MacBook M1 chip do's and don't

Dear all,

I am building an Audio-Visual experiment using sounds and video’s separately and together. This will be used to perform EEG-tests and behavioural tests. During the tests the participants will also have an active button-press-task in which reaction times will be recorded.

As I have read and discovered myself, the M1 chip in Apple causes some problems using PSB. Images and sounds can be displayed and played, but that is how far it seems to go. As such, I wanted to ask for your opinions.

To play the video’s and sounds I can also use the Image Processing Toolbox, including its VideoPlayer function. Also the tic/toc function built in in Matlab could be of use for measuring reaction times.

In this regard, I wanted to consult you and ask, are these options good enough to perform experiments and collect data for a study or are the PsychToolbox for these measures way way better? The option for now is using the videoplayer, imshow, etc. and tic toc function without PsychToolbox OR buying a non-mac laptop for testing purposes only on which I could then use PsychToolbox.

Please let me know your advice.
In addition, is there a list of which functions I can and cannot use with this M1 chip on Apple MacBooks.

Thank you in advance,
Have a nice day,

If you care at all about timing, using that alternative route is an absolute no go. it “works” because it doesn’t check nor try to be remotely accurate. Buy a separate machine that ideally runs linux well, and just use the mac for development

I totally love my M2 Macbook Air, with its utterly insane battery life and phenomenal performance given its energy use (I’ve been running Stable Diffusion, and it can generate images so quickly compared to my beefy workstation). I still prefer macOS for day-to-day use by far (it is both a *nix with a package manager, yet offers integration and consistency in a way Windows nor Linux desktop can). But I would never think of using it for data collection (I use Linux workstations with AMD GPUs for that). Science needs to be reliable and dependable, and PTB does this better than any other available system. Jerry rigging something together out of bits and bobs like image processing toolbox and tic and toc is ripe for failure (unless you really do not care at all about timing)…

As long as I start MATLAB from the terminal (to work around Apple’s security annoyances) and don’t utilise high-precision flip timing, almost everything else is working perfectly on my M2; it makes a great development machine (I use most advanced PTB functions apart from precise sound). PTB in this mode will still be much better than jerry-rigged generic code IMO…

Yep, I fully agree with Dee’s and Ian’s advice (apart from Ian’s opinion about macOS qualities and concistency and pleasantness wrt. to non-data-collection scenarios, but that is purely a matter of different personal taste).

You won’t find anything better than PTB for such tasks, but also currently macOS in general is not great for data collection, and macOS on any Apple Silicon Mac is an absolute no-go if precision and trustworthiness is required, and especially in the visual domain, although there seem to be various other shortcomings as well in other areas.

This is due to severe design limitations of Apples macOS OpenGL implementation for M1/M2/M… And bugs and limitations in Apples Metal and/or CoreAnimation graphics and display api’s.

So far I have spent multiple hundred work hours, only a fraction of them sponsored by Mathworks, trying to find ways to make it work, and have discovered that at least all other toolkits I know of do this incompetently. They are not only broken wrt. timing on Apple M1, but also (as opposed to PTB) on almost all old Apple IntelMacs. I would literally not trust any results collected with other software if visual timing is of importance.

The research wasn’t fruitless though. I have some ideas or hope on how one might improve the situation on macOS + M1/M2/… based on my previous research work. However this will not only require me to rent a Apple Silicon Mac, but also spend more work time on this, and this work has to be paid. So far less than 100 hours of ~275 hours have been paid, and due to the disappointing lack of financial support by the vast majority of our users, we can no longer afford to give much work away for free.

Another venue I want to explore, as the better solution or as backup plan, is if we could make Asahi-Linux work well enough to be suitable for data collection with PTB on Apple Silicon Macs, so users could dual-boot Linux/macOS for work vs. play. Asahi-Linux is an effort by partially volunteers, partially funded people, to get Linux running well on Apple Mx Macs. They have made great progress in the last 2 years, but one of the current main limitations is wrt. precise and trustworthy visual stimulation timing and control. Their focus is on getting the basic use cases working well right now, not special snowflakes like vision science. Also the Apple M1/M2/… display engines seem to have some serious limitations which make proper timing difficult. This is one area where i could chime in to see if I could come up with some workable solution.

My gut feeling is that it would be more likely to get a well working Asahi-Linux implementation than a macOS implementation.

All this requires substantial funding. I will probably try to get my hands on such a Mac for some preliminary tests about feasibility of such work, once current projects are completed. But if I’d find the time to do so, and if I’d think it might be feasible and worth giving a try, our dear users will have to fund it, e.g., with another crowd-sourcing attempt. I can guarantee though that certainly nothing will happen without users paying the bill.


Thank you all for the great and honest advice. As it seems that PsychToolbox is a highly valued tool within the neuroscience research, I’ll thank you in name of everyone for all your hard work and effort, and agree that funding and help from the community is essential for progress.

Have a nice day,

If you agree, then certainly your lab will have bought Psychtoolbox memberships over the last 3 years to support us financially to put actions behind words? Just checking…