M1 Performance: Geekbench vs Reality and heavy multitasking?

[ed.]

Another real world comparison. I used Ebsynth (a graphics processing engine that takes frames from a movie and re-renders each one based on one or more 'input frames' with a different visual... to make a short example: you can give it a short movie, then take the first frame and do a painting of it. The software will then be able to produce the whole movie in the 'painted style'. It's rather impressive).
Anyways the same project takes about 3:50 on the 2017 iMac (32GB quad core 4,2GHz i7), and 2:55 on the macbook pro (M1 16Gb). Ebsynth is not optimised for M1 and doesn't use the GPU on either machine.
As usual, the macbook was silent (temps went up to 70 celsius), and the iMac whooshed away like a mini hurricane.

 

[DWHH1]

There may be a bigger problem with the new Apple Silicon processors. How many of the libraries used by developers and researchers are available for Apple Silicon? For example, and copied from https://scikit-learn.org/stable/install.html#installing-on-apple-silicon-m1-hardware,

Installing on Apple Silicon M1 hardware​

The recently introduced macos/arm64 platform (sometimes also known as macos/aarch64) requires the open source community to upgrade the build configuation and automation to properly support it.​

​

At the time of writing (January 2021), the only way to get a working installation of scikit-learn on this hardware is to install scikit-learn and its dependencies from the conda-forge distribution, for instance using the miniforge installers:​

In a post elsewhere on Macrumors Forums I asked if anyone had used multiprocessing Python (ie had used concurrent.futures) with Apple Silicon. Apparently no one had.

Is Apple Silicon and Apple Dev software Xcode/SWIFT etc only able to deliver Apps that provide gateway access to the Apple ecosystem?

Does Apple Silicon mark the end of Macs being used to develop technical AI, NLP, datascience, engineering code?

 

[Ritsuka]

What would it? There isn't nothing preventing developers from fixing those libraries, if only you had tried to follow the links on that page, you would had seen that the configuration issues are already mostly solved.

 

[leman]

Is Apple Silicon and Apple Dev software Xcode/SWIFT etc only able to deliver Apps that provide gateway access to the Apple ecosystem?

Of course not. Swift is an open sourced programming language, you can use it to develop Windows apps if you want to. Xcode is based on open-source compiler toolchains and supports all the standard stuff. There are of course Apple-specific libraries, but that's something else.

Does Apple Silicon mark the end of Macs being used to develop technical AI, NLP, datascience, engineering code?

Why would it be? If some libraries do not support ARM or Apple's new frameworks, it is up to the community and the maintainers of those libraries to improve this.

P.S. I don't use python much, but some other libraries I use regularly had issues, so I submitted patches to some of them and other people did for others. In fact, the effort of supporting Apple Silicon was nothing but impressive. A huge number of bugs and incompatibilities were fixed in the course of a few weeks.

P.P.S. ARM today has released more information on their next-generation cores targeted at HPC systems. Nvidia has an ARM-based high-performance data center product in the pipeline. ARM is quickly gaining presence in HPC segment, consistently chipping away at x86 market share. While ARM won’t replace x86 in the personal computing space (Windows), any time soon, it is becoming a popular choice for servers, and this in turn means that you ideally need ARM hardware to develop server-side tools. And right now Apple is the only one offering high-performance ARM-based personal devices, which will make Apple Silicon Macs the platform of choice for HPC and similar applications.

 

[ed.]

Here's another unscientific comparison. I've rendered a project in After Effects on my iMac 5k (32GB 4,2Ghz 2017 i7), that took 8 minutes and 20 seconds. The frame size of this version is 2520x1080, rendered in ProRes.
I re rendered the same project on the M1 Macbook Pro (16GB), but upscaled it to 4096x1714 (working in AE felt snappier on the Macbook than the iMac btw - with AE still running under Rosetta).
The render on the M1 took 10 minutes 40 seconds. This is with more than 2,5 times the pixels per frame.
The Macbook was completely quiet as usual, while the iMac... ✈️🌬

 

[Jorbanead]

Here’s a great super in-depth video about Logic Pro X and how it runs on M1:

And this video showing 1,000 plugins on M1

 

[poorcody]

Stumbled across this video earlier today, doing some real-world comparisons of various apps between an M1 MacBook Air and a Windows MateBook Pro X :

Pretty impressive. They also did the three hours+ of testing all under battery, and compared the final battery levels of the two machines (see minute 18 for results).