Having had to fire up my AMD 9950x machine today to test an ‘issue’ i htought it would be interesting to see what the performance between these 2 machines was like.
So the Mac:
M3 Studio Ultra with 1TB internal and 96GB ram -MAc OS 15.7.3- RME UFX III
The PC:
AMD 9950x-64GB DDR5-1TB system drive. Windows 11- UR 22-C
both running the same 256 Buffer @32/96k
As you can see the Windows machine is performing a lot better than the Mac Studio Ultra… which is not what you want to see after dropping $4000 on a new machine
Anyway, i htought it was an interesting comparison runnign the same project side by side.
?? once the ASIO guard meter hit’s the red you’re finished….. so it’s as an important metric as the peak.
ASIO guard is usually the limiting factor I’ve found when testing DAW performance in Cubase. ASIO guard will hit the red long before CPU use is maxed out. And, as I’ve said, you’ve then hit the limit.
Not sure how one would think a completely different device using a fundamentally different core audio engine, connected with an entirely different bus structure is “interesting,” but I suppose that’s a personal inference. But also a really great indicator of how these comparisons and comments are little more than specious corollaries based on non sequitur metrics.
I think it’s interesting that the M3 ultra, is performing worse than my AMD 9950x windows machine on the same project, despite having an ‘inferior performing’ Audio device. You’d think a 28 core $4000 MAc would outperform a 16 core AMD machine.
the M3 wouldn’t playback a mix I did last year as it was completely in the red and crackng and popping. I had to go back to the windows machine to do a mix revision.
That was in reaper too which utilises threads well.
I can’t speak for Steffan, but the reason I think it means nothing is because there’s no such thing as ASIO on macOS in the first place. As such, Steinberg had to introduce an ASIO-to-Core Audio translation layer for its “ASIO-centric” DAWs to work. APM talks to ASIO, which doesn’t really exist on macOS. Core Audio handles all threading and audio DSP itself, irrespective of what ASIO thinks or wants. I don’t even bother considering what APM says on macOS other than if I needed to see what track Cubendo thought may be contributing to DSP-thread saturation. There are certainly considerations for the underlying ASIO dependency, but insofar as caring what ASIO Guard says, I couldn’t be bothered
But ASIO guard hitting the red is what I’ve found to be the limit of performance in a project, you can’t continue once it’s at it’s Max. You get dropoutas and garbled audio.
If it had no bearing i wouldn’t be bothered either. As I said, the M3 coudn’t play back a mix recal done on the 9950x today, that’s an issue for me and ASIO guard is what is Maxing out on the machines with mixes.
Oh, I never saw that you said that. I saw a screen shot of ASIO Guard ~ 55% vs 25% and you comparing the two based on that. If you’re saying you have applicably identical (meaning, same plugins and setting though the plugins themselves are Apple Silicon vs x86 and you’re not running in Rosetta) and the exact same project runs without glitches/artifacts on the AMD yet has dropouts on the M3 - irrespective of what ASIO Guard says, then that would certainly be interesting.
The last time I saw you discussing your new M3, it was in the context of how that Dawbench app failed to provide any value for your decision to buy the M3 (which, of course, it wouldn’t anyway) and instead of you sharing your project, you opted to do something with the benchmark tool instead. It thought that was a bit weird, but certainly your call. It’s just odd that you continue to have such a distinctly different experience than what I’ve experienced with my Ultra.
This is the first time I recall you actually having a project that runs fine on the AMD but that glitched on the Ultra with [near] identical plugins and hardware. I see ASIO Guard being different, but my ASIO Guard is different just between Cubase and Nuendo with the exact same project AND preferences (as in literally copying the Cubase Preferences over to Nuendo). ASIO Guard is one thing (which I’ve always ignored without any issues at all) but the same project glitching on the Mac is indeed interesting. That makes me wonder what plugins you’ve got and how differently they operate on x86 vs Apple Silicon (again, presuming you’re not using Rosetta, which I would doubt you are).
DAWbench isn’t a benchmark “app” , and the session Marcus based the information on in the initial report, is based on a RW MIX session that is part of the DAWbench Suite, that can be run by anyone in multiple DAW’s.
Instead of just attempting to cast shade on DAWbench, it would perhaps be a better idea not to be so quick to dismiss the reports in an ignorant and arrogant manner.
This whole angle that ASIOGuard has no correlation to the overall performance of Cubase, and/or that there is ‘No ASIO” on Apple OS , instead running some wrapper/translation layer, and that somehow renders the ASIOGuard Performance reading useless, just shows the complete ignorance of what is actually being displayed and conveyed by the respective parameters in the Performance Meters.
No matter what Steinberg decide to label the parameter, the simple fact that the extended buffering of the audio engine accumulates and overruns prematurely is not in dispute , and your assertion that “Core Audio handles all threading and audio DSP itself, irrespective of what ASIO thinks or wants” is based on what exactly ? The amazing thread management being displayed by the M3 Ultra under RW mixing environments ignoring 50% of the Logical cores ?
Re plugins , I have already ported 3 separate RW end user mixing sessions to DAWbench, and duplicated the exact behavior and performance dynamics using a complete suite of alternate but similar plugins, so it has zero to do with plugins, zero to do with ASIO wrappers and abstraction layers, we have a pretty clear idea where the bodies are buried.
Why? It’s got no real-world value. If it did, Marcus wouldn’t have been complaining about how he wasted money buying a Mac, right? You guys obviously think it has value, and that’s fine. I don’t. Sure, it’s neat and all, so well done and all that. You’re clearly upset about it, so I’ll just leave it at that. FWIW, you seem really quick to attack people who disagree with you and your benchmarks. The last time you got all upset about how “Mac people refused to participate” in your testing and even went so far as to insult an entire class of Mac user by claiming we all “knew something was wrong and was hiding it” or some other such nonsense. I guess if anyone knows “ignorant and arrogant” it’s you, huh?
I won’t waste anymore time with this nonsense. I’ll make sure I don’t see posts from you or Marcus anymore. It’s not worth it. Good luck though.
Again, you are just showing you ignorance and arrogance !
Marcus’ choice to go the Mac was based on comparative performance testing we had done across both x86 and Mac AS systems over a period of 6 months, based on one of his own Real World sessions, but the initial testing was done on Marcus’ M1 system with lower core count, that showed better thread management and ASIOGuard dynamic.
Thats what he based his decision to go cross platform on.
This whole new adventure was due to discovering that once you get above a certain logical core count, the thread management and performance dynamic on MACOS also falls on its arse , and as much as you may hate to accept, its DAWbench that revealed and highlighted the performance dynamic and discrepancy.
I have no problem with people questioning DAWbench, but I have little tolerance when the dismissal is coming from a position of complete ignorance, as yours has been.
Re Mac users refusing to participate in the testing , yep, still an ongoing enigma, and re insulting a whole class of Mac end user , really ?
Where did you pull those quotes from ?
With you now speaking for a collective of “we”, what class are you representing exactly, as none of what you just blurted out has any relevance nor technical comprehension of what is being reported.
This isn’t that what the screenshots are showing. They show nothing performance relevant.
I’m just seeing irrelevant information regarding the real world performance.
look at the ASIO guard levels…. when that reaches the end it’s game over…
the M3 is more than twice the level of the AMD9950x…
what are you NOT seeing?
really I’m baffled by your comments.
I don’t have anything to prove here. I’m doing this as a ‘service’ to my fellow Cubase users hoping it will help people…. that’s my ONLY motive here…. to HELP people….
How many people here have got a top of the line Mac computer and a top of the line windows computer who can run the same session on both for comparison???
I think I”m in a pretty unique position as we speak to be able to do this so thought it might be helpful….. rember that word again… help?…. to post the same project running on both machines.
