Higher latency needed for tracking audio in Cubase 13?

Cubase
1

Is anyone else finding that, with the same computer and audio interface setup, and a very minimal amount of other things going on in a project, higher latency is needed to track audio without artifacts in Cubase 13 than in Cubase 12?

My scenario here is that I’ve frequently had issues in the past with the potential of artifacts like crackling or clicking (that sounds like crackling, but iZotope’s RX only fixes it with de-click, not de-crackle) getting into mono vocal tracking (I work at 96 kHz). When this has occurred in the more distant past, I could usually reset the audio interface (sometimes more than once) to eventually get it to not happen. In the more recent past, certainly with Cubase 12, I found that just changing the sample buffer size within the ASIO control panel would usually do the same trick. Late in Cubase 12 days, I also found, quite unintuitively, that changing it from my past norm (256 samples) to a lower rate (192 samples) usually made it better.

With Cubase 13, though, the only thing that has worked thus far, and sometimes only partially, is going to a higher sample buffer size (384 samples). While that seems to mostly avoid the constant crackling-type sound, it still seems to get some other artifacts occasionally.

It is worth noting that, in these Cubase 13 cases, I have been tracking against a frozen submix – i.e. no active effects. In the early case I saw with V13.0.10, there was only one stereo audio track for tracking against. In yesterday’s case with V13.0.20, there were two frozen tracks (they were going through a group channel to adjust the combined level of the submixed tracks, but with no plugins on any of the tracks, just different fader balances). It may also be worth noting that yesterday’s case did have a tempo map active (the previous case did not).

It may also be worth noting that my configuration with Cubase 13 should actually be quite a bit faster on the overall disk I/O front now that my system disk and sample libraries disk are SSDs. However, with respect to playing back and tracking audio, that wouldn’t likely come into play in that my audio disk (also SSD) has not changed between Cubase 12 and 13.

My configuration is an Intel i7 5820k CPU with 16 GB RAM running Windows 10 (fully updated), with an Nvidia GeForce GT 640 graphics card, running 2 1920x1080 monitors. The SSDs are all Samsung 870 EVO (2 TB apiece for the system and audio disks and 4 TB for the sample libraries disk). Audio interface is a MOTU 828x (running via Thunderbolt II). LatencyMon reports that my system is up to running real-time audio.

While I know my system is dated on the CPU side, and I have had to work around that by freezing, rendering submixes, etc., the tracking scenarios here are super minimal, and I’ve certainly had more going at times tracking in Cubase 12 without this issue (i.e. after making the above-mentioned adjustments if there were issues initially), without having to resort to sample buffers higher than 256 samples.

2

Not here. We just finished our first major tracking and mixing project for a client using Cubase / Nuendo 13. We found performance for both tracking and mixing to be virtually identical to V12.
Machines are i9 12900K (Nuendo) and i9 13900K (Cubase). Both on Windows 11.
As you point out, you’re running an older machine and tracking at 96K, so I suppose it’s possible a machine running at or near its limits may get pushed over the edge earlier by Cubase 13 than by V12, circumstances depending.

3

Graphics card , i had it with C12 and the Gt730 , i hate to say the lower 700 cards are not up for the job now . Even thou they are supported they are a MAJOR bottleneck

4

Yeah, perhaps. That is part of why I’m asking – just wondering if anyone else may be seeing something similar. There was something that emerged in an earlier thread where a Steinberg engineer was speculating whether some of the tweaks they made for hybrid CPU performance may have adversely affected older CPUs. That was with 13.0.10, so I was hoping 13.0.20 might help, but there was nothing in the release notes to suggest it, and yesterday was the first time I got to tracking audio with 13.0.20. (My instrumental tracks are virtual instruments, so it is only for the brief periods when I’m tracking vocals that I see this issue.)

Unfortunately, replacing CPU/motherboard/memory is not likely to be in the financial cards anytime soon.

5

That may be one area I get to check out in the foreseeable future as my son is planning to give me a GTX 1050 left over after upgrading his card to a newer gaming card. That one also has Studio drivers available for it, which mine doesn’t. It’s just a matter of “when” on that front…

Of course, I was using the same old card with C12 without having this specific issue.

My current card nominally isn’t supported by Photoshop 2024 (due to lack of DirectX 11 support), though it actually works fine there despite the warnings I get every time I load it.

6

I recall the same thing although I’ve seen no mention of it since.
I will say though that one of our test machines is an i7 6700K (our former Nuendo production box). We ran V12 and V13 back to back on it and noticed no difference in performance there either. The graphics card on that box is an AMD 5450 passively cooled unit that is as old as the hills, but it didn’t seem to make any difference between V12 and V13.

7

I just looked up the specs on the 6700k to compare to the 5820k, and it looks like mine has more cores (6 vs. 4), but lower CPU speed (3.6 GHz vs. 4.2 GHz), and I do at least understand these days that the CPU speed is likely more important with respect to DAW latency.

The main issue I’d seen, other than this with audio tracking, is that I’ve sometimes needed to apply the “trick” with adding a HALion track that isn’t doing anything but has the focus in order to allow better CPU thread balancing. But I even tried that yesterday with my audio tracking, just in case it might make a difference, and it didn’t.

Of course, it is conceivable that something changed on my system (e.g. due to Windows updates) that is affecting this area, rather than something specific to C12 versus C13 since I haven’t even attempted going back to C12 since upgrading to C13 (e.g. to see if I have the same issue in C12 now). And my computer cannot be a dedicated DAW, so there is always the possibility that something other than just Windows has contributed.

Anyway, thanks for the inputs.

8

Yes, if I was forced to choose I’d always pick scale-up over scale-out when it comes to audio processing.
Good luck, I hope you get it sorted.

1 Like
9

I have a 5820K system but I overclock to 4.6Ghz. I also stick to 48Khz sample rate 99% of the time. I have not noticed any difference between CB12 and CB13. This is with orchestral stuff with 60 tracks or so and up to 6 Halion copies running. I usually use buffer set to 256 but sometimes use 128 or even 64 if I am noodling on guitar just to keep the latency low. So mostly no audio just synths and a few VST FXs .
Do you hear any difference between 48Khz and 96 Khz ?Do you really need 96Khz for what you are doing?

10

My recordings are mostly songs, most typically with anywhere from 1-12 tracks of vocals, and typical band configurations (with virtual instruments including Superior Drummer, various options for virtual basses and guitars, piano and/or organ/synths) plus whatever else might be needed for flavor on the virtual instruments front (e.g. more synths, orchestral instruments, brass, etc.), though the one I was working on the other day is just stripped down piano-vocal (with NI NOIRE and Arturia Piano V3 doubling, but both tracks frozen when tracking vocals). I don’t think I’ve ever gotten as high as 60 tracks – I’d guess most are in the 16-32-track range (though most virtual instruments would be stereo, so maybe that doubles the count?).

I’d pretty much standardized on 256-sample buffers at 96 kHz (equivalent to 128-sample buffers at 48 kHz), though, in C12.0.70, sometimes I inexplicably had to use 192-sample buffers to address the crackle/click issue when tracking vocals. The 256-sample buffer size felt like a happy medium for latency versus performance considerations, especially when tracking a virtual guitar through an amp simulator that might add some latency to the basic ASIO driver latency. And I could just leave it there for all my tracking, only changing to larger buffers for mixing.

I can still use the same rate in C13 on the virtual instruments side of things. It is just the audio (vocal) tracking where I seem to be running into problems. (That is the only area where I had to deal with the crackling in earlier versions of Cubase when it was there at all.)

As for 96 kHz versus lower sample rates, I occasionally use 48 kHz if doing something like instrumental cues, but I’ve been using 96 kHz for a very long time on my mainstream recordings. I did some testing on that front a long time ago, while I was still using SONAR as my DAW (I didn’t get Cubase at all until V9.5, and only started using it as my main DAW for new projects at V10.5).

I don’t recall all the details at this point, but my early experimentation led me to feel strongly that I was getting better-sounding (more like analog???) results when using 96 kHz. I think I initially experimented with only using it for mastering, and, even there it made a difference to my ears. At that point, the computer I had couldn’t handle the increased demands of 96 kHz for the actual recording and mixing stage.

After upgrading my computer, I was able to continue experimenting with 96 kHz all the way through. Beyond my perception of better quality results, I also considered it to be a degree of future-proofing as I also tend to go back and remix older projects as my production tools and capabilities improve.

In any event, the question here is really just about changes in latency considerations for audio tracking in C12 versus C13.