Well, I was curious so I did a direct comparison using the same project running on the same computer. The only thing different is one time I ran the project in Windows 7 and the other time in OS X Yosemite 10.10.5. I did a screen recording of the results on both platforms and slapped together a video. The results are surprising to me as to how close the performance is.
I’m now wondering how close the performance gap is between Logic and Cubase on OS X. Here is the link to the video below:
Best I’ve seen in a cross-platform A/B DAW comparison without the usual caveats. disclaimers and variables. Excellent work. Thanks.
The only thing I can imagine as a variable is whether there could be an advantage using a Hackintosh vs Apple hardware. Unlikely, but I suppose it’s possible that Apple has some proprietary whatnot going on behind the scene that loads the system slightly more. But that is total speculation on my part.
This sure challenges the concept that Core Audio is a problem. At least with RME hardware/drivers.
The good old “CoreAudio makes the Macs not as efficient as PC” topic was only particular to the buffer sizes smaller than 128 samples running in real time. Even with the testing done when ASIO guartd was not available, at the buffer sizes larger than 256 samples, Macs and PCs performed similarly IIRC.
What may illustrate the difference could be, tracking 24 tracks of audio at 32 samples buffer on both platform and see which one performs better. I think the PCs should theoretically win in this case because of lower DPC latency.
If you are not doing a full band tracking, and have the ASIO guard turned on, then yeah, you should theoretically have the same performance on Mac and PC.
This is not quite the correct way to test this, because both the ASIO meter and the operating system provided CPU load are notoriously unreliable indicators in realtime applications, where dropouts can occur much before you hit 100%.
I think the result you got was close to correct, because with ASIO Guard 2 the performance is likely quite similar, since the OS X kernel’s poor low latency performance is mitigated by it.
To get an actually accurate result, you should for example download the test sessions from dawbench.com, or build something similar. The method is to keep enabling tracks until you experience dropouts and repeat that with different buffer sizes on both platforms. Whichever platform runs more FX or VI without dropouts is the better one – you should disregard the ASIO/CPU load meters entirely.
The problem is that DAW bench is old. Or at least it looks like everything is old on their site. I could load up a bunch of Halion Sonic instances or something and see which one breaks first at 32 samples. The thing I’ve noticed now is that actual 32 samples is really only for armed tracks to record and I believe that is due to ASIO guard. So all unarmed tracks are buffered using ASIO guard and then tracks actually being recorded to go down to whatever latency you set it at.
Recording live audio tracks shouldn’t really be different between each OS unless you are monitoring through your host and not doing direct monitoring through your audio device. I was able to record a full drum kit with an IPAD running Cubasis.
I went back and disabled ASIO Guard and the project behaved pretty much the same on both operating systems. 32 samples with this stuff going is a no go. 64 sample buffer was tough and operating around an average of 75% ASIO performance but would peak red every so often on both operating systems. 128 sample buffer was the same result on both operating systems at about 60% without any overloads.
2 things I have realized here is that ASIO Guard works and the performance gap between the different operating systems isn’t large now. Those old tests from DAW bench are from Cubase 5 I believe and OS X 10.6 and Windows XP.
I will say that regardless of OS Cubase seems to work really well now.