can take advantage of the hardware. Such benchmarks don't scale well with the number of CPU cores, because many tasks involve sequential bottlenecks. With higher-end hardware, you should get more meaningful results by running a few copies of the benchmark software in parallel and reporting the sum of multicore scores as the true multicore score. That would reflect the common use case of running several independent tasks in parallel.