Hector Martin

I expect things will improve over time, but before moving your workflow to M1 Macs, *test* whatever it is you're doing. Still looking forward to doing my own tests under Linux, especially at low buffer sizes :-)

I suspect that the difference in memory size between the Macs affected the result. The author used reverb effects that might consume a large amount of memory. Still looking forward to an additional comparison with the same memory size.

I can't imagine a single DSP reason for a reverb plugin to need more than a few megabytes of RAM at most.

Honestly I'm not confident of my inference though :P but at least the Intel Mac in the article has 4x bigger memory than the M1 Mac and it might result a difference other than the difference of computational capacity. (BTW I started to donate Asahi Linux last month. Keep it up!)

I want the author to show more values like a graph of memory consumption and CPU % etc.

I agree that it would be nice to confirm, but I thought about this and concluded that a reverb plugin shouldn't be eating RAM. i.e. if it is that would be a giant bug, but I don't think the fabfilter guys are that terrible at plugin dev ;)

I got some info related to reverbs:https://twitter.com/t_okmt_12sound/status/1345287358917197829 …

Yes, I considered that. Sampling reverbs (i.e. convolution-based ones) should use at most 4x reverb tail length for the impulses (stereo cross pairs), which is less than 8MB per instance for a 10-second reverb time (which is quite long).