I’m currently in the finishing stage of a new project and I’m having massive trouble getting Nuendo to playback the project without any glitches, even with all tracks frozen/rendered and running at a 4096 buffer-size.
I thought my system is pretty potent:
AMD Ryzen 9 5950X (16-Core)
128 GB RAM
NVIDIA GeForce RTX 4070 (Studio-Driver)
Samsung 980 Pro (m.2 only)
Windows 11 Pro
RME Babyface Pro (4096 buffer-size)
Nuendo 12.0.70 Build 464
The project is running at 96khz/32-bit-float and it contains only 37 tracks including 10 VSTi-tracks and about 9 audio tracks for vocals and external synth recordings. The rest are subgroups and FX-channels. I have some automation running and I’m having some cpu-intensive plugins on the MIXBUS as well.
When I hear people talking about how many tracks they can run on this CPU, they are always talking about 100-300 tracks, which is far more than what I’m getting here.
It’s not that the project doesn’t play at all. But sometimes the timing is really off or it might stutter randomly. I’m not getting the typical clicks and pops.
I’m also an IT-professional, so I guess my system is pretty optimized as well. LatencyMon doesn’t show any problems at all. So maybe it’s related to using subgroups in an inefficient way?
I’ve read somewhere, that routing is a critical factor for multi-processing. So I hope someone might enlighten me on why I can’t playback this project properly.
Since I’m in the final mixing/pre-mastering stage, I have frozen all tracks, but kept most of the FX live. I have also tried to render all tracks to audio, which essentially should be the same as freezing, right? And I have tested different ASIO-Guard-settings too.
My Laptop is doing live recordings of 60 tracks. (Dante Virtual Soundcard)
My Studio system plays 80 tracks with ease (never tried more), so I guess your system isn’t optimized at all.
The interface is a UA Apollo USB @ 48kHz with a buffer size of 256.
Try to set the buffer size to 1024 first. Maybe it helps. Sometimes is the highest setting not the best.
That was the mixing setting with my Fireface 800 (FireWire).
And check your USB drivers, there could be an update available.
Thanks for your input! Of course I have gradually increased the buffer-size and didn’t start at the highest value right from the start.
I think recording audio @48khz is just not as demanding as mixing with a lot of plugins @ 96khz. So that’s not a real comparison here.
I have followed all recommendations from Steinberg related to optimizing the OS, so I don’t know what to do any further? The latest drivers have been installed.
Does it work when you disable the VSTi tracks? Depending on what they are, there is lots of potential for things going wrong, especially at 96 kHz. (BTW: … is there a real need for that SR, or is it just because it’s said to be possible?)
Does it work without UAD plug-ins?
I see several PA Overstayer instances in your mixer. This is a plug-in that caused problems on my system. Does it work if you de-activate them?
Since I have frozen all of the instruments, there’s no active VSTi in this project.
Interesting info about Overstayer though, since this is in fact one of my latest plugin additions. I’m gonna try it without it and I’m also going to test this project on my M1 MBP, just to check out an alternative system.
Thank you for your help!
So, you think grouping in general has nothing to do with it?
BTW, the UAD-Plugins are all from the native Spark-subscription.
The only time Nuendo slowed down here (and I’m doing Dolby Atmos, so 48 kHz, 24 bit, but a 512 buffer, no more), was when I used a lot of stretching on sound effects without rendering the processing in a new replacement file. With around thirty stretched events, Nuendo was running out of steam, as it has to calculate everything in real time (on top of everything else). But otherwise it’s not normal. Maybe it’s Babyface…or a poorly designed plugin. Keep us posted, I’m curious to find out.
I’m assuming you’re running all of this off of the NVME m.2 drive, correct? And that drive is doing nothing but streaming that audio?
Well, your signal chain ends up being very long for some paths. If you look at your VOX-FX tracks for example they have 10 plugins on them. Then they go to your VOX bus with 2 plugins. Then that goes to the mix group with another 6. That’s 18 plugins in series which is a hefty amount I’d say. I’d look at that as twice the amount at half the rate, i.e. 36 plugins in series at 48kHz.
So if any of those plugins are ‘heavy’ plugins it’s going to make things worse of course. “Heavy” to me would mean either poorly / inefficiently coded plugins or plugins doing a lot of modeling… which… UA?..
I would simply try to deactivate the most ‘suspect’ plugins one by one to see if there’s one in particular causing a problem. I’d start at the end on the master and go backwards.
I agree with Mattias. It’s asking too much of the system.
The trick, once the plugin tests are done, would be to render with effects for at least the source tracks (first level). Or simply change the way you work by streamlining the processing (which is always a good thing, after all).
All DAWs recommend that playback of audio, and the project itself, be stored and used on a dedicated audio drive, and NOT on the operating system drive. You do not mention if this is what you are doing here.
Was there a moment after which this particular session began having isues? As in, you instantiated a new plug in, or synth, or channel, and it began having problems? If that is the case, then look at that situation. Maybe a plugin is not the right version for your system, maybe it is just a glitchy plugin…
I just mentioned the UAD-native-plugins to make clear, that I don’t get any additional I/O-latency by using a DSP-card.
Is your system game optimized, btw?
I’m not a gamer. This system is a workstation for video and audio-production.
I’m assuming you’re running all of this off of the NVME m.2 drive, correct? And that drive is doing nothing but streaming that audio?
All DAWs recommend that playback of audio, and the project itself, be stored and used on a dedicated audio drive, and NOT on the operating system drive. You do not mention if this is what you are doing here.
Yes, I’m not using this drive for anything else, the OS is on a separate NVME SSD.
I guess this is really a case of using a lot of CPU-hungry plugins in series and I have identified Shaperbox to be one of the main offenders here.
Anyway, thanks for your input, I just wanted to know, if the routing might be an issue at all.
I ended up ditching Nuendo for work because the audio engine stalls just by adding some plugins to a few realtime chains (a group, fx, master bus). You can play hundreds of tracks with several plugins on them rock solid as long as they’re just streaming audio, but can’t get away with some subgroups with some processing on them, even if none of the tracks routed to them are monitoring any audio.
Only workaround I found was to use something like Audiogridder and do the bus processing outside Nuendo, but it’s a workflow killer. Also using UAD plugins for buses help. I’ve been waiting for ages for this to improve, and finally gave up. Still hoping they fix that thing someday, though (and hope to be still making music by then…)
I went through the hassle of rebuilding the whole project in Reaper, because people kept talking about its small cpu-print and performance advantages, which I always tended to discard as fanboyisms.
I was really sceptical, but the project is running in realtime now, without any hiccups and I still have all of the heavy mixbus-effects like Gulfoss on there. This way, I can even record new vocal takes without any freezing or rendering at all. Crazy!
