problems w/ high-end CPU and more ...long post

after some extensive testing and tweaking, I wanted to share my experience with C8.5 and C9 on Windows, running into problems with audio-dropouts, even though my PC-system is very new, built by a known audio PC store in Germany.
I managed to have pleasing results finally, but I am not shure my system performs as well as intended.

(Careful, this is going to be very long. Those affected or interested, bear with me, as I want to give detailed information. Also, please excuse any inaccuracies on the technical side.)
For all the impatient users, here’s a quick overview of what I did and actually helped. But if you read on, you’ll see that it’s probably different for each user and DAW…


  • BIOS settings are very important. In my case turning Hyper Threading OFF was the biggest help of all, even though Steinberg says it is supported, and many users confirm this.

  • obviously Drivers are important (graphics, network, audio of course, doesn’t have to be the newest either)…

  • This gave me better performance as well: Apparently Windows Defender gets stuck in a loop sometimes (, searching itself, hence driving CPU usage way up, so it helps to exclude these folders in Windows Defender Settings:

C:\Program Files\Windows Defender
C:\Program Files\Windows Defender Advanced Threat Protection

Also, if you haven’t done this already, exclude all drives but the system drive, otherwise Defender will search all samples/projects.

  • Cubase Preferences:
    I work with folder tracks a lot. When all folders are closed, zooming gets VERY slow in edit window. I disabled Preferences → Event Display → Folders → “show event details”. This got rid of the slow zoom problem.


now to the whole story…

I recently moved from working on Hackintosh PCs (3 systems built), to working entirely on Windows. Yes, it’s true, Hackintosh can be a pain, but once they are set up correctly, they work just fine.
So I was well surprised, like many of you here, to see that my brand-new, top of the line i7 6900k system was very unstable, be it with 2 VSTi tracks or a dozen random plugins on audio-tracks, I had constant audio-dropouts and ASIO spikes, slow graphics and crashes on a regular basis… Benchmark Results were impressive nonetheless.
C9 crashed so often I quickly stuck with C8.5, as I could work for sometimes 2-3 hours before the slugishness and crackles began.

So, what the hell is going on, right?..

Since I had worked in the OSX world for about a decade, I was patient, I knew there was lots to figure out in the Windows world. I tried out many many options, configs etc, some things helped, lots of things didn’t… then I found this post a few days ago:

The film was an eye opener. A lot of things make much more sense now. I can’t really blame the companies who produce the hardware. Audio production might be a big deal for us. For Intel, Nvidia, Microsoft, etc. it’s a different story. They just don’t have to address problems so quickly, if they concern <0.5% of all the customers (that’s a wild guess here).
Keeping in mind what Richard Ames said, I started looking into my system again

Windows (System) Settings

  • Although I think Windows Defender is important and works fine, excluding the folders I mentioned above seems to help. I don’t use other Anti-Malware programs.

  • I found out that Chrome browser uses a lot more resources than Edge browser.
    So I switched.

  • Did some other tweaks here and there, can’t say it did anything noticeable. With CPUs and GPUs as powerful as they are now, this shouldn’t weigh so much IMHO. Plus, it seems we have to get used to Microsoft / Apple dictating how we use their OS, with the way they force updates etc… at least for the time being.

  • That said, I set Windows to defer driver updates. Also, I try to delay OS updates, so I can at least chose the time of updates.

    Other Software

  • I use Dropbox. I actually work on sessions inside the Dropbox folder. This worked great on Mac, so I expect it to work on Windows. Dropbox does use some RAM and CPU, but never conflicts with Cubase. Disk performance is not affected by Dropbox syncing active session folders.

  • I also started using Synology Cloud Station Backup for other files such as sample libraries or other. This program uses more resources. For the time being I turned it off, doing manual backups of non-critical stuff every other week or so.


  • I had to learn the hard way, but video cards are big deal. My geforce 1060 with 6gb is a powerful card, but it’s apparently a bottleneck. Why?
    I run 3 displays (1 hi res, 2 fullHD), and that could get demanding I guess, even in 2D. But the problem seems to be the CPU latency as explained it the video above. On my PC, the program “LatencyMon” shows Nvidia and Directx drivers as being the culprits, blocking access to other apps using the CPU.

  • I haven’t found out how to improve this driver issue. (Installing the newest driver made things worse, so I went back to the older version.) I turned off every Nvidia service I don’t need. Though, again, I don’t believe this is the real problem. As described in the film by Richard Ames, I don’t care if other apps and services use the CPU, cause there’s plenty of resources there. I care about how they use it.


  • By accident I discovered on the RME forum that RME Fireface settings have the option to enable/disable MMCSS for Windows. This is supposed to suspend CPU blocking by different hardware/drivers, which can cause the realtime audio drop-outs. For some reason the option is in the “about” rider of the Fireface settings (where I’d have never looked myself).
    Anyway, toggling the state of MMCSS did no noticeable changes.

    Thread Affinity / Process Lasso

  • This user is definitely on to something:
    I don’t know enough about the backend of operating systems or software in general, but to me, it sounds quite logical what user Reflection says.
    I installed Process Lasso. The company who sells it claims it can do a bunch of stuff, for example automate the thread affinity. I saved only 2 threads (1 core) for Nvidia- and a few other services, and assigned the rest of the threads to Cubase, so they wouldn’t conflict with each other.
    Process Lasso can also “tame” processes when they want to draw too much CPU resources at once (similar to MMCSS?), thus reserving them for the app that’s most important (in this case all audio relevant apps/drivers).
    It definitely looks great to see Process Lasso blocking hundreds of CPU spikes a day (mostly Edge Browser, some Dropbox, some Network stuff); BUT, i can’t say either option works for good. I thought it helped, and maybe it did, but it didn’t give me the stability in Cubase I was looking for. So I removed it again. I suppose if it were all that simple, ppl would have caught on earlier, or at least this program would be better known among PC audio producers…

Btw, setting thread affinity manually in Task Manager had no noticeable effect either.

Intel Turbo Boost / Hyper Threading

  • For about one month I didn’t bother tweaking BIOS CPU settings. First of all, I assumed it’s all properly set up, being a designated audio workstation. Second, I had read that Cubase now handles Hyper Threading just fine.

I turned off Turbo Boost. I don’t remember if there was a big difference.
Then I turned off Hyper Threading, and this was the one setting that made a HUGE difference.

Until I found this out, I hadn’t done much with C9, cause it crashed every hour or more, and was just worse in every respect. All the new features I didn’t care about, and performance was real bad, not just crackles and drop-outs, but also loud noise bursts etc.
C8.5 did give me spikes and slow performance, but I could work in a session.

Now I can work very well for long periods of time in C9 as well. I still get the occasional crash, once every 2 days, but that I can handle.
C8.5 and C9 run quite decently. I haven’t turned Turbo Boost on again, cause I’m so happy it works well now. But I might try that soon.
Now, there’s still things I’m unhappy about (read here:, but I’m a lot closer to a high-end workstation workflow.

During all the testing, I noticed that if I start a session from scratch, I generally have higher stability in a session. For example, some sessions that I started on OSX a year ago and use as templates, they tend to be way more demanding on the system. This has gotten better with the tweaks, but it’s still bothersome. I have to be able to rely on my old sessions/templates. Otherwise I think the whole Idea of software updates and backwards compatibility etc. would be kinda useless.

So is there any conclusion?

From what I’ve gathered around the forums, there is no real solution. Definitely no “one-for-all”. Apparently any Display/USB-device/Driver can mess with CPU realtime-processing (reminds me of Hackintosh days), and given the almost infinite number of different hardware combinations, no one can guarantee for a perfectly tuned audio workstation.
ATM it seems that: Cubase on OSX (including Hackintosh) handles realtime audio better than Cubase on Windows, but OSX users have more trouble with slow/bad graphics. Very generally speaking. Would you agree?
A good friend is building his new PC right now, with OSX and windows installed. Will be interesting to hear out about his results…

Why doesn’t Steinberg recommend builds for Cubase? Cause they could be held accountable if the system doesn’t work as expected? Doesn’t Steinberg use certain PCs to test their Software? Even if it’s just mainboard, CPU, RAM, GPU… wouldn’t that be a great starting point?

I’m happy to hear your thoughts, and I hope we can gather useful information. I’m looking forward to 9.0.20, I know Steinberg are working on bugs etc… so let’s try and keep this constructive.


So true. This is exactly the same conclusion I arrived some months ago with my i7 5960x (8 core) system, and it seems to be applicable to any i7 CPU with 8 or more cores. In my setup, disabling hyperthreading makes audio dropouts completely disappear when running Cubase at very low latencies. In fact, I’m currently running Cubase at a latency of 32 samples by default with an RME AIO PCIe soundcard, and the performance I get is truly fabulous. With hyperthreading on, I had to raise the latency to a minimum of 256 samples to avoid the dropouts. Big difference.

The optimum always seems to be between 6 to 8 threads in my experience if you want anything like usable low latency,
with “Real” threads being better than hyperthreading ones if you have a choice.
On my machine 6+6HT is not as good as 6 real only.
4+4HT is not as good as 6 real and about the same as 4 real only
Turning off HT makes for and easier overclock and the overclocked 6 real gives the best performance of all.
This is on this machine with my usage pattern but every one is individual all I would suggest is experiment.

The only problem with experimenting with HT is, every time you turn it off (or back on again) I have to re-authorize some of my plugins (I guess the authorizations are tied to the cpu).

See if the plugin manufacture will give you two authorization’s, many will and register one with and one without HT.

This is interesting re HT. I’ve always been recommended to keep it enabled so have.

What I recently found out though was the huge difference my GPU settings made. I’ve been using an AMD R290 for years and just changing the graphic profile from ‘balanced’, to ‘optimise performance’ by right clicking the Radeon softwares system tray icon gave me a significant boost which like said above, allowed me to regularly play at 64-96ms buffer rather than the previous 256ms min when using my default production template.

Full thread on this here.