Mainstage CPU spikes

Clicking on the Release Notes in the Help menu takes you to non-existing webpage.

http://support.apple.com/kb/TS4027

does that get you there?

Hi D3,

First of all, http://support.apple.com/kb/TS4027 doesn't mention that you have to go to the App Store to find any mention of the upgrade. The upgrade is not mentioned any where else except in TS4027. I'm not used to going to the App Store yet, so it took me quite a while to find it.

I did the upgrade on my MBPro 3.1. It shows that MS2.2.1 uses almost exactly twice as much CPU load as MS2.1.3. The older version runs reliably. The new version simply requires more horsepower than I have to run my concerts.

My solution (albeit just a bit spendy) is to get a UAD Apollo quad for IO and processing, and use a new MBPro for musical instruments only - and all connected together with Thunderbolt. UAD is claiming less than 2ms latency for the entire rig, and enough processing power to keep my ten little fingers busy... The Apollo will be about $2500, and the new MBPro (with 8 gigs, 512 SS drive, and hi res screen) will be about $3700. But hey, won't it be nice?

Maybe you are running into http://support.apple.com/kb/TS4157, MainStage 2.2: Performance may appear to decrease on some dual-core computers. I have noticed that although the CPU meter reads higher in 2.2.1 compared to 2.1.3, I get better performance (i.e., I can run more channel strips) compared to 2.1.3 before overloading it. Also I found this tech note which relates to MainStage's processing: http://support.apple.com/kb/HT5155

Because the audio engine in MainStage 2.2 is completely different from MainStage 2.1.3, you can't compare the CPU meters between these versions. The only point interesting was already stated: how much can you actually do.

Hi everybody,

I ran the same concert in 2.2.1 and 2.1.3. I measured the CPU usage using Activity Monitor, because I was getting such an awful response from 2.2.1 on my MBPro 3.1.

This is what I got:

2.2.1 idled at between 15% and 30%. It would spike to 100% erratically, every few seconds, and the audio would glitch or cut out if I played a single note. 256 samples, no buffer.

2.1.3 idled at 37%. It would go to 70% with pushes to 89% based on demand when I played normally. 256 samples, no buffer.

128 samples doesn't work. In 2.1.3, it glitches the audio, and in 2.2.1, it puts out a loud, weird artifact that sounds like massive digital distortion, and no real tone at all. In 2.2.1, I had to put the samples to 1024, and back to 256 to get the machine to recover. I shut off 2.2.1 and started up 2.1.3 while 2.2.1 was making this noise, and 2.1.3 worked just fine. When I restarted 2.2.1, it still made the noise until I did the 1024 to 256 entry. I'm guessing that the overload on 2.2.1 when I set it to 128 was so bad it didn't record the 128 entry properly.

In addition, with the fan running at 5000 rpm (I force it there to keep the processors cool), 2.1.3 ran at 208º with regular excursions into 212º territory. 2.2.1 went to 218º, which I know is high enough to force the processor to slow down to keep from overheating. My poor old MBPro can't handle 2.2.1 the way I use it.

I don't use the buffer at all, because the latency goes from not very good to totally unacceptible.

In 2.2.1, the roundtrip time was 18.9 msec, and the output time was 11.7 with samples at 256.

In 2.1.3, the roundtrip time was 18.9 msec with samples at 256.

lohogan wrote:

I measured the CPU usage using Activity Monitor, because I was getting such an awful response from 2.2.1 on my MBPro 3.1.

That's the first issue: Acitivity Monitor doesn't work with realtime applications. On the other hand: a MacBook Pro 3.1 can't be faster in MS 2.2 compared to 2.1.3: the CPU doesn't have more resources to be used.

lohogan wrote:

In 2.2.1, the roundtrip time was 18.9 msec, and the output time was 11.7 with samples at 256.

In 2.1.3, the roundtrip time was 18.9 msec with samples at 256.

Please read the articles from Apple, because you are actually comparing the wrong things. 256 in 2.1.3 is not the same as 256 in 2.2:

http://support.apple.com/kb/TS4027 MainStage 2.2 release notes

http://support.apple.com/kb/HT5155 MainStage 2.2.1: Optimize latency

Hi Blueberry,

Activity Monitor has a readout very much like the CPU readout in MS. The two jibed, but Activity Monitor gave separate readings for the two processors, and also had per-app percentages and a history meter. Activity Monitor has been very helpful for realtime applications for me. It uses about 1.5% of CPU resources.

I know nothing is the same between the two, but the performance of 2.2.1 was noticably less than the performance of 2.1.3. I get barely satisfactory results with 2.1.3, and I can't get 2.2.1 to run well enough to use. My concert has 9 instruments running on some patches, and 5 guitar pro with bass patches, plus three audio channels and 8 output channels. It takes up 1.1GB of my 4GB memory. It sounds awesome when it works!

So is the bottom line that 2.2.1 will probably give me worse performance on a Core 2 Duo 15" MBP? than 2.1?

lrizzo2001 wrote:

So is the bottom line that 2.2.1 will probably give me worse performance on a Core 2 Duo 15" MBP? than 2.1?

That's been my experience, yes. CPU and RAM usage is approximately doubled in 2.2.1 vs. 2.1.3 for me

You know, as long as CPU doesn't go over 100%, I'm not sure why I care about the actual performance. I don't need CPU cycles left over to read email or something during a performance.

Reliability is much more relevent.

I can't give an apples to apples comparison on CPU/RAM usage between 2.2.1 vs 2.13, as I upgraded everything at the same time (computer, OSX, and MS version). But if usage doubled between versions on the same setup, as in madcow's case, how and why would that be happening? Its not like the application added a lot of powerful features. I would expect it to get more efficient with bug fixes, based on the number of complaints. Maddening indeed!

And I agree with dhjdhj: reliability is #1. This is a live performance application; its not like you're sitting at home with no one listening or caring if you have to restart your application.

Reliability can be managed. Keeping CPU usage under control is one of the more direct ways to control it.

2.2.1 uses twice as much CPU on my machine as 2.1.3. I like the improvement in the concepts on 2.2.1, and if they can find the problem and get the CPU usage reduced (hopefully below 2.1.3) that will be a great thing. Until then, I have to use 2.1.3.

I wonder how the other brands, Ableton, Max, and even Logic, keep the usage down...

lohogan, have you tried removing all of your third party AUs in case one of them is conflicting?

The other apps are not really comparable since they do not achieve the same task -- it would be difficult to achieve the exact same end result with those apps.

Maybe. I've developed an environment for myself that was originally designed to give me the same functionality as MainStage but at this point has gone much further. It's not as pretty as MainStage and you need to have MaxMSP installed to use it but I've used it now in three live gigs as well as numerous rehearsals and it has been 100% reliable.

I've implemented the same kind of indirection so that basically any physical knob,slider, or pedal can be mapped to an an arbitrary parameter of a VST

An old version is already available but I plan to replace that with my latest version in the near future as soon as I'm done with the latest tweaks.

More at http://deskew.com/blog/replace-apple-mainstage-with-a-custom-maxmsp-implementati on-part-4.html