When trying to preview the De-Reverb settings after a few seconds SL says that my CPU is too slow. Well, my CPU is a i9 12900K which I assume should be capable to handle the task - at least it does the same thing flawlessly in RX 10 Advanced.
2 Likes
Yep , exactly the same here with my i7 9700 running all cores at 4.90khz . I can use the Fx racks but the Dereverb is a no go in preview
2 Likes
Hm… interesting. i9 12900 and i7 9700 also are not able to do the preview?
1 Like
Assuming it’s using the demucs model, that really needs GPU. In UVR5 I wouldn’t even attempt to work without the Nvidia doing the work, it would take too long on even my 13900K using the higher settings and FT and MDX ensembles. Some models are upto 100x faster using CUDA than the CPU.
1 Like
You’re right, Sam. I use Whisper with CUDA support to do transcriptions and the increase of speed is insane. Instead of 1,5h of processing with LARGE model I get 3,5 minutes with the same model and the same file. So I agree, GPU support is very promising.
I was a bit suprised Robin saying it isn’t enabled for GPU processing yet.
As a developer I know this problem. Deadlines.
Could be, but demucs does comes with a cpu or gpu environment out the box on Windows & MacOS. without GPU you can’t use --shifts functionality and probably means htdemucs_ft (4x slower) can’t be used either.
Maybe the limitation is the ArA support with Cubase ?
Yes, but we don’t know the internal architecture of SpectraLayers and the necessary changes to be made. Otherwise, every change needs to be tested, and every test is the cost and time. Robin wrote that GPU support was unstable in current implementation and that’s the reason for skipping it. I think they need more time to improve the implementation and for testing. In my opinion, the release of version 10 could (and should) have been delayed, but we don’t know the whole picture. Oh well…
I’ll look into that (and no, de-reverb is not using demucs).
Regarding the question of lack of GPU support at launch, it’s been answered here: SpectraLayers 10 - #63 by Robin_Lobel
Again, the plan is to add GPU support with patch 1 or 2. Even though it’s already working on my side, I really need to be sure it’s stable before releasing it, you can’t imagine the support nightmare and complains otherwise 
3 Likes
As someone who has been trying to avoid putting any GPU into a (desktop) DAW, and the only reason to do so would be to accelerate SL, I’d welcome any recommendations for an upcoming new build.
@MrSoundman the safe recommendations is to buy a NVIDIA GPU. Any NVIDIA GeForce RTX product will do wonders for acceleration, whatever your budget.
1 Like
So the hardware acceleration is only for processing? Thats good to know, I also hope other things like fft size and resolution and refinement (with multiple layers) would also be considered apart of the hardware acceleration implementation/improvements.
1 Like
I wouldn’t be concerned with Nvidia RTX GPU in an audio workstation these days. At worst all you’ll have to do is flick the driver to have an affinity if you get glitches on playback. Last two machines I’ve had with them have been fine, better than the M1 Ultra in fact for audio and stability.
The following explains the fix, pretty simply to follow if you have an issue. Use RTX Studio Drivers rather than the bleeding edge Gaming ones obviously.
Cheers, I’ve seen all those fixes before, until now my fix has been to not use a GPU. As one of the commentators says, why should normal people have to deal with this? 
Nvidia have clearly acknowledged the issue in the form of the Studio Drivers so that’s good to know. I’ll just have to wait a little longer while the virtual money folk go virtually broke, and supplies of good cards become available again.
AMD Ryzen 5900X running Reverb Reduction on some events here. On several shorter events there were no problems. On one longer even I got crackling. I never got the dialog box with the “slow CPU” error message though.
The CPU is 12-core, AM4 platform (previous generation) and it seems threading is solid and hitting all 12 cores.
Just for the record.
Nvidia don’t release Studio drivers to address any issues, be they video or audio they release Gaming drivers that Gamers use to push the performance or to address certain demands of gaming at the consequence they haven’t been real-world tested so cannot be classed as Studio level stability. Basically Gaming version is a live beta for a few % of performance gain. Gamers see advantage in higher performance because it’s part of how you win games, to have higher performance machines than rivals.
End of the day, the chances of having an audio issue is really slim these days. My only advice is some Dells do have DPC latency issues with ACPI this has nothing to do with Nvidia and DPC latency isn’t necessary a bad thing for audio despite what the internet tells you or DPCLatencyMon reports.
1 Like
Can highly advise running through Pete Browns guide, guy runs Microsoft audio dev and is a huge audio and studio geek. I’d say 80% of the windows audio advice out there is misguided, totally out of date by 20 years and simply recycled nonsense that somehow ends up in every Windows studio optimisation guide.
Unofficial Windows 10 Audio Workstation build and tweak guide - Part 1 - Windows MIDI and Music dev (microsoft.com)
2 Likes
So , is this a Cpu or a Gpu problem , will the Gpu patch you mention cure this problem , im runnig a N Geforce Gtx 1650 card (latest version 2020 )that never breaks into a sweat (less than 20%) with 3 monitors , id hate to think with this machine running so well id have to upgrade the graphics card (again ) just to get the Dereverb preview going ?