Understanding CPU load

Not completely sure about this, but crackles on one insert-overloaded track may be eliminated by transferring some of its inserts to other Groups and FX tracks, then sending the originally overloaded track to those Groups and FX tracks.

Wow, just a new 24bit/48kHz project with a single stereo guitar track with a Neural DSP plugin, the Cubase CPU it’s gone. How is it possible?

Following @xerogh suggestion, the channel latency is 0.1 ms (4 samples). wtf?

Are you getting crackles and drop outs?

If not, maybe best not to worry despite what the Cubase monitor shows, until/ unless you actually start hearing problems.

@alexis unfortunately that’s not the case, sometimes the sound stops. It makes no sense, my CPU is very powerful and we’re talking about a single plugin. I hope there’s a solution :disappointed:

Can you provide more details about this “powerful computer”?

If you need low latency for real time guitar sims and drum MIDI recording it is highly dependent on your audio/MIDI interface and drivers and how they interact with your mysteriously yet undefined “powerful computer”.

In this case, more speed is generally favorable over more cores (from what I understand).

What is your currently used ASIO audio device?

Also, there is this you can try …

Resplendence Software - LatencyMon: suitability checker for real-time audio and other tasks

Im fairly confused, and I am stunned that this issue is obfuscated in this way.

Ive been following this IDENTICAL issue, and have made dozens of posts about it, since cubase 10.5, and I have confirmation of the issue (ASIO internal overload from hyperthreading/multi-media multithreading) from steinberg in writing via email.

I keep seeing this over and over, 1000s of times since last year alone, and each time the response from steinberg is “have you checked the manual/do you know how to admin a computer/have you bought new hardware/have you replaced all your hardware”. Nearly 100% of the time.

And each time, as I follow these threads and notice the much larger megathreads about the issue, I see the same pattern of customer service and it frankly enrages me.

To the OP:
Have you tried any BIOS changes, such as totally disabling all virtual cores, disabling all CPU p-states/power management, and disabling all dynamic clock frequency(speedstep, etc) features, then try this again?

I have seen the identical issue 1000s of times, Ive seen it on windows and mac over the years, and I have seen it on $10,000 workstations as well as $2000 workstations. I have a $1200 laptop that does not experience the issue because of a specific BIOS feature set that MSI built into that unit.

Perhaps the $1200 MSI laptop from 2017 is the “powerful machine” that Scab_Pickens means.

Im not sure if there is a 12.0.30 megathread on this, but here is the megathread for this issue for 12.0.20.

It might be helpful to review this thread. Hope you are able to work around this.

The problem is not average CPU load.
The problem is what your system interrupt latency is.
A vast variety of devices in a system have been known to lengthen the interrupt latency by milliseconds at a time, and if this happens, no sound card is going to be able to play back without crackles.
Some graphics cards do this.
Some network cards do this.
Almost all BIOS / RGB / system-monitor / SMBus tools do this.

They do this, because it’s either the easiest way to write the driver, OR because it leads to higher benchmark numbers in whatever benchmark that particular vendor is optimizing for, even if the overall enjoyment of the system goes down.

So, rip out anything you don’t absolutely need. Uninstall all drivers and add-ons.
Update BIOSes, both for your motherboard, and for each device (including graphics cards!)
See if it gets better.

Run a system latency (interrupt latency) monitor.
Sometimes, you have to break out the windows driver developer kit software to debug the system, and if you’re not a systems software programmer, well, you’re kinda SoL then…

This may be helpful, @mmoreo :

Unofficial Windows 10 Audio Workstation build and tweak guide - Part 1 - Windows MIDI and Music dev (this is one part of three, generally applicable to W11 per the author)

Also the latency mon link by scab-Pickens some posts up may point you to some processes you can turn off. Would run it for a long time, maybe overnight before checking results.

Also using latest versions of plugins, hardware drivers, check specific plugin forums to see if others have similar problems.

Finally, turning off things like WiFi, antivirus, anti-malware, windows “indexing”, cloud sync (like Microsoft OneDrive), back up programs, basically anything else you can think of that might divert the computer’s attention from real-time audio.

Having only the necessary tracks being record-enabled can help.

Finally, if you can run it till it crashes, sending a crash dump file (instructions on one of the other current threads) may point out the problem.

Sorry if you’ve done all this already, don’t mean to post things that may be obvious already to you!

:slight_smile:

@Scab_Pickens my setup is:

  • MSI CASE ATX MID-TOWER 100R, 7 SLOT HDD, SIDE TEMPERED GLASS, 3X120MM FAN FRONT, 1X120MM RGB FAN REAR
  • GAMDIAS MWE GOLD 750W - 80PLUS GOLD, ACTIVE PFC, 120MM FAN
  • MSI MB MPG Z490 GAMING PLUS ATX LGA1151 64GB DDR4 PERFORMANCE GAMING
  • INTEL CPU COFFEE LAKE I7-10700KF 8 CORE 3.70GHZ SOCKET LGA1151 12MB CACHE BOXED SE
  • COOLER MASTER MASTER LIQUID LITE 120, 120X38MM RADIATOR, 1
  • ADATA RAM GAMING XPG SPECTRIX D60G DDR4 3000MHZ CL16 4X8GB RGB FULL LED TITANIUM
  • ADATA SSD GAMING XPG GAMMIX S11 PRO 1TB M.2 2280 PCIE GEN3X4 3D NAND FLASH 2ND GEN NVME 1.3 R/W 3500/3000 MB/S BLACK/RED HEATSINK
  • SSD GAMING XPG SX950U 960GB 2,5 SATA3 TLC 3D NAND FLASH R/W 560/520 MB/S
  • MSI VGA GT 710 2GB DDR3 DL DVI HDMI VGA LP

Enough to monitor a single track with a single insert at low latency?

@Scheissberg thanks a lot, so I’m not alone. I’ll study some resources and try to understand how multi-core feature works under the hood. I see that with the multi-core feature activated a single CPU is used at 70% (the other 15 CPUs are idle) and then de-activating the feature this load disappears. I’m really confused, and the problem is still there in any case.

Thank you @jwatte I need to study a little more (the support just said that it’s normal and to use a higher buffer size, which makes the playing impossible for me, as I’ve just tested).
Thank you @alexis I’ll ready it carefully trying to find a solution, even if I spent 2k€ to delegate this stuff to a shop and they aren’t able to help me (this is a “Cubase computer”, what a bad experience…).

I’ll update you all with more information as soon as possible and hope there’s a solution to this very, Very, VERY frustrating thing. If interesting for you too, I paste here a link from the Steinberg support answer: CPU Performance vs. Real-Time Performance in Digital Audio Workstations (DAW) - YouTube

:crossed_fingers:

1 Like

I don’t know what you mean by “higher buffer size.” A well put together Windows system should be able to use 128 frame buffers without too much trouble, assuming there are no “power supply monitor” or “RGB LED controller” or “system temperature optimizer” or “custom fan curve” or other such bloatware on the system.

This is assuming that you have a high-quality USB sound interface – if you’re using the built-in headphone/speaker jack, then chances are that the problem is in those drivers.
So, if you’re using the built-in sound card in the motherboard, and you have already removed all “fancy” “gamer” “featureware,” then the next step is getting a high-quality USB sound interface with good ASIO drivers. Typically you should be able to get one for less than 200 dollars/Euro, and which particular brand you choose, depends on personal preference. Maybe something like a Scarlett Solo for guitar/microphone/headphones, or Scarlett 2i2 for stereo in/out ?

1 Like

lol I have all of this and I hate it. :laughing:

Hi there,

I also have a very powerful computer (18 cores, 36 Threads and 64 Giga RAM).

I run a reasonably heavy project and pops and crackles started happening near the end of the project for no reason at all.

I love Cubase and Steinberg, but Cubase is full of bugs, period.

I am also on Cubase 11 pro.

I don’ t know what made me try this, but here is what happened.

I went to - Task Manager->End Task for Cubase ->Restart Cubase->Safe Mode console came up → I opted for ‘Disable Program preferences’ aaaaand , all my problems just went away.

No pops, no crackles, Audio performance went down to near zero…

I initiated the Safe Mode again, and then I opted for ‘Delete Program Preferences’

Cubase runs like a dream, what happened? No idea…But it worked…

2 Likes

It sounds like you have a nice system. However, your ASIO device and it’s drivers have a big impact on performance. So, again, what audio device are you using?

What were the results when you ran LatencyMon?

Is it only with this specific plugin that you experience problems? What about recording a “raw” signal without any plugins involved? Any problems?

Have you tried in “safe start mode”?

I know it’s not something that people want to do, but have you tried to disable hyperthreading and see if there is any difference?

Just throwing the common things out there to see if there is some kind of rhyme or reason to your problem.

I know this must be frustrating, but there must be a cause for this.

Then get rid of it, dork! :stuck_out_tongue_winking_eye:

:smile:

Thanks, @somecomposer, I’ll try soon. I’m testing my old Scarlett 2i2 with its specific drivers to bench it before spending 2k (!!) on a thunderbolt interface. I need good performance so will spend some money to be ok for some years. But I want to be sure to understand how the system works before. I’ll install a latency tool and will give you some feedback soon. Thank you guys

1 Like

Ouch…Cubase runs fine for me. However, after long use of a Cubase file, it all goes to hell…

At least at this point, I can work.

It’s a Cubase issue.

before spending 2k (!!) on a thunderbolt interface

As long as it’s not an Analog Devices interface – those are not a great match for modern DAWs in my experience.

I ended up spending almost 1k on a RME Babyface, and it’s very good – almost tailor made for my particular need. (It replaced an original, version 1, UR-22) So I can recommend RME, at least if you’re on Windows.

Curious - did you disable cpu p-states, speedstep, etc? The stuff that dynamically clocks the cpu & controls power? I would suggest disabling all of it at first, testing cubase, then adding back individual BIOS-level CPU management features. I cant tell if you disabled more than hyperthreading.

On my system and on dozens of other similar platforms that I have observed, disabling virtual cores alone no longer stops the issue (it did in C10, 10.5, 11, but not 12) and I’ve had to fully disable all virtual cores, all CPU management features. I almost never see the ASIO overload or actually hear a pop/click anymore. I have not changed my hardware or software environment since cubase 10.5 outside critical windows updates, and the issue changed its behavior on 12, leading me to disable ALL CPU management.

1 Like

Ive done testing on this issue for 3-4 years now, and I find that I can do ridiculous things with my system without seeing any impact on cubase normally.

Ive purposefully decompressed or compressed 5-20GB archives, run windows updates, run java-based minecraft servers hosting 4-5 clients, maxed out my NIC across the LAN, etc, and cubase is never impacted.

But turning on virtual CPU cores (hyperthreading), enabling CPU power and clock management? That brings the ASIO internal overload and the pops/clicks every time. Ive installed and patched huge video games on this system, for fun, just to challenge this issue, and oddly the ONLY thing that triggers it is CPU management. Years ago I followed my own systems admin philosophy and kept this machine clear of everything but cubase and audio apps just to make sure, but over time I realized the ASIO overload issue has very little to do with anything else on my systems.

Edit: apologies, I could have combined those 2 comments. : /

3 Likes

I studied this topic a bit, and the problem seems to be related to Windows itself (ACPI.sys process in particular, if I read the logs correctly). Even talking with experts, I’m not able to push my DPC latency under 1ms in a totally idle scenario (I used Dpc Latency Checker closing every process possible).
Here are some screenshots:




I also tried to disable hyperthreading and all the stuff: it’s a complex topic and this is the main reason why professionals typically use MacOS instead of Windows. I decided to delegate this stuff being sure that a specific assembled DAW PC would have great performance. I decided to stay with Windows in order to build a powerful computer at a lower cost, but no one told me about this kind of problem.

I also tried Process Lasso (what a great idea to configure CPU affinity for specific processes!) but it didn’t work and the performances are the same. I believe that being the system itself there’s no solution.

I’ll try to focus on creativity but a bit sad about these results…