I did this because I am having a lot of crashing in OSX and someone advised Bootcamping the computer and running Windows 7, suggesting one would experience a considerable performance increase. So I did.
Here’s the setup:
Cubase 7.5.2 on a Mac Pro dual Xeon 2.26 (8 cores with hyperthreading turned off) with 32GB 1066 DDR3 comparing performance on OSX 10.9.2 versus Windows 7 (all updates and usual DAW-friendly settings).
The project is at 24/48 and has:
31 stereo tracks and 4 mono tracks
SD2.0 (just under 4GB)
9 VSL instruments loaded into Kontakt
5 instances of Absynth
~43 stereo plugins
uses 5.56GB RAM
I compared many aspects of this setup - boot time, Cubase launch time, and, of course, the playback performance. If you want to go straight to the latter, it starts at 4:13.
Please understand I’m posting this simply to furnish information (and hopefully save others from wasting the incredible amount of time it took me to do this). Further, my closing conclusions solely address and are based on this experiment and its results. I’m not saying one type of OS or computer is better than another.
All the music in here is mine. In order of appearance:
Brock, whilst these sort of comparisons are always useful, your test needs a few extra things to be useful to many other people.
I don’t see where you mention what buffer you are working at. Historically Windows has been better than OSX at lower buffers, and this tends to even out the higher the buffuer gets.
You have turned off Hyperthreading, so have already partially crippled your machine. Again historically OSX has been pretty poor at thread synchronisation.
Your soundcard is one of the worst performing soundcards ( using Windows 7) at low latencies, so again it depends on how you like to work.
Certainly on your system, set up the way you like, there seems to be no advantage either way. However, as your system is crashing (hence you doing the tests) it seems more likely that there is something more basic going on.
Interesting, I was tempted to try bootcamp on my machine i hope of getting better visual GUI performance(less lag, snappier zooming etc ). Did you notice any difference in that area between the 2 os:es?
In the video, I do cover the buffer setting (@ 4:01). It was set to 256 using the Focusrite driver on both machines. If you pause the video (and squint really hard!) you’ll see the the latency was actually lower on OSX.
Steinberg recommends disabling hyperthreading on PCs (see https://www.steinberg.net/nc/en/support/knowledgebase_new/show_details/kb_show/optimizing-windows-for-daws.html). It, in part, reads “7. Disable Hyper-Threading if your CPU supports it (e.g. Intel i7). Further information on how to disable it can be found in the mainboard or computer documentation.”. In OSX the performance is about the same either way. To me the CPU use seems just a bit less spastic with it disabled so I’ve been running it that way.
I don’t think latency has anything to do with the issues I was tackling. I merely mentioned the buffer setting to show that both systems were set the same so people didn’t think I crippled one by setting the latency really low and high on the other. It’s quite evident to me that latency doesn’t have anything to do with the crashing.
I do agree with you that something basic is causing these crashes and I look forward to finding out what it is!
Hi claesbjo,
The GUI performance (like the responsiveness of the meters and zooming) seemed better on Windows 7. Otherwise the performance was snappier on OSX.
The hyper threading information is woefully out of date. DO not turn off hyperthreading on any recent i7 you’ll lose performance, in fact as already stated you’ll really cripple your machine under windows 7 /8 when running Cubase.
Steinberg really do need to take that info from the data base, it was relevant in certain situations about 4 years ago. Since the on die memory controller i7’s this has not been the case.
I did the same test a few years ago and OSX was like working on Valium in comparison to windows at the time.
So not a particularly low buffer, so the differences in any case would be less than if you were working at 64 or even 128
Yes that is very old advice, and I haven’t done that for years. In Windows the performance would be at least 30% better with HT enabled. I can’t believe that there is no difference on OSX, or performance would be appalling when compared with Windows.
No, I don’t think latency has anything to do with your issues either, but was commenting more from a performance point of view. In fact, the only people I know who have been having frequent crashes have managed to solve them by switching off ASIO Guard, but I don’t know what to suggest to you, because I would imagine that you’ve already tried that.
I’ve looked at the logs. Far as I can tell, it’s a moving target. As to not derail this thread, I’ve given a pretty thorough recount of what I’ve done so far here Steinberg Forums. I’m pretty sure it’s a matter of a couple bits of software not playing nice together and that it’s something will require programming to fix.
Fabio at Steinberg wrote in part “As you correctly guessed, the problem is with the Steinberg Hub. In the crash report, Cubase is loading and invoking the Hub, which in turn calls the WebKit, the WebCore and eventually crash the system library.”
@DG - While 256 might not be a low number, the effective latency (of my aggregate device - Liquid Saffire 56, Axe-Fx II, and B&W MM1s - reports 8ms input and 6.938 output at 24/48) is imperceivable - even when programming MIDI (drums included). For most projects I can run at 128 (4.649 in, 5.805 out). For audio, I’m using direct monitoring from my Focusrite so it’s moot in that regard.
With the Hub off, I don’t get crashing on startup. I’ve known this and am presently operating this way. That said, I’d like to be able to use it.
RE: Hyperthreading, I wish I’d known that while doing the test! After spending 1.5 days installing Windows 7 (and all my plugins and sample libraries, and running the tests) I took it right off. It was so time consuming that I’m not doing it again. A Steinberg rep told me it doesn’t really provide much, if any, performance difference (Bootcamping Win7 and running Cubase from that partition) - and this is a support tech that’s been with the company for several years and did this to his own system.
I just ordered new processors for my Mac Pro (dual 6-core Xeon 2.93 boosted to 3.33). When I get them I’ll experiment again with the hyperthreading on OSX…if for no other reason to see 24 processor windows. lol
I wonder if having hyperthreading OFF is the reason why my mix console freezes on monitor #2 in full screen mode when not used for period of time while monitor #1, #3 and #4 are just fine. I turned it ON again.
I actually had a Steinberg tech support guy recently tell me the exact opposite. He said Hyperthreading should be turned on and that the advice on their website to turn it off was outdated and ought to be changed. This was with reference to a Windows machine.
I believe you misread what I wrote. The tech told me using Win7 via bootcamp on a Mac did not provide a performance increase. His comment didn’t pertain to hyperthreading.
@SteveInChicago
Good find. You’ve been pretty vocal about Mac/OSX. What’s your take on using hyperthreading in OSX (as RE: Cubase)?
According to that SB link, it sounds like enabling hyperthreading with ASIO guard on is the way to go (regardless of OS) and that the only reason one would disable it is if you are recording a significant number of tracks simultaneously.
The link pertaining to hyperthreading did not work for me. I would like to know what it said.
This is quite a useful topic and whilst I would like to try it out I am a bit nervous that I might destabilise a system that is working quite well without. Presumably I would be able to run more stuff.
lol