CPU Overload / Dropout detected with CPU at circa 5%

Is there a such thing for AMD Ryzen 9 processors?

If it’s about the virtual cores, mine are enabled on 3900XT CPU, and haven’t had issues so far. Load seems to get spread out pretty well. I don’t think AMD does it the same way as Intel. It should mostly be at the OS/Driver level (as opposed to the APP) anyway?

There is a video on CPU performance vs Real Time performance in YouTube.
Below is the link. Understand before pointing fingers at Cubase

Regards

2 Likes

Excellent video.

I’ve been building realtime a/v/dev production environments and custom machines for 30+ years, and I understand this particular issue QUITE well at this point, thanks to my own research and testing. I’ve worked around the issue, myself, on my particular platform, and I’ve shared this workaround with many others. I’ve also had Steinberg (de) confirm the issue to me in writing, which I still have in email form. I understand the diff between software internal processing vs total CPU resources, etc.

One can attempt to defray the blame from Steinberg to their heart’s content, but those who understand this issue arent demanding Steinberg fix what they cannot or demanding anything unreasonable. Before “pointing fingers” at Steinberg, I did a year of research and testing. You?

What we do demand is HONESTY and transparency about the issue + better documentation about it coming FROM Steinberg, not their forum participants or 3rd party websites.

QA’ing the platforms experiencing the issue, then documenting and publishing the findings, would be the right thing to do, but that hasnt been done and instead these forums have seen literally YEARS of thousands of identical posts to this one, about ASIO overloads, multiprocessing, multimedia multi-threading under windows, etc. On each system/platform the issue can be slightly different and different CPU/MB/multicore/power management/runtime environment changes can affect it. This is where Steinberg could provide the biggest assistance to those struggling with this – document the issue more thoroughly, and PUBLISH IT.

1 Like

I don’t expect you to take the time to answer this, as it’s a potentially MASSIVE TOPIC. I am curious about the cost involved in testing systems and publishing the results, and how that’d effect pricing for end users.

I agree it’d be nice to get an official list of equipment confirmed to work well, and a list of things ‘known’ to have ‘issues’. Still, that requires a whole new dept. and another set of lawyers on retainer if they’re going ‘official’ with the reports.

It would be nice to get access to deep research analysis on these matters straight from Steinberg, but I don’t see any other ‘DAW Company’ doing and ‘publishing’ this ‘research’. At least not for mass consumer ranges of products.

Steinberg isn’t in the business of selling system builds. Sure, they could set up a lab, source every possible combination of equipment they can get their hands on, spend months or even years compiling relative bench marks and doing publications, but then the price for the software shoots up to thousands of dollars per copy instead of being well under a grand. Much of the data would probably be several cycles behind the current crops of system components on the market.

Third parties could be ‘contracted’ by Steinberg to do these things, but that still runs into expenses and legal concerns that could have a significant impact on the price range of Steinberg products.

I’m looking at an amazing set of products that for less than half a grand, of which I can obtain a Demo and try it out before buying. I can download and within’ a few hours have tremendous production capability.

When I started tinkering with electronic music in the 80s, it would cost up to 4 grand just to get 4 to 16 different voices/instruments to work with! I could NOT bring it home and hook it up to different things and try it out. The demo in the store was it. I still needed another $300 worth of cables and junk to make it work. I still needed the $500 mixer, the $1,200 PC/Mac, another chunk invested in ‘software’ and all the ‘contracts’ that came with that stuff. I still needed ‘media’, and even had to change out ‘capstan motors’ and ‘clean heads’ on a regular basis.

It’s unfortunate that there are still risks when it comes to picking software and system components. Some companies have better ‘return policies’ than others if things don’t work out. Some have better ‘demo’ terms than others. The good news is that in 2022, almost every DAW on the market allows you to resell your keys, and almost all of them offer ‘cross-grade’ pricing that takes a considerable chunk of the ‘bite’ out of ‘some users’ needing to invest in multiple DAWs (and you still get to keep the keys to all your software).

I won’t argue that there is not, or should not be MORE and BETTER attention to these matters by software dev companies; however, what would be the best way to do it, and what kind of impact does it have on the bottom line pricing of end products?

For what it’s worth. If I were going to build an enterprise class studio, I’d be calling an engineer like yourself who would sit down and ask my needs and budget. I kind of doubt you’d just plunk the ‘latest and greatest’ models and versions of everything in the system. You might even strongly advise that no one installs any updates/upgrades to anything, PERIOD, until you’ve personally had a chance to check it out and run some tests.

It’s been years ago, but I once worked in a pretty modern TV station that had access to the most cutting edge tech on the planet (I.E. it was the first station to the planet to set up an HDTV transmitter and start a process of converting EVERYTHING in the building to digital formats, so engineers all over the planet were interested in what happened there). They also had a few in house engineers gunning for tech and software patents themselves. The fact is, they came across many things that worked for us, and not for other stations. Sometimes it was the weirdest issues that ruled out a given process choice (user errors, weather, temperature, static build up issues, consistency of power coming from the city grid…you name it, they ran into it, and it forced constant slates of system and policy/usage tweaking). They also came across stuff that worked for us, but was a big fat FAIL at other stations. Our transmitter and half the gear plugged into it also happened to be a prototype. By the time ‘other stations’ were installing transmitters based on that tech, designs had changed somewhat, and DIFFERENT issues popped up, all needing ‘different solutions’.

Not one single independent manufacturer of all the bits in that station could predict and give us ‘research and reports’ for everything. All they could do was give us known specs on the protocols they were doing their best to support. It works fine on their test bed. The scopes in their dev lab say it’s performing to specs. As for QC at manufacturers trying to make ‘copies’ of the tech and distribute it all…yes, they all had problems there sometimes, but that’s often NOT the software people’s fault! In fact, the software people are often called upon to ‘salvage’ bad production runs and find some way get enough of it working to perhaps change the model number and sell it as a lower end product.

Sigh…apologies as none of my rambling is meant to be ‘personal’ in any way. Brain over-load.

Bottom line is that the $500 price range for access to all this tech is dumbfoundingly AMAZING to me. It puts it in a price-range that almost anyone can afford, yet it’s not like the OS, or an office suite, that pretty much every system is going to need. It’s something of a niche product, with a LOT of competition out there.

What would it ‘cost’ we end users to support a big systems research team? Would it come anywhere close to the quality of service we could get if those who need it instead…hired Engineers who aren’t ‘connected or biased to any given ‘company’’ to work it all out? Where is the ‘line’ that spending time/money/resources doing this ‘research’ would push the price up to a point where ‘the younger and smaller DAWs, who obviously ARE NOT doing this kind of ‘system build’ research either, with lower prices’ take over significant chunks of your base markets?

1 Like

That was indeed a lot to respond to, but allow me to respond minimally.

I’ve worked with software and game devs for decades of my life, also architecting, tuning, and supporting systems used by professionals, including in live streaming shows. In all these decades, I’ve not encountered the philosophy “QA testing is too expensive”, even while working for small, cash strapped startups who were gambling on what they could cut costs on.

Maybe I’m wrong – but Steinberg / Yamaha have the budget for a QA team, and the ethical imperative to do so for a reliable, usable product. A QA team doesn’t even have to be a team of 800 senior engineers or even software devs if it comes down to it, and one QA tech can automate the testing of 100s of systems at the same time through virtualization and desktop automation, such as Sikuly. My point about all this is IM NOT steinberg’s QA engineer, and I would be a lot more amicable to working through this with them if they were simply truthful about what cubase can and CANNOT do at current.

Besides all that, I have to return to the meat of your comment. Are you under the misapprehension that QA teams or QA processes are a luxury, and that Steinberg would somehow reduce reliability and customer service options even more if they were “forced” to QA their $600 software before shipping it?

This isn’t the 80s, or the 90s anymore, and the technologies and expectations for those technologies here are not misaligned. Others always say it, and I’ll add my voice: if most other audio dsp/daw vendors on the market have been able to work around 3rd party problems with hardware and software over the last 20 years (this began with the core2duo in like 2002. Had to disable virtual cores), the Steinberg engineers should also be finding a way or at least documenting that the issue still exists and advising users pre-sales.

Instead, Steinberg (in public) says “we have ASIO guard, and recommend you turn it on. It should solve the ASIO overload issue for most users.” And admits in private email chains that - yes just turn off virtual cores for less headache.

2 Likes

I got my first ‘modern’ copy of Cubase Pro for a Windows based PC starting with version 7.5 (had been using older systems, and different ‘synced’ methods and tech to ‘record’ before). Documents on Steinberg Support like this were around then, and they still are.

Windows: How to set-up and optimize a Digital Audio Workstation – Steinberg Support

Tweaking the computer
Modern systems with a fully updated Windows should not need any of the default settings to be modified. However, if certain drivers are not fully optimized and you experience audio drop-outs, it is worth having a closer look. Note that the following settings are often not accessible, especially on laptop systems.

Disable Hyper-Threading (Intel)/Simultaneous Multi-Threading (AMD) in the UEFI BIOS if your CPU supports it and if your BIOS allows you to modify this setting.
Disable advanced power-saving and dynamic performance options for your CPU if applicable. This usually needs to be done in the UEFI BIOS of your computer and includes ‘Enhanced Intel SpeedStep (EIST)’, ‘AMD Cool 'n' Quiet’, ‘Intel Turbo Boost’, and ‘AMD Turbo CORE’.
Disable C-States in the UEFI BIOS if this is accessible. C-States allow your CPU to sleep when idle, which may interfere with real-time applications such as audio. This option is often called ‘Disable CPU Idle State for Power Saving’.
1 Like

PS

I too would like a list of ‘known problem’, and ‘known to run well’ hardware.

Always wondered why it doesn’t exist. Even if a third party got ‘sponsors’ and put it together…it’d be good to have!

1 Like

Likewise are there pages like this:

https://helpcenter.steinberg.de/hc/en-us/articles/115000535804-Windows-10-audio-dropouts-on-multi-core-CPU-setups

And this:
https://helpcenter.steinberg.de/hc/en-us/articles/206625630-Hyperthreading-Simultaneous-Multithreading-and-ASIO-Guard

Which contradict the older advice to DISABLE virtual cores. This is where Steinberg should do much, much better with documentation and updating their overall spec on hyperthreading on windows and Mac.

2 Likes

Oh man, I could rant for days with my ‘pet peves’ for every software title I own.

People laugh at me all the time for harping on 10 year old stuff, that in theory would require little more than a single coder taking half a day to add some fields to an XML file under the hood of the Steinberg engine, and adding a way for the user to manipulate that field. Or simply adding some things to the UI that grant ‘user access’ to abilities OBVIOUSLY already built into the ENGINE!

There are more than half a dozen things I can do outside spec by ‘opening an XML file and making a change’. Sometimes it’s a matter that can save me HOURS of work through the provided ‘editors’.

I’m starting to see similar things pop up with Dorico. I.E. No way to truly ‘black-list’ devices from being touched when launching. Properly implementing legato/portamento pedals from expression maps, etc. It seems like a very small and unimportant thing, so why bother? Tweaking the default sounds so they aren’t down right ‘offensive’ (makes the HALion engine look pretty bad, when the truth is, it’s a really nice/solid engine). Only that ‘one goober’ is complaining about this! Move on…

Adding a massively complex and brand new ‘feature set’ is more important than taking half a day tops to fix some of the most ‘basic’ of issues. The longer they WAIT to fix the ‘foundational’ things, the more they have to consider adding a lot of ‘legacy support’ stuff IF and WHEN they do get around to ‘fixing it’.

Drives me CRAZY sometimes. What I couldn’t do with a few simple additions to elements that the Logical Editor can access! Just add it to the UI! The engine can ALREADY DO IT!

Instead…sample tracks, new skin jobs for old plugins, blah blah.

Oh well. Steinberg isn’t alone in this. I have multiple DAWs, multiple Scoring Suites, and it’s the same old song and dance. It’s just as bad for all the products I’ve ever owned (sometimes even worse).

One step forward, 2 steps back.

I just have to keep reminding myself how much I do get for the relatively small investment. It wasn’t all that long ago that I couldn’t even dream of being able to afford to make sounds like this on my own! I sure as heck could not instantly share it with thousands of people at the click of a few buttons.

Did OP find the solution? I also have a 5900x, 64GB RAM, NVMe for OS drive, 2 x NVMe’s sample libraries and a NAND SSD for project files. I was getting dropouts and crackles all day having recently built. CPU load at 8% for Cubase 12’s Pop Demo track, with 64 buffer, Medium Asio Guard with PreSonus Quantum Thunderbolt 2 interface.

I read @Brian_Roland talk about SSDs and proceeding to install the latest drivers/utilities for all my SSDs and the dropouts disappeared! :smile:
I was getting LatencyMon failures too. I shall try that again, to see if it made a difference, but I’m not sure it will for LatencyMon.

I’m having the same constant dropouts and I can’t trace it to anything… Reaper, Studio One, Pro Tools all work completely fine. I stopped using Cubase on v10.5 because of the constant and same issues and totally regret coming back even with the cool new feature sets.

UPDATE: Unfortunately, my previous reply needs changing because I continue to have dropouts, even after running firmware on my SSDs. The firmware helped but I’m getting dropouts even when the project is idle.

1 Like

Thanks for highlighting this. It made me re-read the advice again. I seemed to have missed the part when Steinberg talked of older Cubase versions and that Hyperthreading / SMT should be Enabled if using Asio Guard for Cubase versions after around 7.
I’ll enable SMT on my 5900x with Asio Guard on, and will see if that helps my spikes disappear.
UPDATE: Still getting the spike problem :frowning:

Do you have a network card (ethernet or wireless) active? Bluetooth antenna?

No idea if disabling these will help, but it’s worth a try. Sometimes those things do system interrupts that screw up anything demanding ‘Realtime precision’.

1 Like

I think you should provide a bit more information.

Firstly let us know what you mean by “dropout”. Clearly you can’t have a dropout if the project is idle, at least not the way the word is typically used. So explain what you mean by it and let us know exactly when it happens.

Also - let us know not only what CPU you have and amount of ram etc., please be specific. Give model numbers for the motherboard, memory and drives. Give version/build number of the OS.

I’m currently also using a 5900X on this version of Windows:

image

I use an ASUS x570 ProArt with BIOS 0801 which contains AGESA 1207. Memory and more in my sig.

Thanks @MattiasNYC

Firstly: “Dropout when idle” was clearly a mistake in language. I have dropouts when audio plays and overload spikes when idle.

Secondly: on this thread, I’ve stated 64GB if you look just 2 posts above. I’ve also mentioned details here.

But below, are fuller details, including the info that I missed and which you suggested to include:

My system:
CPU: AMD 5900x

RAM: 64GB (2 x 32) Patriot Viper Steel DDR4 3600 MHz Module - PVS432G360C8

Motherboard: Vision D-P Gigabyte motherboard

BIOS: F15a

Windows: Windows 11 Pro, Version 21H2, OS Build 22000.652

GPU: AMD RX 6600-XT Gigabyte Eagle

Interface: PreSonus 26x32 Quantum Thunderbolt

256GB SSD Samsung 970 EVO Plus, Firmware: 10.022000 for OS/programs

1TB NVMe SSD Samsung 970 EVO Plus, Firmware: 10.022000 for sample libraries

1TB NVMe SSD Gigabyte Auros GP=ASM2NE6100TTTD, Firmware EGFM13.2 for sample libraries

1TB SSD NAND Western Digital WDC WDS100T2BOB, Frimware X61190WD for project files

Hi @Brian_Roland I only tried disabling the LAN Ethernet, because I needed to be online, but I will try disabling the WiFi too, as well as the Bluetooth and will see what happens. I only have 1 mouse which is Bluetooth, so will need to borrow a wired one. I love PC but coming from Mac, it seems very silly to have to disable parts of the computer and make so many trial and error moves with things like BIOS, Reg Edits etc just to get a stable system. After 10 days of trying things, I will still persevere, because I prefer my PC build and Windows, so will hope to get it stable at some point.

I can also confirm that the overload spikes (whether idle or dropouts during playback) happen when I run Cubase 11 Pro too.

With disabling the Bluetooth and WiFi/network cards etc, I assume it is actually going into device manager, right clicking on them and disabling the device? Instead of merely turning WiFi / Bluetooth off via settings?

Thanks.

That’s pretty weird.

Just some random thoughts on things you could try (coffee still making its way into my bloodstream so…):

Maybe try using a different interface driver. TB should be fine but perhaps there’s something with the Quantum driver and Cubase that doesn’t work.

Also, you’ve updated drivers for SSDs and TB, right?

You could check if your motherboard has any known issues with TB. I included my BIOS and AGESA versions because there was a case where Ryzen 5xxx series CPUs would sometimes drop USB for a fraction of a second which would give some users audio glitches. In my case it happened once per day and depending on what I was doing it was sometimes zero issues and sometimes I’d have to address it. TB should be different than USB, but it might be worth looking into (I know it’d be weird if Cubase is the only DAW that has the problem if this was the case).

One issue I had on an early build version of my current computer was a UAD-2 card interfering with my graphics card. The UAD-2 driver wouldn’t install properly and the card was left in the slot. As soon as I took the card out the system became stable. So maybe make sure you have no other peripherals/cards attached for the sake of troubleshooting.

You could also try something like Hardware Info (HWinfo64) to see a maybe more ‘granular’ set of measurements of your system. It might catch some peaks in core temperature or speed that seems off…

Anyway, that’s all I have. Really hope you can get past this…

Well, Apple has full control over all of the hardware that goes into their computers where as for any given Win builder it’s all hardware from different vendors and it’s up to us to make it work. So it’s really not so much a Win v OSX thing as it is Apple builds versus average Joe builds.

1 Like