This for anyone moving from Intel to Apple Silicone. I’m playing with VDMX on a base iMac M1 with 16GB unified memory. For a low spec machine it performs really well but you can see the high memory pressure which would be an issue on a long set. Saying that I have a few layers going but nothing crazy and it feels rock solid.
I also tried the ISF Water simulation which kills my 2017 MBP and the M1 ran it without any real issues. This is a good test if you visit the Apple store and try out various machines.
My conclusion so far is that 32GB unified memory would be the base but you can get away with 16Gb if you are doing simple video mixing. Obviously the more GPU cores you have the more memory you would want.
The question is not necessarily M1 or M2 but the balance between GPU cores and memory.
It would be really useful if someone who had access to multiple Mac configurations could come up with a practical and cost effective base CPU, GPU, memory spec. This would save people a lot of time trying to work out what to buy (we can’t always afford the high spec machines).
A lot of the earlier, intro / tech troubleshooting cues we received from M1 Mac Mini users came from weird errors, issues, and crashes when they tried to push the M1 Mac Mini too hard with VDMX and other applications. These are entry level machines. Same as a Macbook Air. You wouldn’t expect them to out perform another mac or PC 2-4x the cost, but we had a number of people pushing them too hard, and the built in activity monitor was not a good indicator of what was going on in the backend.
When the CPU/GPU, Memory, and SSD were all being pushed hard, and the OS started diverting more memory to SSD swap, all sorts of strange non-reproducable issues popped up. The universal build of VDMX helped a good bit with that issue, but they are not high performance machines.
I’d say you are right, and to avoid swap related issues, 32GB of memory is a solid starting point. I’m curious how the M2 machines compare.
Yes, I agree the Activity Monitor doesn’t provide the right info to make a proper evaluation. Also traditional benchmarking doesn’t feel as it it fits for the VJ / realtime workflow either. My simple logic suggests more memory helps with loading files into RAM/swap and more GPU-cores speeds up frame tiling/rendering.
Seems like the best strategy is to choose a machine based on your budget with a minimum of 32GB unified memory. Then if you can stretch the budget to bump up the GPU-cores and unified memory to the next level to give you some headroom.
