x99 i7-5930k & sonar platinum thread balance

just got an x99 i7-5930k 3.5 ghz, gigabyte ga-x99-soc champion board, 32gb corsair ddr4 2666 (running 2133), samsung ssd for os & audio, win7 x64, sonar x64, all x64 plugins. I'm running fixed oc @ 3.8ghz, all cores active & running, no turbo boost, no eist, no c settings, no ram xml active in bios.

I have mmc in sonar set to 2. but I seem to get the 1st thread noticeably more loaded than the other 11, and none of them are close to maxing out. yet I get audio engine drops with a relatively sane amount of tracks (<20), busses (10 or so), and plugins (20-30), and no vsti. buffer read/write @256, midi play buffer @ 2048.

granted, I'm using a few heavier plugins... like acustica nebula, but it is their local server version, which is supposed to be heavily optimized, running their core6 version, vstgui4, sse optimized.

it just seems like this pc should be able to load up the sonar cores, and running at least half core load for each thread with ease. but the audio engine is dropping off adding any more plugins, and with what looks like relatively little loading of the cores.

any ideas?

Start removing plug-ins till you find the one(s) that are chewing up the first core.

Have you taken a look at this?
http://forum.cakewalk.com...this-Out-m3103263.aspx

So did you verify ThreadSchedulingModel = 2 in AUD.ini on this machine? Usually SONAR config will get it right but it is worth verifying.

I was having a similar issue and I limited the thread count to 2 less than the total threads available with my CPU (another .ini variable) in which my theory was to  leave 2 threads for the OS and other system overhead.  I have 12 threads with a 6 core processor but only use 10 threads for SONAR.
 
Don't know if it will help but it is worth a try.  Sorry I don't remember the exact variable name but it is quite specific to what it will do.
 
Jim

Turn off Intel Speedstep. Also try disabling any "turbo" performance features in BIOS.

If one or two cores are getting massively loaded and not much is happening on the others try deleting or renaming Sonar's aud.ini file and let it create a new one.
 
I found Splat when first installed (X2 and X3 were already installed) was thrashing two cores to the point of dropouts while completely ignoring the others. Letting Sonar create a new aud.ini from scratch solved the problem.

Well a chain of plugins is by definition serially processed since each plugin needs its predecessor's output.
What would you process in parallel? :)

I'm running an i7 4770k and regularly load 12 to 16 instances EW Play with Hollywood orchestra along with a couple of Kontakt multiple output instances.

My CPU usage is usually pretty low but distributed more or less evenly. Processor 1 always has a bit more of a load on it. Play works out of the box but Kontakt needs to be set to support multiple threads . Perhaps the vstis are part of the issue.

Noel Borthwick [Cakewalk]
Its fairly common for one (or more) thread to get loaded more than the others. Multiprocessor load balancing is not perfect since plugin loads vary. Also if you have 10 plugins on one track/bus they are all processed on a single thread. If your project had all other tracks with a single plugin for example, the thread with 10 plugins would consume more.
The same would apply depending on how expensive the plugins are.
In short its unrealistic to expect fully balanced cores in a real world situation. I hope to get some time to revisit this area at some point since there are some ideas we have to improve this...

That was very informative thanks. Plugins like Maschine and Kontakt apparently have multicore support? Is it best to set them to use a single core at present?

Okay. So after a month or two of problems, I've found the cause. Noel hit on it in his previous post, so I'm sure he's aware of the particular problem. It's Nebula Core6 v1.3.903 attempting to take care of its own multi-threading within the host. Apparently it doesn't jive with Sonar's implementation of multi-threading control. I finally did a complete new install on my new PC of Nebula 1.3.846, and it works great! Tons of Nebula plugins able to run in a project with ease. But now Acustica have decided to control thei own multi-threading for Nebula within the host. And apparently, they think Sonar's multi-threading is not implemented exactly right. I really have no side, except to wish they played well together so I don't have to switch to Reaper to use new Nebula upgraded plugs from now on. And although I definitely don't know the technical workings going on between the DAW and plugin, it does sound much more logical to me that the host would control how it implements plugin multi-threading... not the plug-ins implementing their own multi-threading within the host. Kinda sounds backwards.
 
Anyway... I just wanted to post what I posted over at the Nebula forum, and the response I got from Acustica concerning Nebula Core6 and Sonar. And hopefully get Cakewalk's side of it here... so I know how to proceed. As in... is there any chance Acustica and Cakewalk are gonna try to work this out? I've been using Sonar since I left Pro Tools III back in 2005. For what I need it... I know it like the back of my hand. I'd really hate to have to switch to a new DAW and different work flow. So I'm considering... I guess based on the replies and info from each company... and how much I want to use Nebula in the future. I REALLY like Nebula. I REALLY like Sonar. I wish you guys would work it out.
 
So... here's the correspondence and reply from the Nebula forum...
 

javahut wrote:God... I don't know where to start.

First, I've spent a month and a half trying to solve a new problem with my PC, and in the process... rebuilding my audio PC to a 5930k from a 980x, new 32GB DDR4 RAM (from old 24GB DDR3), new SSD Samsung Pro 1TB for audio and samples (OS is and has been SSD Samsung 840 Pro 500GB), reinstalling everything from scratch. Uninstalling recently updated plugins... trying to figure out what's causing the problem. Thinking it must be a "bloated" Sonar. Purchased Reaper and started trying to learn it. Realized there's too much in Sonar's workflow I can't do without enough to spend time learning Reaper. Back to trying to figure out the problem with Sonar. Purchased server version of Nebula. All to no avail... as I continued to have the same severe audio engine drop out problem that's been plaguing me for the last month and a half continue.

Then on a whim... I uninstalled the 1.3.903 server completely... reinstalled the Pro version of 1.3.846, making sure to completely rebuild plugins, xmls, and VST registry for Sonar. And... voila! Everything back to normal.

I don't know what you guys did to 1.3.903. All I know is... I can run a metric $h|T ton of Nebula plugins in a single project without problems in 1.3.846. I can't play through a project once with less than a quarter of that amount of Nebula plugins in 1.3.903 on the server version.

So anyway... relieved that all's working again after a month and a half of problems, and several thousand $$$$ later. Nice to have a newer "upgraded" PC (though I lost my PoweCore PCI in the process).

I just hope the guys at Acustica can figure out what it is exactly that's different about 1.3.903 from 1.3.846, and can fix it. Or figure out a way to explain in English what the difference is and what we need to do to make the newer versions of Nebula work as well.

For everyone that says 1.3.903 works better? Either you have something entirely different with your system than several of us here do... or you don't really use Nebula heavily. Because there is without a doubt a HUGE difference between the two versions... and I'm using a LOT of Nebula in my projects. And this makes 2 systems I've used 1.3.903 on now that both exhibited the exact same problem with the update (one wasn't an update, but a completely new install on a completely new, modern X99 system).
Can't wait to see if I see an even bigger increase with the 1.3.846 Server version (now that I own it) over the Pro version.

But I'm really kinda worried that Acustica doesn't recognize that there's a problem. And what's different in the newer version that might be causing it to perform so very much worse than 1.3.846, in some systems at least. Is there a setting in an XML or something for the new version? Something not assembled correctly? I don't know. But I'd hate to be stuck on 1.3.846 and now never be able to upgrade again. As much trouble as this has been, and as much time and money as this has cost me, I'm afraid I'll have to abandon Nebula after all this time using it, and go back to algo plugs for everything. Please, Acustica, try to solve this problem. There definitely is one. I really hope you don't ignore it. Maybe it'll work itself out in the next update (fingers crossed).

At least I'm back in the audio business for now, and out of the technology troubleshooting business... at least for a little while.

Acustica wrote:
Hello.
 
First at all, it is very important to be careful in measurement and conclusion about those measurement in order to not draw in wrong conclusions.
 
9xx has several code re-witten to ensure multithreat, SEE4 and OSX 10.10 compatibility. Nebula and Acqua Effect load emulation presets and vectors using multithreat, several audio host can not handle in a efficient way this approach, and CPU cores are not synchronized correctly (Core1 CPU peak for example). In i7 v3/4/5 9xx works better with similar DSPBuffer setting than your ASIO/AC setting from 256 to 2048 values, before 9xx values around x2/x4 from ASIO/AC setting was the best setting for i7 v2 to v5.

We already test several times Sonar Producer x3e in i7 v2 and v4 and we can confirm that CPU efficiency is lower compared with other audio hosts like Reaper v4/v5b, also stability is lower. We contacted Cakewalk and report this issue, but they did not replay us.
 
Please if you need more details reopen your support ticket, and sales department can refund you Nebula3 Server if you don't want/need CORE5 & CORE6 features.
 
Best regards, Enrique Silveti Acustica Audio technical support www.acustica-audio.com

Perhaps your cores are multi threaded (as are most multi core intel chips)... 2 threads per CPU, 6 cores, 12 threads.
 
Jim

javahut
Okay. So after a month or two of problems, I've found the cause. Noel hit on it in his previous post, so I'm sure he's aware of the particular problem. It's Nebula Core6 v1.3.903 attempting to take care of its own multi-threading within the host. Apparently it doesn't jive with Sonar's implementation of multi-threading control. 
Anyway... I just wanted to post what I posted over at the Nebula forum, and the response I got from Acustica concerning Nebula Core6 and Sonar. And hopefully get Cakewalk's side of it here... so I know how to proceed. As in... is there any chance Acustica and Cakewalk are gonna try to work this out? I've been using Sonar since I left Pro Tools III back in 2005. For what I need it... I know it like the back of my hand. I'd really hate to have to switch to a new DAW and different work flow. So I'm considering... I guess based on the replies and info from each company... and how much I want to use Nebula in the future. I REALLY like Nebula. I REALLY like Sonar. I wish you guys would work it out.
 

 
I haven't personally got any correspondence from them. I'll see if I can track down their call. Do you have a contact there? I don't see how this can be a DAW problem. We have been doing this since SONAR 3 and every DAW implements multiprocessing differently. It seems like the person on that thread is jumping to conclusions without fully understanding multiprocessing. When you do MP you need ONE scheduler. SONAR has already allocated threads for each core and is trying its best to load them up. If the plugin now tries to circumvent this by doing its own scheduling it can quickly turn into a mess. Aleksey (from Voxengo) and I discussed this at length some years ago and we wanted to come up with a MP spec for plugins but its difficult to get vendors to agree on anything so nothing came of it. That is the proper way to go - let the daw do the scheduling and the plugin is the client.

Just in case anyone is still interested in Nebula 1.3.903 Core 6 & Server working within Sonar... there's some updated information at the Acustica forum. And Acustica replied to my support ticket implying that they will be testing a setup using Sonar with Nebula Server installed locally and running with processor affinity set to run on all but the first few cores.
 
http://www.acustica-audio.com/phpBB3/viewtopic.php?f=11&t=28728&start=84