here’s the link to the audio gridder /Cubase thread
1 Like
Thst bothers me most. Enabling monitoring on a track without plugins and in a one-track project can raise audio performance monitor higher than in previous Cubase versions.
In order to achieve this did you have to declare plugins as external processes?
I don’t personally have much experience with AudioGridder and none at all when server and client is running on the same machine. But, I have seen examples where in use, a better balance of system resources appears to have been achieved.
I understand the “one thread per audio stream” concept that Cubase employs and while it makes logical sense to me, it is apparent it is not the only concept that works. If it comes at a cost of higher latency, I don’t know but it would surprise me if there is no tradeoff.
I don’t think it is a secret that when it comes to utilizing system resources efficiently, Cubase is not the leader of the pack.
3 Likes
When you say “Cubase will max out,” do you mean to say that specific tracks actually have audio dropouts, or that the APM shows higher real-time dsp load? It matters, and in all the discussions around this I’ve never actually had anyone give precise information.
The APM isn’t a CPU meter; it shows the peak load of whichever thread has the highest requirements for real-time DSP. Meaning, just because the APM is showing “pegged” for one thread doesn’t mean that you can’t have 32 other identical tracks all running just fine on their own cores. It just shows a “higher risk of audio dropouts” for that thread.
By way of example, if I have 50 tracks in Cubase and my APM is showing the peak thread is at risk of dropouts, if I move 25 of those tracks to Ableton Live and sync the transport and use ADAT (or whatever method I want) to bring the mix together, then the APM can certainly go down for Cubase. But my system itself is still processing that data between both DAWs. This is in effect the same thing Audiogridder is doing. It’s just moving the process outside of Cubase’s visibility while actually increasing your latency. That probably doesn’t even matter with you mixing, but it matters if one is using the APM as a “cpu meter” if you think it applies to the entire system. It doesn’t.
I just think it’s important to be using the same metrics, and of course to define what those metrics are in these discussions.
3 Likes
Yet that is exactly what should happen within Cubase. The end user shouldn’t see anything – seamless. But internally it should do whatever it needs to do to use more cores/threads. If that means opening more processes and if that increases latency then so be it. It would be nice if there was a mode to allow that.
1 Like
No that was just regular Reaper default.
M
By max out i mean rhe ASIO guard meter hits the red and you get dropouts.
I’ll add that on Mac Os and apple silicone, this doesnt happen. The cubase performance meter and the Apple performance meter are well in sync. So on my M1 max i can load a project that is maxing our my 9950x in cubase ,but only using 25% system, whereas on the Mac it’s using 90% system usage.
The issue windows users are having is we cant get to 90% system usage.
M
3 Likes
Try disabling SMT in BIOS. I know it’s not what you asking for, but at the moment it might actually improve multiple core usage and improve performance.
Here’s an example of 11 instances of Serum2 playng heavy preset on 5950X with SMT off:
1 Like
Got it. And when this happens, if you use audiogridder, that solves the problem with that specific project? My specific question is if during an actual dropout scenario, moving to audiogridder has solved it - THAT would be very interesting to know (to me).
Moving plugins to AG completely solves that problem as the cubase ASIO guard stays low.
M
1 Like
Damn! I never heard of Audiogridder. Just tried it and it works marvelously! What a gem this thing is!
Anyone know how to get instances of HALion 7 to work with Audio Gridder?
I have been saying for YEARS, literally years now, on this very forum, that the Cubase audio engine on Windows is seriously deficient, that there’s something very wrong.
It’s so validating to finally see you all have uncovered some more hard proof in this thread, despite the Steinberg Defense Squad doing what it reflexively does.
6 Likes
and without disabled SMT? all 32 molasses, for the purity of the experiment
I’ve never tried instruments - only effect plugins.
Seven instances of Serum 2 with SMT On vs seventeen with SMT disabled:
Windows 11 Pro 24H2 26100.4652
1 Like
So is this showing that 7x instances of Xfer Serum 2 with AMD Simultaneous Multithreading (SMT) Enabled has worse performance in the Cubase Audio Performance Monitor compared to 17x instances of Xfer Serum 2 with AMD Simultaneous Multithreading (SMT) Disabled ? Is that correct ?
If I remember rightly, it’s long been recommended to turn off SMT or HyperThreading, as it is sometimes called.
Yep. I also have Hyperthreading disabled on my Intel CPU.