Running parallel using build in 10-core GPU and 16-core neural Engine in MacBook Air M2

Now I am considering to buy the new macbook air M2 with respect to the following details; 8 core CPU, 10 Cores GPU and 16 cores neural enginge. I would like to use it to perform simbiology in MATLAB as well as the deep learning. Is it possible for simbiology and deeplearning to utilize all capacity of the program, including the calculation using GPU and neural enginge cores?. Do they have a new version for apple silicon M2 chip update?

Thank you,

SIncerely yours,

Teerachat Saeheng

Answer 1

[Arthur Goldsipe]

You can use run the Intel version of many releases of MATLAB on an M2 Mac just fine. I'm a SimBiology developer with an M1Pro MacBook Pro, and that's what I do.

At the time I write this, there is an open beta of a native Apple Silicon version of MATLAB R2022a. I refer you to the answer on this question for the latest information and how to download the open beta.

As described in this question, MATLAB currently only supports NVIDIA GPUs for computation.

[Arthur Goldsipe]

Oh, I also wanted to link to this blog post, in case you find that helpful.

[Teerachat Saeheng]

Dear Arthur Goldspie,

Thank you for your quick response. Accoding to your answer, it is better to run the Simbiology in Windows and using the eGPU NVIDIA for the acceletating the calculation. So I will use the previous Macbook mid 2019 with Windows installation rather than buy the new one. Another question is that how can I set up the Simbiology to running on my eGPU?. It is due to when I use the parallell runing it takes a long time even my intel is 8 cores with 32 GB ram

Regards,

Teerachat

[Teerachat Saeheng]

How can I set up the external eGPU (Titan V) for the accelerating the calculation in Simbiology? in Windows Matlab version 2022a

Thank you

Regards,

Teerachat

[Arthur Goldsipe]

Although MATLAB supports NVIDIA GPUs, SimBiology does not have any built-in GPU support. SimBiology does have built-in support for parallelization using parallel workers with the Parallel Computing Toolbox.

If your work takes a long time even when using local workers running on each core of your computer, then I suggest profiling your code to see if there are any obvious bottlenecks that can be eliminated. If not, then the most obvious way to speed things up would be to obtain a larger pool of workers using MATLAB Parallel Server.

[Arthur Goldsipe]

Oh, and I should clarify that the built-in parallelization is achieved primarily by running each simulation as a different task. There is currently no way to parallelize a single simulation.

[Teerachat Saeheng]

Thank you

[Joss Knight]

There are no supported NVIDIA GPU drivers on Apple computers so you cannot even get an external GPU working unfortunately.

[Jeremy Huard]

Native Apple Silicon MATLAB is available as of R2023b. More information here.