Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Forum Jump  
VEP 7 CPU performance meter
Last post Tue, Jun 30 2020 by Dewdman42, 12 replies.
Options
Go to last post
Posted on Fri, Jun 26 2020 19:59
by GunnarC
Joined on Thu, Aug 08 2019, Posts 7

According to the manual, the numbers in the instance header in VEP7 server is "real time CPU performance". From what I can see, the number in the instance header is always about 5 times that given by Windows Task Manager for the VEP7 server process.

On my system, a completely empty instance with no plugins loaded uses about 8% of the CPU according to VEP7, and about 1,5% according to the Task Manager. With one (1) idle plugin (Kontakt with no instrument loaded) it uses about 10% according to VEP7 and about 2% according to the Task Manager.

Why is "real time CPU performance" 5 times higher in VEP7 than in Task manager?

Also - in my network, a completely empty VEP7 instance with a connection from a Cubase track (not playing) uses about 3 Mbps network bandwidth, is this normal? For doing nothing?

Posted on Sat, Jun 27 2020 06:31
by Paul
Joined on Sat, Aug 03 2002, Vienna, Posts 12232

Hi Gunnar, 

Our performance meter is always in real-time, while the Activity Monitor takes an average value. 

The network bandwidth also depends on your assigned outputs/inputs in the VE Pro Preferences. 

Best,
Paul

Paul Kopf
Product Manager - Vienna Symphonic Library
Posted on Sat, Jun 27 2020 07:22
by GunnarC
Joined on Thu, Aug 08 2019, Posts 7

Originally Posted by: Paul Go to Quoted Post

Performance meter is always in real-time, while the Activity Monitor takes an average value.

I would say that something is wrong, either the design or the implementation. If your meter constantly shows 10% and the Task Manager constantly shows 2%, there can't be a difference just based on real time vs average value. And even if all numbers actually are correct - in what way does your 10% value give me any usable information for a process that have an average load of 2% on my system?

I have measured the CPU load for the VEP process over time and have attached the graph. During this period (1.5 minute) my impression from the performance meter in VEP was that there were about 10% constant load. According to the performance monitor, the CPU load from the VEP process averaged around 4% (the scale is x10 in the graph) and only once managed to get above 10%.

The thing I'm getting at here is that even if the numbers technically would be correct (which I honestly doubt), the implementation gives me the wrong impression. As I said - during this period when I measured the CPU load using Performance Monitor, my impression from the CPU meter in VEP was that there were a pretty stable load of about 10%.

Maybe having a CPU meter going all over the place in real time isn't the best way to give the user a usable estimation of the CPU load?

GunnarC attached the following image(s):
VEP7U0020CPU.png (72kb) downloaded 5 time(s).

You cannot view/download attachments. Try to login or register.
Posted on Sat, Jun 27 2020 19:46
by Dewdman42
Joined on Tue, Feb 27 2018, Park City, Utah, Posts 590

I question how you are using Task Manager to measure CPU utilization.  on my mac, I can run VePro with a small test, Vepro shows 6% and my Cpu utilization meter for the whole computer also shows 6%.  If you are trying to measure just one particular process with Task Manager you might not be getting the whole story.  I'm not that familiar with the Windows version of Vepro but its possible that multiple processes are involved that need to be measured by Task Manager.  But I think if you just monitor the overall CPU utilization of the machine it should be pretty close to what VePro is saying also.

Remember also that cpu utilization measurements are always an average, but different tools may use a different time period to measure that average.  A shorter time period would result in a curvier line.  

5,1 MacPro 12core X 3.46ghz, 128gb ram, RX580 video, OSX Mojave, VSL (almost everything), LogicPro, Cubase10, StudioOne, DP9, Reaper, Dorico, Finale, MuseScore, Notion6, EW Hollywood Orch, Kirk Hunter, GPO, much more..
Posted on Sat, Jun 27 2020 21:28
by GunnarC
Joined on Thu, Aug 08 2019, Posts 7

So on a Mac it obviously works. On Windows it don't. When it comes to reports on CPU utilization I trust the Operating System more than VEP, the OS is after all in control of the whole situation.

And it isn't about measuring the correct processes. While VEP is constantly reporting around 10-11% for a single instance, the complete load on my CPU from all running processes including the OS itself i pretty stable at 7% (and I have a lot of things running in that machine...)

Believe me - on Windows VEP is reporting an incorrect CPU load. No matter what other tools I use for CPU performance reports (built in or 3rd party) they show way lower CPU load than what VEP does.

Posted on Sat, Jun 27 2020 21:31
by Dewdman42
Joined on Tue, Feb 27 2018, Park City, Utah, Posts 590

As I said, in task manager you are trying to measure the VePro process itself and there may be other hidden CPU activity related to VePro at work in the kernel or other processes which you are not measuring with Task Manager in that case.

Change your Task Manager configuration to measure the entire overall system CPU utilization, and I suspect that it will be similar to what you see in the Vepro cpu report...as long as you don't have anything else big running on the machine at the time.

5,1 MacPro 12core X 3.46ghz, 128gb ram, RX580 video, OSX Mojave, VSL (almost everything), LogicPro, Cubase10, StudioOne, DP9, Reaper, Dorico, Finale, MuseScore, Notion6, EW Hollywood Orch, Kirk Hunter, GPO, much more..
Posted on Sat, Jun 27 2020 22:14
by GunnarC
Joined on Thu, Aug 08 2019, Posts 7

Originally Posted by: Dewdman42 Go to Quoted Post

Change your Task Manager configuration to measure the entire overall system CPU utilization...

Windows Task Manager is built to measure the entire overall system CPU utilization, it can't be configured. The entire overall system CPU utilization is quite a bit lower than what VEP reports to use all by its own. But I give up. If you do not think it is a problem that your program shows the wrong CPU load, then I should not push the question anymore.

Posted on Sat, Jun 27 2020 22:21
by Dewdman42
Joined on Tue, Feb 27 2018, Park City, Utah, Posts 590

The image you attached above is specifying the VePro process.

No I don't particularly think its a problem, I'm sure it can be explained and probably one of the VSL engineers at some point will explain why the numbers may look different. 

5,1 MacPro 12core X 3.46ghz, 128gb ram, RX580 video, OSX Mojave, VSL (almost everything), LogicPro, Cubase10, StudioOne, DP9, Reaper, Dorico, Finale, MuseScore, Notion6, EW Hollywood Orch, Kirk Hunter, GPO, much more..
Posted on Sun, Jun 28 2020 06:47
by GunnarC
Joined on Thu, Aug 08 2019, Posts 7

Originally Posted by: Dewdman42 Go to Quoted Post

...No I don't particularly think its a problem...

I appreciate that clarification.

You list this as a new feature and as a selling point for VEP 7.  In the promotional video for the feature (introduced with "control is an important factor"), the combined "real time CPU load" from the instances averages around 150% and sometimes goes above 200%. For me, it is extremely obvious that this is not correct.

I have been using the trial version of VEP 7 for some time and was about to purchase a couple of licenses now in the June promotional prices. I'm glad that I in time got to realize that support of this product is more interested in finding excuses than finding the cause of the problem.

Posted on Sun, Jun 28 2020 06:50
by Dewdman42
Joined on Tue, Feb 27 2018, Park City, Utah, Posts 590

FYI, I don't work for VSL.  I'm just another user like you.  Maybe give them a little more time to respond to your question.  VePro is a great product, not sure why the big concern about the CPU meter.

5,1 MacPro 12core X 3.46ghz, 128gb ram, RX580 video, OSX Mojave, VSL (almost everything), LogicPro, Cubase10, StudioOne, DP9, Reaper, Dorico, Finale, MuseScore, Notion6, EW Hollywood Orch, Kirk Hunter, GPO, much more..
Posted on Tue, Jun 30 2020 06:47
by MS
Joined on Wed, Feb 19 2003, Vienna, Austria, Posts 1759

1. The VEPro CPU meter measures real-time thread use only. If you are running at a buiffer size of 256 samples, and samplerate of 44100 Hz, every buffer has to be processed within 5.8ms. If VEPro manages to process that buffer within 0.58ms, the real-time CPU usage would be 10%.

2. Plug-ins may spawn threads that process things in the background. This will affect the process/system CPU load, but may not show up in the real-time CPU meter. There is no way for VEPro to know about these background threads.

--
Martin Saleteg
Software Developer
Vienna Symphonic Library GmbH
Posted on Tue, Jun 30 2020 06:56
by Dewdman42
Joined on Tue, Feb 27 2018, Park City, Utah, Posts 590

Originally Posted by: MS Go to Quoted Post

1. The VEPro CPU meter measures real-time thread use only. If you are running at a buiffer size of 256 samples, and samplerate of 44100 Hz, every buffer has to be processed within 5.8ms. If VEPro manages to process that buffer within 0.58ms, the real-time CPU usage would be 10%.

Martin thanks for that explanation!  In many ways that is much better metric than the actual CPU utilization.  Because it will show on a scale of 0-100% how close you are to drop outs...regardless of what the CPU utilization is actually doing on the system.  

Do I understand that right?

5,1 MacPro 12core X 3.46ghz, 128gb ram, RX580 video, OSX Mojave, VSL (almost everything), LogicPro, Cubase10, StudioOne, DP9, Reaper, Dorico, Finale, MuseScore, Notion6, EW Hollywood Orch, Kirk Hunter, GPO, much more..
You cannot post new threads in this forum.
You cannot reply to threads in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.