CPU spikes in Synthogy Ivory

[ochikubo]

I have spent some time trying to isolate where some of the peak loading is happening in my setup when using MainStage. The one AU plugin that I have seem to consistently have issues is with Ivory (AU 2.1.3b138 x64 Coa).

 

With a program that has Arturia MiniV, Silververb, EXS24, Channel EQ, Chorus, and Ivory. Watching the CPU Load History (while running in performance mode), it will idle around 4 to 6 percent, with an occasional spike around 15%. Every couple of minutes it will have a VERY big CPU spike over 1000%. The CPU spike can be attributed fully to Ivory.

 

Not clear on how to eliminate this. I have tried reinstalling Ivory, but it continues to happen.

 

 

Mainstage 3.2.3

Using some modest audio settings (512 I/O buffer size and 44.1 kHz sampling)

MacBook Pro (13-inch, mid-2012)

2.9 GHz Intel Core i7

16 GB RAM

OSX 10.1..13

WiFi is OFF

Samsung 250 Gbyte SSD (with about 50 Gbytes free)

MOTU Ultralite Mk3 Hybrid (using FW)

MOTU Microlite (for MIDI)

Using built in Logic AU plug-ins, u-he Diva (very heavy CPU load), Synthogy Ivory (very "spiky" CPU load), Arturia MiniMoog (medium heavy CPU load).

[DolmensDude]

I've observed the same kind of issue, not a CPU spike but a kind of "lag"... not latency, it was as if the Ivory plug in caused the entire sound stream to "rubberband" and "catch up" to what I was playing. Haven't had this for a couple of years though, but I've seen it.

 

this is just a hunch, but EXS24 and Synthogy both are streaming-based sample players. Each of these companies write their own code for how to stream samples in best-way possible - they may "get in the way of each other"...

 

Does Ivory perform okay using the standalone player w/o any spiking?

[ochikubo]

The nice thing with mainstage is the CPU Load history gives you visibility on the processing taken by each individual plug in. So I can see Ivory relative to other instruments and effects. And that is where I see Ivory spiking up and other instruments not.

 

I have made a very simple concert with only Ivory in it and it doesn't have the problem - so definitely interaction. I am starting to go through and exporting different combinations of sets/patches and see if it possible to isolate the offending interaction.

 

One thing I do notice is that the native instruments are very stable and efficient in the latest version 3.2.3 of mainstage. The latency issue also seems to have gone.

 

Just nothing in the native instruments has the quality of Ivory.

[DolmensDude]

agree on NI's pianos. An alternative to Ivory could be Ravenscroft 275 from vilabs... and in my rig "it plays nice" with the other guys (such as NI). The troublesome patches I had were EXS, Kontakt *and* Ivory. EXS, Kontakt, and the Ravenscroft were okay together.

 

Just to throw a loop in your thinking... I've recently tossed em all out, and went with Pianoteq and their new "Model B" - set up my own ultra close mic setup in that, and knocked down memory requirements in large concerts by a nice amount. It's just another route, one which I would not have done until this very recent update.

 

That CPU spike can be better described as a kind of "panic" - it needs memory, and it needs to get stuff loaded in and out of that memory (streaming) in a hurry. Not sure what the memory load is in your concert at the moment you're seeing spikes, but it can *also* be OS X converting compressed, not-recently-accessed memory getting uncompressed so that it can be used, and if that's happening, something else is getting compressed... that would spike the CPUs as well, so double click on the memory indicator and see what the total in use is at that moment in time.

 

You also write that you have a Samsung SSD, which is 3rd party to the Mac. I've heard these slow down, considerably as TRIM isn't natively supported on 3rd party SSDs... you might want to check if your SSD is still as fast as it should be.

 

Perhaps you know all of these already, but if not, just suggesting again.

[ochikubo]

The macbook pro that I am using has 16 gig of memory in it, so memory is not tight. Looking at both the MainStage memory monitor or the activity monitor utility it sits around 5.5 gig of memory used (stays constant). But I do notice that the spikes occur after some time, and so that might point to a memory leak or something else going on.

 

Read a lot online the Samsung SSD's and TRIM. Main issue I read is regarding benefit of TRIM on the write side performance on drives at a high level of utilization. The drive sits at about 75% utilization so their is still a fair bit of space available. And looking at activity monitor, for the concert I have had running (with all the patches in it), it is 20 Mbytes of writes versus about 4 Gbytes of reads.

 

Did go ahead and do a backup and then enabled TRIM since it is supported natively on El Capitan, waiting to see if the problem comes up.

 

Also, read some online regarding kernel_task spiking for El Capitan CPU requirements, but I am not seeing any of that here.

 

On a side note, did take a listen to the Ravenscroft and Pianoteq pianos online demos. They sound good, First time checking out Ravenscroft and it has been quite a while since I last took a look at Pianoteq, and that sounds a lot better than what I remember.

[Mark K]

Have you tried 3.2.4 yet to see if the spiking is still an issue? Apple confirmed in the 3.2.4 release notes that there was an issue with external instruments in 3.2.3.

[ochikubo]

Just completed the backup and install.

 

Still getting the smaller CPU spikes when not playing anything, where CPU usage goes from about 3% and jumps to 25%. But I haven't seen the really big 1000% CPU spikes so far. Going to let it run for a while and see if they come up. Not too clear from the release notes what the other stability issues they fixed might be.