New 24-inch M1 Apple iMac destroys every iMac that ever came before it in Geekbench single-core tests with +37% gains over the nearest Intel Core i7 and i9 models

[Redaktion]

The new and colorful 24-inch Apple iMacs with M1 chips have started popping up on Geekbench and have revealed strong performances, especially in single-core testing. The M1 iMac was able to surpass any average single-core score by all previous iMac and iMac Pro models regardless of size or Intel processor variant.

https://www.notebookcheck.net/New-24-inch-M1-Apple-iMac-destroys-every-iMac-that-ever-came-before-it-in-Geekbench-single-core-tests-with-37-gains-over-the-nearest-Intel-Core-i7-and-i9-models.539353.0.html

[Gustavo]

The fair result should be M1 running Windows compared with Intel running Windows too, then the geekbench has credibility, Mac OS and Windows cannot be compared eye to eye because it is different kind of operating system environment altogether, I don't understand why the one called expert always compares apple with orange :D ..it is so naive then

[Wolfen]

Intel Tiger Lake can get 1700+ on single core with Linux.

Also consider that MacOS is Linux based.

[Andreh]

Intel Tiger Lake can get 1700+ on single core with Linux.

Also consider that MacOS is Linux based.

Is UNIX/BSD based no Linux, complety different, Linux cpu performance is much better than this UNIX sometimes 50% better, macOs is simply bad in performance, even windows is much faster, this geekbench is a mac oriented bench put some real things in the table like I did last mounth and m1 is always week than 3/4 years old intel cpu and new ryzens cpu, in multi is a complety disaster but ok continue with this non sense, such big cpu with bigger cost than amd/intel ones a much bigger than others arm...

[nothing_to_see]

Has anybody actually ran Geekbench? Geekbench generates pretty much useless numbers.

Try running Geekbench with some sort of system monitor on. You'll see NONE OF YOUR CORE is utilized nowhere close to 100%, during the 'single core' part of the Geekbench. Same goes for the 'multi core' part. ALL OF YOUR CORES won't be utilized nowhere near 100%. Not even close. Average utilization during the multi core test is much closer to idle than 100% utilization. And this happens to almost every CPU on every platform.

Most workloads of Geekbench don't care about maximum performance you could squeeze from the CPU. It just runs a set of libraries on 'as is' basis, hence it's not actually gauging the absolute maximum performance, nowhere close to that. What it measures is, how well each CPU run those libraries on 'as is' basis, which doesn't mean much. Random optimization by the library could play impact the results to a degree that it renders outcome useless.

Geekbench tries to mask this inherent flaws by introducing many sets of workloads, including some real-world workloads (because faster CPU would do better on average, even if it's not utilize the CPU to its fullest extent and there are some optimization hiccups here and there), but still it doesn't do any good job. The result is basically a combination of anecdotal measurements of some random libraries.

So don't treat the number like it means something. If you still incline to do that, at least get your fact straight. M1 scored higher than older Intel Macs. Tiger Lake hovers around 1600-1700 on Windows, Linux and even on macOS VM. Rocket Lake around 1900-2000, and Vermeer could get 2000-2000 on macOS VM.

[Boobaleh(sweeter)]

Kiddin'? GB is the only reliable CPU test, right after CineBench 10.5.
Do take it with a pinch of salt; as a reminder, BSD/Linux scores on Geekbench are usually 10% higher, compared to running the same chip under Windows.
'Sides, ARM has always been looking unnaturally strong against x86 on Geekbench.

[Bill]

The title could also be "4 year old Mac blows brand new M1 iMac our of the water in multicore scores". Why make this article so biased and only gloss over scores in which the M1 is bested by an Intel chip?

[Abhishek Kylasa]

Kiddin'? GB is the only reliable CPU test, right after CineBench 10.5.
Do take it with a pinch of salt; as a reminder, BSD/Linux scores on Geekbench are usually 10% higher, compared to running the same chip under Windows.
'Sides, ARM has always been looking unnaturally strong against x86 on Geekbench.

The title could also be "4 year old Mac blows brand new M1 iMac our of the water in multicore scores". Why make this article so biased and only gloss over scores in which the M1 is bested by an Intel chip?

This is because Geekbench only runs on a single-thread for CPU's that have built-in hyperthreading. That's why even the ultrabook versions of 11th-Gen Intel (let alone full power laptop or desktop versions) match the M1 in single-core performance. Comparing Geekbench single-core scores between these systems is literally comparing apples and oranges.

Plus, as others have noted Geekbench scores tend to be slightly higher on MacOS versus Windows. And in the case of non-native apps, even the very efficient Rosetta 2 drops performance far below Intel/AMD systems.

Now this shouldn't take away from the M1 being impressive. It is absolutely amazing for an ultrabook processor because it has not only high single-core performance on par with Intel's 11th Gen ultrabook CPU's, but much higher multi-core performance. Not only that, but it has incredible thermals (far better sustained performance and very quiet) and lower power usage (insanely high battery life). But let's not put it on a pedestal and say it crushes everything else. The most exciting part is that this is a chip primarily meant for ultrabooks and compact systems. A future theoretical M2 and M2X (with more high-power cores) might truly blow other systems out of the water.