M2 Pro Mac mini

Is anyone here planning on getting one of the new M2 Pro Mac mini systems? I would love to find out if it can handle these settings:

SDM Integrator: FIR2
SDM Conversion: XFi
48K DSD: Checked
DAC Bits: 20
Default Output Mode: SDM (DSD)
SDM Defaults:
1x = poly-sinc-gauss-xla
Nx = poly-sinc-gauss-hires-lp
Modulator: ASDM7ECv2
Bit rate: 48K x 512
Multicore DSP: Greyed
Adaptive output rate: Checked

I know many DACs don’t support 48K DSD so that is not required but DSD512 is what I am curious about. The M1 Mac mini can handle the above but only at DSD256…

1 Like

no idea… but when my trusty 2012 mac mini roon core goes to the great beyond, I’ll definitely be looking at the new M2 Pro to replace it

2 Likes

In couple of months I may upgrade my current M1 to the new 12-core M2Pro.

But my estimate is that it won’t do DSD512 tho ASDM7ECv2.

It can certainly do more filters and such compared to old M1 Mini.

3 Likes

I have the basic M1 MacMini and I’m getting the M2 model. I’d like to turn the M1 into my core. Can you advise of any “gotcha’s” I might look out for?
Thank you all for sharing.

Roon Server headless (but keep a blank HDMI connected, just a connector with no monitor, If none the mini is just searching for a monitor). No sleeping, disable all wifi, bluetooth, locations, software upgrades, time machine, Spotify search… just all that you can find. A wipe and reinstall of osx would not hurt before installing roon. Later if you want I can send you screenshots with all my settings (m1 mini)
Enjoy
edit: with all I could do the sistem it is between 1.4/2% of cpu

5 Likes

Trying to do exactly that…moving from my Intel i3 to Mini M2 or Mini M2 Pro is decided. Need to know how I can keep this Mac M2 fanless box running for my Roon ARC in my car (I do use it on the go) without having the Mac on sleep mode.

Hello Jussi, I would like to use Mac Mini M2 Pro as Roon core with your great HQPlayer software. Regarding configuration: will HQPlayer make use of the GPU or is only the CPU relevant? Is there a significant advantage for this application in going for 32 GB RAM vs. 16 GB? Thanks, Andreas

Apple GPUs cannot be used at the moment. Apple is not offering suitable programming interfaces for their GPUs.

Right now, only Nvidia GPUs are supported on Linux and Windows.

I have some support implemented for AMD GPUs on Linux as well, but unfortunately they want to protect sales of their most expensive data center and workstation GPUs by limiting their programming interface support only to those models. This is like Nvidia over a decade ago…

So on Apple hardware, only CPUs matter at the moment. Going from 16 GB to 32 GB is much less advantage than going from 8 GB to 16 GB. Although it is good idea to scale amount of RAM with your CPU. So more cores you have, more complex things you can run and more RAM is being used. Things like convolution and sinc-L type filters are RAM hogs.

2 Likes

Hello Jussi, thank you very much for your answer - it is highly appreciated! Thanks, Andreas

fyi, I I tried DSD512 ECv2 on an M2 Mac Mini and, as expected, the 10% increase in clock speed was not enough to make it happen. Maybe the M3 …

4 Likes

Which filter and source rate though?

Does “poly-sinc-gauss” also fail?

Yes, I tried a very simple controlled case. 44.1 → 44.1 x 512 with poly-sinc-gauss-long. If this doesn’t work them nothing will.

Again, I could see that the cpu load was improved by roughly 10%, in line with single core improvement, but it’s just not enough. One of the power cores is pegged and hitting it’s head frequently - I am guessing this is the problem. The other 3 in the 85% range. The load is not evenly distributed - I assume because of the way the modulator works. I suppose if @jussi_laako could optimize load across the power cores one might be able to eek something out?

1 Like

gauss is lighter CPU load than gauss-long

But not much.

Thanks for testing

Sure, I tried poly-sinc-gauss. Possibly a titch better but still can’t keep up.

Final work distribution is up to macOS. It doesn’t let applications control thread affinities the way other OS do.

3 Likes

You’ll get more success with more filters at DSD512 if you use AMSDM7EC512+fs on M1 and M2 chips.

On another note, I have more working combinations on Ubuntu i9 9900k with titan x cuda than with a M1 studio, hence the reason I keep the Ubuntu machine around. :wink:
The M1 studio is my roon core, everyday use hqplayer host, plex server, mail spam server, and zfs server (lacking ecc memory though :-1:). Lots of memory pressure over here, usually 80-85% utilization. Works well considering.
Probably should’ve opted for 64gb vs 32gb of ram. Oh well. :face_with_raised_eyebrow:

2 Likes

Does this provide some hope?

2 Likes

I could be wrong, but it was my understanding that ASDM7ECv2 @ 256 was generally speaking better sounding than non-ECv2 @ 512. But obviously it would behoove me to try it out … :expressionless:

[Edit: AMSDM7EC 512+fs @ 512 works without issue on my M1. Will do some listening later on. Thanks for the nudge ]

I’m currently running a M1 MAC Mini )16 MB of AUM and 1TB storage…for my Roon Core with zero issues…though my Library resides on my 2012 MAC Mini Roon Server… I will probably replace that server with an M2 Pro MAC Mini this fall…