I use Audiogridder, too. But none of this should be needed. Which is why Steinberg need to address this asap.
Best,
Magnus
I use Audiogridder, too. But none of this should be needed. Which is why Steinberg need to address this asap.
Best,
Magnus
Now…..if those three bandaids solve, let’s say 50 other systems in the same manner, what would Steinberg hypothetically create?…..
1 Checkbox option on the asioguard settings area to “ignore core 0”?
Stuff like that as a relatively quick SB bandaid to your bandaid….just throwing the off top of my head thought.
That is, IF your 3 steps seem to have the same positive result on bunches and bunches of systems.
Or something like that.
Could you share a few more details about your setup? For example, what kind of projects, effects, instruments, and audio settings are involved when you notice performance issues?
Your system sounds quite powerful, yet you’re experiencing problems that most users don’t seem to encounter as frequently.
Having more information would really help our team build better real-world test scenarios.
My projects are audio only, at most some 20 audio tracks (mono and stereo, my own recordings), with 4-5 group tracks and 5 or so fx tracks, plus some mix buss processing. The mix buss is what sometimes maxes out the asio guard, but that is not surprising.
I almost exclusively use Acustica Audio VST3 plugins, a few from other devs, like Hornet, Waves, UAD and Pluginalliance. All my projects are 44.1/24. Cubase is set to 64 bit processing. I have an RME AIO Hdspe soundcard running at 512 samples buffer size, and a Geforce RTX 3060 gpu for Spectralayers Pro.
I run the Acustica plugins in OS mode when needed (but not excessively high OS).
I could send you some actual cpr files if you like.
Best
Magnus
Would you believe if I told you I’m still using one Delta 1010 with a 8700k, Nuendo 14, Windows 10, and going strong? With minimal latency? ![]()
I’m also an enthusiastic endorser of Audiogridder for heavy plugins (still haven’t tried it on vstis, what’s your experience latency-wise?)
Many drivers do end up on Core 0 because they force themselves there. Core affinity is a really annoying thing.
However, that’s not at all guaranteed, as the matrix of CPU types and drivers is pretty huge. But if it works for now, all good. But I wouldn’t put options into any software to specifically exclude Core 0, or any other single core.
Pete
Microsoft
I ran a VSTi load test not so long ago (comparing HT on vs. off). Watching the resource monitor I noticed that Cubase (14 in this case) was avoiding core 0 until such time as the other cores started to reach their limit, at which point core 0 started to load up, although it never got beyond around 50% util. With HT on, both Core 0 and 1 behaved in a similar manner.
I didn’t pursue it as it wasn’t what I was looking for at the time, but it appears Steinberg has already built this in.
Please get back to me regarding my reply.
Best
Magnus
If you could sent me a message with a download link to your project, I will forward it to the team. Thank you for you support.
PM sent with a link to a cpr, which is a good example of how Cubase on Windows distributes the processing load very unevenly between cpus 0 and 1, and the rest of the cores on modern cpus, resulting in Asio guard overloads long before the total cpu load is even high, in mixing projects with a lot of processing on mono and stereo audio tracks, on group tracks and on the mix buss.
It would be so great if Steinberg could finally address this.
All the best,
Magnus
Did you try disabling Nahimic ? that is the one change that helped my setup a lot.
Win search services, find it down, disable.
It took a few days to gather more information, but I had a longer discussion today with my colleagues who are working on the audio engine, and they shared the latest update with me. Hopefully, this will give you some answers.
Short answer: No.
Long answer: As I mentioned before, we have a team that is constantly working on performance improvements in many different ways. A big part of that is making the audio engine more flexible and robust so we can explore new strategies needed to adapt to emerging technologies, both hardware and operating systems. This takes time, as we’re dealing with by far the most complex, layered, and interdependent part of Cubase, where even minor adjustments need to be made with great care. We’re fully aware that there’s room for improvement in the core handling, and we’re actively working on it. We’re just not there yet.
As far as I know, Cubase performs quite well with single instrument instances. The issues tend to appear in more complex setups with many groups, track connections like sidechains and sends, and similar configurations.
Unfortunately, that’s not how it works. There is no simple solution to such a complex problem.
I believe it!
We’re fully aware that there’s room for improvement in the core handling, and we’re actively working on it. We’re just not there yet.
Thank you Matthias for acknowledging the issue. Do you have any timeline for when we can expect the updated audio engine?
All the best,
Magnus
Thanks for providing an update to the community. It is greatly appreciated when users are not left completely in the dark.
Thanks for digging into this with the team, Matthias. As these audio engine optimizations move forward, many of us are really hoping true gapless playback is part of the plan — engine hiccups and dropouts are still a major workflow killer in Cubase/Nuendo.
So what would you recommend instead when issues arise as described above?
Thanks Matthias for this detailed message. I feel you and I keep my fingers crossed for your success.
From the Cubase perspective, we’d rather recommend taking another look at your project and routing layout, as well as your plug-in usage.
CPU-intensive plug-ins in groups, buses, or output channels can cause a large amount of processing load to shift onto a single core. Also, sends and side-chains that connect tracks with a high CPU load can create situations where all connected tracks end up being processed on a single core.
It’s also worth identifying plug-ins that are particularly demanding on the CPU and seeing if there are lighter alternatives. Another option is to freeze tracks with lots of plug-ins when your system starts reaching its limits.