Single Core and Multi-Core Performance Figures?

Single-core performance is generally very important because much software is not fully parallelized to use multiple threads. Increasingly, Web Apps are replacing traditional platform-specific compiled binaries, yet Web Apps are inherently single-threaded. It could take many years until languages and frameworks are improved to harness multi-threading on Web Apps. Until that happens, only improvements to single-core CPU performance will make those run faster.

FCP is significantly multi-threaded, but some workloads still run mainly on the main thread. Thus, only faster single-core performance will improve that.

You did not state the version of your Mac Mini, nor the version of GeekBench. You can't compare, say, GeekBench 5 to GeekBench 6. But even within a major version number, it's best to use the exact same version on both machines.

If your Mac Mini is an M4, it's fully expected the single-core performance is faster than any version of M3 (including Ultra). The M4 micro-architecture was significantly improved to produce better IPC (Instructions Per Cycle). Within a given generation of Apple M-series chip, they use the same micro-architecture. In general, the single-core performance of M1, M1 Pro, M1 Max and M1 Ultra are about the same. Likewise for M2-series and M3-series. The single-core micro-architecture was tweaked from M1 through M3, but M4 was the first big change.

Traditionally, Unix-derived operating systems (which include MacOS) have displayed process CPU utilization using the convention 100% = full utilization on a single core. Therefore if the process contains multiple threads, it can be potentially running on multiple cores, where the theoretical maximum is 100 * number of cores. If you have a 10-core machine, that would be 1000%.

If they only used 0-100%, a single-thread process and multi-thread process would both show (say) 50% CPU, but the multi-thread process would be consuming far more system CPU resources. So the > 100% display method is more informative.

Unfortunately, there is no built-in MacOS tool for monitoring per-thread CPU activity for a given process. However, there's a free command-line utility 'htop' which can be installed on macOS using the "Homebrew" package manager. Google "Homebrew" or "htop" for details. See the attached example.

[Edited by Moderator]

Ian R. Brown wrote:

Thanks for the explanation. How can I tell which apps are single-core?

Does this Activity Monitor window show the number of cores an app uses?

No, this shows core usage overall for your system.

You can, however, get a general idea if the system is not doing much except for an application that is doing some hard work, like FCP.

On the other hand, if you open the main Activity Monitor window and see something using

more than 100%, you will know it is using multiple cores.

Try, for example, a file transcoding using HandBrake, and you may see 1000%

Ian R. Brown wrote:

That's over 1,000% (Thousand)

Is that because 10 cores are at work?

More than 10, actually. This screenshot was made in my M4 Pro MBP. This machine has 14 cores, and HandBrake is one application that does use multiple cores.

However, this still does not make it faster than Compressor in some jobs; that is because Compressor can use the dedicated video encoder hardware, and so it can usually produce H264 or especially H265 (HEVC) faster.

Or you can also use MacPorts as the package manager to install htop. I prefer MacPorts for open source projects since it does not rely on MacOS system installed software that can change/break when Apple does Software Updates. Also the ports packages live in the /opt directory which Apple has indicated they will not step on.

Here is HandBrake converting a file to H265: