Poor Macintosh Performance

I am posting this in case anyone has any suggestions for getting Cubase10 to perform reasonably on a MacPro. Right now I am getting poor performance to the point that the product is really not usable for me.

I have been doing a lot of disciplined performance testing lately with LogicPro, Cubase10 and Vienna Ensemble Pro. In all cases, Cubase10 is performing so badly that I stopped testing it. I finally came back and did some tests of just Cubase without VEP involved, and still just as bad.

How bad?

I have a test project with around 100 midi instrument tracks, each one feeding VSL instrument, and MirPro used for reverb. Nothing else.

On LogicPro I have this same configuration, same tracks, same music.

I measured average CPU usage of the computer every 3 seconds for a 3 min period of time, playing through the same section of music.

LogicPro uses 25% average cpu. Cubase10 uses 3x that amount and cannot play more then 30 seconds of music without dropping and glitching and eventually halting audio. Cubase is basically unable to play the project through.

With less than 50 tracks it seems to work on on my mac, but over 50, I’m afraid that something in the way cubase10 is coded is simply not performing anywhere even close to what LogicPro is able to do on the same computer with the same tracks.

Any suggestions to improve performance?

The Mac is 2012 5,1 MacPro, which is still very commonly found in the Apple community. Part of me is wondering if Cubase is using newer CPU instructions because most windows users are on newer CPU’s, but somehow this is making the mac version on the cheese grater run poorly. Just a wild guess though.

Kind of a bummer, I was hoping to try out Cubase more, but this feels like a waste of time if it can’t play 100 tracks like everyone else.

Hey stranger! :slight_smile:

Is there any difference in the VST/AU version of VSL do you know?

I’m yet to compare Mac/Win performance, and only just got Pro 10 installed.

But initial impressions in standard use are it runs slightly better on mac for me (Similar spec i7’s) - that said, my mac has an SSD and i’m yet to upgrade my Win machine to SSD, it’s possible the Win machine will fly once i transfer across to SSD. I’ll find out end of the week! :slight_smile:

I don’t know if i dare try it on my 8 core 2008 Mac Pro, but would be interesting to see how it runs on Xeon cpu based on what you’ve said.

What are your latency settings on both Logic and Cubase? Are they both set to 64bit?
I have a 2011 iMac and have done extensive testing between both Cubase and Logic Pro. My results were that both Cubase and Logic Pro ran about the same number VST instruments ( au for Logic ) but the culprit for Cubase was that the GUI would start falling apart ( laggy) before Logic Pro would.
Logics GUI would hold solid right until the audio would drop out whereas Cubases GUI would become very laggy before audio would start degrading.

“OSX”, but what version of the operating system are you actually using? That can make a huge difference in performance on the old Mac Pros. Certainly a 7-8 year old Mac Pro will not provide the same performance as the newer ones do (have owned both).

In answer to a few questions and comments…

I also had decent performance with low track counts. Up to about 10 tracks the CPU seemed normal use and I was hopeful. However as I built up the project into more tracks I could see the CPU usage increase steadily. At 10 tracks it was fine. At 20-30 tracks it was starting to be a little high, but still entirely usable. At 50 tracks, quite high, but still usable. At 75+ tracks it would not even play my cue all the way through.

The VSL plugin comes in AU and VST3 versions. Obviously I have to use the AU version in LogicPro. In Cubase I have to use the VST3 version. Performance is much worse with Cubase+VEP then it is with Logic+VEP, but the interesting thing in that scenario, is that when I look at activity monitor, its VEP that is grinding the CPU, Cubase isn’t doing much other then playing midi tracks at that point. However, when I remove VEP from the equation entirely… LogicPro actually uses a little less CPU then Logic+VEP, but Cubase uses much more…and can’t even play the project.

Latency setting is 1024 in all cases. I did try Cubase at max latency setting of 2048 but it still would not play all the way through. I also tried with ASIO guard on and off, no difference. I also tried both 32bit and 64bit engine, no difference. Its currently set on 64bit.

I am currently using OS X 10.12.6. The cheesegrater macs are old indeed, but they are still very common because Apple has not released a decent MacPro alternative since then. Steinberg can complain about that fact all day long but it will not solve the problem because many users are still using the cheese graters and that will remain to be the case for at least af ew more years. For my own part I have heavily invested in various other hardware, some of which depend on the PCIe bus, so I will not be changing my Mac any time soon. I will use LogicPro rather then cubase since Cubase seems to suck on this commonly used hardware.

I also have SSD’s, lots of ram and upgraded video card to RX580. VEP performs worse with HiDPI video display settings, its possible that Cubase is having a problem with that also, but the results I have posted were WITHOUT hiDPI modes.

The issue is not whether a newer IMacPro can out perform a 2012 Cheesegrater. The issue is that a cheese grater is easily capable of playing back 100 instrument tracks, people have been doing it for years, and LogicPro can certainly do it easily, but Cubase is not even remotely capable.

Seems as if you’ve made up your mind that this is the way it is, so guess no reason then to comment further. But obviously if all was as you describe and have concluded, few people would ever use Cubase. I too use both Logic Pro and Cubase 10 (on a 2014 Mac Pro ‘trashcan’) and my conclusions are totally different. Hope you are happy with Logic. Cheers.

Did you check your Asio Guard settings? Kind of the equivalent of Logic hybrid engine. I don’t know if the Logic trick of selecting an external midi track when playing back a project in Cubase also applies. It seems to save the butt of a lot of Logic users having CPU spikes.

I also have both C10(artist version though) and they perform rather similar since V9.5 but like you observed, my projects are rarely more than 20-24 tracks. Before V9.5 the fans of my MBP would kick in very soon when playing a small project in Cubase even though the ASIO meter would not indicate any significant load.

The current promo is very tempting but I’ve decided to let it go. I’ve tried to love Cubase often as it has a very good feature set(and Steinberg has been an innovator in a lot of areas) but it just doesn’t “click” with me. Just a personal thing. Emotional side of me would like it but rational brain says to don’t fight it. No reason that the Pro version would change it… :slight_smile:

Blink

I already said I am open to suggestions. Just clarifying everything I have tried. So far no bueno. My mind will be made up if I am unable to get reasonable performance from it. I cannot Reccomend cubase for Mac users right now. Do you have specific suggestions for how to make it perform @playz123?

One thought I have is that Cubase10 is coded to take advantage of newer CPU’s at the expense of running very slowly on the 5,1 MacPro. Just a guess since a few people with newer Macs seem to have no complaints. But i have no idea if they have actually pushed Cubase and tested it properly. What I can say is that Cubase is not capable enough to run 100 tracks on my 5,1 MacPro, where LogicPro handles it very easily with a lot of room to spare.

If anyone has specific suggestions about how to configure Cubase so that it will perform, please let me know.

Screenshot 2019-05-27 at 11.25.28.png
Not sure how relevant this is, but I’ve just upgraded to 10.0.20 today and it is sucking the life out of my system. Export time has tripled (and more). Tracks with amp sims and impulse responses take forever to kick in to action and keep dropping out. Shocking performance. Thank God I literally just finished an album on Friday.

I did a benchmark between LogicPro, DP, cubase10 and StudioOne recently. I used exactly the same midi tracks feeding exactly the same Instruments and FX, attempted to optimize each app as much as I could. Also tried each one together with and without Vienna EnsemblePro for hosting the plugins.

(I would post the graph results here but this forum won’t allow the image because its too wide or something)

https://docs.google.com/spreadsheets/d/e/2PACX-1vTPIyO1yivggupCLnQ7SeLRI0BY4yWf981bDCOzdLZZCLiAhhiRlm2anw60HEuRSDFEfO9-FwwJyYp1/pubchart?oid=1127774306&format=image

Observations

  1. Cubase10 was not even able to complete the task of playing back 90 tracks of VSL instruments through a reverb. It did ok up to about 50 tracks, though using more CPU then the others. After 50 it was rough and somewhere around 75 it could not even play it back without pooping out 10-15 seconds into playback. So there is no line on the graph for that. FAIL.
  2. Cubase10 combined with VEP was able to play all 90 tracks, so VEP makes it usable, but notice even that scenario is performing significantly worse then all the other DAW’s (see the blue line on the graph)
  3. Most all of the other combinations are clustered together in the same general CPU range, give or take. StudioOne performs slightly better then LogicPro which performs slightly better then DP, but they are all generally performing in the same range. Cubase is the outlier that is performing much worse.
  4. LogicPro really surprised me that it performed extremely well without VEP involved. That is the bottom dotted line on the graph. I have not been able to test studioOne without VEP yet, but I have a feeling it might perform even better or similarly as LogicPro does, both performing better without VEP overhead. DP somehow performs about the same with or without VEP involved. Cubase performs much better with VEP and can’t perform at all without it.

To be fair, you’re using an ancient Mac in DAW terms. Perhaps it’s time to invest in a modern machine, and use you old Mac as a VSL slave?

Basically you’re measuring Vienna performance here, not DAW performance. The DAW is using only a tiny portion of the cpu load in your test. Aside from the way Logic handles audio buffer sizes, all DAWs distribute plugin loads over the cpu cores the same way. Given equal circumstances, it’s unlikely Steinberg could do anything to get significantly better or worse performance out of Vienna than anybody else.

No, I ran each DAW alone also. The results chart shows that. Cubase is the only one that could not play 90 tracks. The others all played the tracks alone without sweating. Cubase however was able to complete the task WITH Vienna Ensemble as a crutch, but still much worse performance then the others when combined with Vienna.

The 5,1 MacPro is still actively used by many musicians world wide, despite being “ancient”, its multicore benchmarking scores are right up there with many current modern machines. It has PCI slots and many people have invested in additional hardware based on that. Its preposterous to say that this machine can not be able to be used.

Needless to say, LogicPro, StudioOne and DP all run fine on it. 90 tracks with much room to spare for many more, both with and without VEP. Cubase10 can only really get to 50 tracks and that’s it. unless VEP is used together with it…then it can complete this task, but still using up enough CPU that its clear it would run out of tracks way before the other DAW’s do, even with VEP helping along.

IMHO, this is a more accurate way to say it: “When I tested Vienna in Cubase, I wasn’t able to play 90 tracks.” Even if they wanted to, Steinberg couldn’t significantly change the number of tracks you can get with Vienna :slight_smile:. The assertion you are making is like saying, “my phone loads web pages faster when I’m in a Toyota than when I’m in a Nissan”. Anything’s possible, so you might be right, but the far more likely explanation is that you have something different in your test with Cubase than with the other DAWs. For example VST vs. AU, buffer size handling (this is why Logic performs differently than the others), streaming performance, or some other factor specific to Vienna.

Dewdman42:
I have three questions…

  1. What video card are you using? I have an older MacPro than you and had to upgrade the stock card a few years back to improve performance. I have a suspicion that some of the changes they’ve made to accommodate new monitor resolutions has impacted the performance on OSX using underpowered video cards. And like you, I also see an improvement difference running duplicated LogicProX projects. I have a feeling your card isn’t the issue due to the performance of LPX. Just thought I’d ask, though.

Here’s the thing…Cubase 10 performance has indeed taken a hit compared to 9 and 9.5…so you have to ask the first rule of beta: if it stopped working, what changed? Hopefully, this gets addressed in the next maintenance update.

  1. What did you generate that chart with?
  2. Do you see any improvement in the settings for ASIO Guard including disengaging it completely?

No, that is not the correct way to say it, you have it backwards.

  1. Cubase+VEP can play 90 tracks, but not nearly as efficiently as the other DAW’s
  2. Cubase without cannot play 90 tracks, it craps out somewhere between 50-75


Even if they wanted to, Steinberg couldn’t significantly change the number of tracks you can get with Vienna > :slight_smile:> . The assertion you are making is like saying, “my phone loads web pages faster when I’m in a Toyota than when I’m in a Nissan”. Anything’s possible, so you might be right, but the far more likely explanation is that you have something different in your test with Cubase than with the other DAWs. For example VST vs. AU, buffer size handling (this is why Logic performs differently than the others), streaming performance, or some other factor specific to Vienna.

You are saying a lot about nothing. Cubase is the thing that can’t play 90 tracks WITHOUT vienna. Only with vienna can it be done and not as efficiently as the other DAW’s. Vienna is actually rescuing cubase…you are accusing the wrong product of poor performance, Vienna performance is extremely good and is not the problem.

RX580 with metal support. Had to be upgraded in order to run Mojave.

As stated before, all the other DAW’s are running very nicely on it. Cubase10 runs like crap.


Here’s the thing…Cubase 10 performance has indeed taken a hit compared to 9 and 9.5…so you have to ask the first rule of beta: if it stopped working, what changed? Hopefully, this gets addressed in the next maintenance update.

I never had cubase9, so unfortunately I can’t compare.

  1. What did you generate that chart with?

tomorrow morning I will post a link to the google sheets spreadsheet that has the numbers and link to another forum post where I describe the tests in more detail.

  1. Do you see any improvement in the settings for ASIO Guard including disengaging it completely?

Yes I tried on and off, large and small and bigger audio buffer too…none of them can play more then 50-75 tracks in Cubase alone…but as I said…Vienna can handle it…even with sluggish Cubase as the Host. The other DAW’s handle it fine with and without Vienna…

Is it possible for me to install and use Cubase9 even thought I purchased Cubase10 as a first time cross graded user? If Cubase9 performs better then Cubase10 on my mac, I would like to try that.

Some people asked for more information about the test i did. This link on the logicForum has more info. Its only there by coincidence, because it came up in a discussion…

https://www.logicprohelp.com/forum/viewtopic.php?f=1&t=142371&start=80#p737261

Happy to answer any questions about the test I did.

I would like to test Cubase9, but it is actually a lot of work to setup the project in each DAW, so I’m not sure whether I will get to it.