VEPro7 - Slave upgrade causes stutter and dropouts

[DonRex]

Hello all,

Some weeks ago, I decided to upgrade my Windows 10 slave with a new CPU, GPU and motherboard. I was lucky enough to get my hands on an RTX 3080 (ASUS TUF). Since I mostly write game-related music and my old TitanX-card died on me, this was a really cool substitute. While I was waiting for my GPU to arrive, I decided to upgrade the machine too, switched from an Core i7 6800K on an Asus X99 motherboard with 64GB of RAM to a Core i9 10850K on an ASUS z490-f motherboard. And that's when the trouble started.

With Cubase 10.5 on the master, a 2012 Mac Pro, which was previously rock solid and reliable, I'm now getting stutters, dropouts and hanging notes and I can't seem to find the source for this. It's safe to say that I've tried many things to fix the issue now: I've changed the power plan on the windows machine, turned off turbo-boost, hyperthreading and all that stuff, upgraded the BIOS on both the motherboard and the GPU, updated all drivers, bought a new Ethernet controller to eliminate issues with the onboard chip.

LatencyMon does occassionally spike on the NVIDIA-driver or NDIS.sys (after about an hour of inactivity), but in general it tells me that my Windows-machine is more than capable of running real-time audio.

On the master (the Mac) I've played around with turning multiprocessing and ASIO-guard on/off, I've changed the VEPro buffers, but nothing has helped so far. Even without any effects and plugins, songs that would play back without a hitch before, stutter and drop out. Export/rendering works as expected too.

Now, the "funny" thing is that these dropouts only seem to occur when I play back multiple tracks. When I solo or simply play around on the keyboard using a single VST, it will play back fine even at low latencies. When the stuttering occurs, the VEPro performance monitor shows a peak of around 15% and the task manager shows CPU usage around the same.

I don't know where to go from here, so I'm hoping that someone, either from the community or from VSL, can help me "move on" before I literally "move on". 😊

I've posted a short video that shows the problem here:

. Sorry for only showing Audio-tracks in the video. The full template has 410 tracks where around 400 of them are VST-tracks (and their outputs) connected via VEPro.

Thanks for any help you can give me,

Thomas Regin.

Specs Mac Pro:

Model Name:Mac Pro
Model Identifier:MacPro5,1
Processor Name:6-Core Intel Xeon
Processor Speed:3,33 GHz
Number of Processors:1
Total Number of Cores:6
L2 Cache (per Core):256 KB
L3 Cache:12 MB
Memory:16 GB
Boot ROM Version:MP51.0084.B00
SMC Version (system):1.39f11
SMC Version (processor tray):1.39f11

Soundcard: RME Fireface 800

Spec Windows 10 slave:

Core i9-10850K
ASUS z490-f gaming motherboard
64GB RAM
ASUS RTX 3080 TUF GAMING OC
TP-link TG3468 Ethernet controller

[Ben]

Hi Thomas,

Right before the drop-out you can see that the ASIO load meter peaks in your DAW, so to me it seems to be a DAW (machine/OS/plugin) issue. You could try to disable (not only bypassing) the plugins hosted in Cubase.

Best, Ben

[DonRex]

Hi Ben,

Thank you very much for helping me! I must confess that I don't see the peak in the ASIO-meter. I've been monitoring it several times because I've previously experienced issues with "bad" plugins, but this time it's been pretty steady at around 15-20% or so.

I did, however, try to disable (delete) every, single plugin in my project, but unfortunately the issue remains.

Where can I go from here? Is there any way to figure out where the problem stems from?

Thomas.

[DonRex]

Small update..

I've started to remove hardware piece by piece from the slave to eliminate potential problems. I started with the RAM and the GPU, but neither had any effect. Still stuttering and dropouts.

However, while monitoring the CPU usage in VEPro, I noticed something a bit weird. Every time the music stuttered, the CPU usage went to 0%. I couldn't catch it in a screenshot, but this is the monitor I'm refering to (in idle mode):

 

So, the big question: Does this new information help in any way to establish whether it's coming from the slave or the master?

Thomas.

[DonRex]

Lo and behold, I think there's a chance that I've finally figured out what was wrong! Further testing is needed to say for sure, but so far it's looking good!

I downloaded a program called "PowerSettingsExplorer", which apparently gives you access to hidden power settings in Windows and changed two properties according to this video here: 

Basically it is to uncheck "Processor idle promote threshold" and "Processor idle demote threshold", go into your power plan in Windows and set the two new values that magically appeared there to 100%.

This change resulted in DPC latency checker going into the green area instead of just being yellow and red. 

Average DPC latency before was 1500+microseconds according to DPC latency checker, now it's below 1000.

Cubase, however, was still causing massive dropouts, which I found strange, since my slave was now pretty responsive. I read somewhere that people had turned the ASIO latency down from 1024 to 512 buffers which actually made their systems perform better. In my case 512 buffers made no difference, but when I went all the way to 256 the system suddenly began to play back smoothly!

All that remains now is to see how stable this system will be over time.

Hope this can help someone in the same situtation.

Thomas.