I just ran a Dolby Atmos session from 14 in 15. It’s pretty big and sits around 70% asio guard in n14 and around 85% in n14. Is anyone else taking a performance hit after upgrading?
update…
i just closed and opened and it’s running better in 15 now.
Way worse. The same project running perfectly in N14 is overloading N15.
That is not good news, as N14 was already a performance downgrade relating to N13… Haven’t yet upgraded, but the waters seem a bit murky,..
I agree. I wish there some sort of clarification as to why N14 and N15 have ASIO Guard maxed out, but only 25% cpu usage in task manager.
Have you tried to troubleshoot it? Tried to find out what causes the overload?
I know that there has been some finicky things with DAWs in the past where loading-reloading can change behavior (not necessarily Nuendo). Maybe try to create a new, clean, empty project in N15 and then import tracks several chunks at a time into it from that older N14 project to see if there’s one chunk that causes the overload, if at all?
All I’ve done, for now, is start a brand new big project in N15 and then open it in N14. N14 ran better.
I’ll be away for a week, but when I’m back I’m going to start a new thread about this. I really, really need to understand why N14/N15 don’t use all of the available CPU processing available. I need some solid advice as to what direction to take. Building a new machine every 2-3 years is just crazy and costly..
I spent most of last year with Vin at Dawbench going through all this after I bought a 10 core M1 max and it ran my maxed out 16/32 core windows projects!! i thought apple had discovered Alchemy or something….. it’s only when I looked at the respective task managers i saw the windows machine was only using 25% of it’s resources when ASIO guard was maxed out, wheras the Mac was runnign 90%… this led me down a lot of testing and eventually spending $4000 pus peripherals on a Mac M3 Studio ultra 28 core machine…….
Only to find this exhibited the eact same issues…. ASIO guar maxed out with 25% cpu usage across cores……
I sold the M3 after a month as it performed worse than my windos 11 9950x machine.
There’s something needs adressing with thread management. If you copy and paste a track with FX, MAc OS and Windows will load the threads perfectly……. however with a normal mix project with lots of busses/Aux routing etc this all falls over and ASIO guard maxes out at only 25% cpu usage as you’ve seen.
The only way round this I’ve found is to use Audio gridder locally and load plugins via the AG plugin inside Nuendo, that way you can actually make use of the 70% system resources sat idle.
There a huge thread on the cubase forum about this. Steinberg are very aware but i think the whole audio engine need a re write and that’s not going to happen soon.
M
3 Likes
I’m gonna defend Steinberg a little bit here (regular readers of my posts will be shocked!) because in the earlier days of multi-processors the Cubendo audio engine handled threading far far better than Windows did natively. Obviously that’s not the case anymore and yes it absolutely needs addressing but it is a massive piece of technical debt so who knows when they’ll have the resources to address it.
3 Likes
@EdGray They are aware of the problem and on it (confirmed by Steinberg in a post someplace else). They also said that it will take time to get ahead of the problem due to its complexity - you’re spot on.
3 Likes
Here is the core handling thread in the Cubase forum. With all respect for the potential complexity of this issue, I really think Steinberg should give it the highest priority, since it actually means that Cubendo cannot utilise the full processing power of modern multicore cpus.
Best,
Magnus
9 Likes
I think so, too…above anything else they are working on IMO. My big scoring sessions are often crippled with a very fast Mac Silicon computer. I don’t mean this in a mean way, I’m just being real.
3 Likes
I can confirm that the same project runs with less ASIO Guard utilization in N14 than in N15. I have tested it with a few (big) projects and get 5-20% more utilization in N15 compared to N14. It’s totally understandable that the core handling problem is a tricky one but it should at least don’t get worse with a new version of Nuendo in my opinion 
I’m working on Win 11, AMD Ryzen 9 3900X, 64GB Ram, everything up to date (sytem and drivers).
It has been mentioned already that using “external” virtual racks like Vienna Ensemble Pro or Audio Gridder will help a lot in our quest to use our machines’ power to their full extent. But there’s a little-known work-around I use all the time to force Nuendo to make better use of multi-core CPUs: “Splitting” the processing chain across several channels.
The method is a bit cumbersome GUI-wise, but actually very easy to implement: Instead of inserting all the plug-ins into a single channel, I simply add a bus and route only the original track to it. When the next two or three “heavy” processors are inserted there, Nuendo calculates them on different CPU cores, obviously.
This is especially helpful in case of multi-channel formats and/or “expensive” mix busses with lots of demanding plug-ins.
2 Likes
FWIW, I think this is recommended practice in Logic Pro, which has always had far less multi-threading capabilities than Cubendo…
I can’t compare it to N14, but N15 has noticeably better performance than C14 on my Mac/Sequoia. C15 does not have better performance than C14, interestingly.
1 Like
I’ve been using N15 for a bit now and I’m pretty happy witht the performance compared to n14. I have a project sitting around 75% asioguard in 15 and it is the same in 14.
As for overall cpu utilisation, fingers crossed there are big positive changes on the way.
2 Likes
That’s a nice tip for when things get really tight, thanks!
I’ve also found that it can help to remove individual processor cores from Nuendo.exe (via Task Manager). Especially on my laptop, it always helps to remove the very first core (CPU 0), because Windows seems to run some power-saving processes on it that cause sudden spikes.
But even on my desktop, removing every other core from Nuendo reduces ASIO Guard utilization by 5–10% for some projects. The remaining cores are then suddenly utilized more heavily, but this actually gives me more headroom in ASIO Guard (counterintuitively). It’s as if this forces Nuendo to distribute its processes more efficiently, while simultaneously giving Windows more room to manage everything else.
I tried now to disable Core 0 in Task Manager by
Nuendo.exe - Go to detail - Set affinity
Will report if I can feel any change - Will someone make a benchmark on this on a larger session?
Are you guys setting the affinity while the process is running via the Task Manager, or are you starting the process via command line (or shortcut) so that the affinity is set while the process is starting? Just curious if you’ve experienced a performance difference based on when the affinity is set.
I’m setting it while Nuendo is already running via Taskmanager because Nuendo seems to restart the Nuendo.exe when loading a new project (or switching from the hub to a project) and I would have to re-set it anyways then. If someone has a workaround for that, it would be great. With Ableton for example, i can start it with a shortcut that sets the process affinity for me and it stays the same as long as the program is running.
1 Like