I use MATLAB and EEGLAB to analyze recordings of brain activity. It’s essentially signal processing at this phase, filtering, independent component analysis, etc.
My 2015 retina MacBook Pro holds its own, but is a bit slower than the lab’s Windows desktop.
So I’m wondering about buying an iMac, maybe a iMac Pro, to gain a little horsepower and cut down on processing time.
MATLAB / EEGLAB doesn’t take advantage of GPUs or multiple cores without a lot of gyrations. So that leaves single core benchmarks to compare.
Here are the GeekBench scores
2015 rMBP 2.5GHz: 4037
2017 iMac 4.2GHz: 5567
2017 iMac Pro 3.2GHz: 4356
2018 iMac Pro 4.8GHz: 5689
So it seems there isn’t a lot of incentive to buy an iMac or iMac Pro, as for single-core tasks, at best they are 50% faster than my rMBP.
One variable here may be how representative GeekBench is of your workflow. You might get better scaling (or worse) than the GeekBench score depending on how your workload stresses memory, disk, and CPU.
For instance, the iMac Pro’s disk access is supposed to be quite a bit faster than earlier Macs, so a workload is I/O bound, that might provide a bigger improvement than the GeekBench score indicates.
And while he’s single core preformance is critical another thing to look at is system multitasking, as a developer most of my tools are also stuck to a single thread, but the value of the multi threaded systems is that I can run an emulator on one core while I compile on another, and run a server or two on another.
So my though would be, if you can’t run a single instance of matlab on multiple cores, why not think of your machine instead as a cluster and run several single instances doing different tasks on different cores?
The only hesitation I have in saying all that is that I have no idea if it is actually possible