Does SL 11 make use of Nvidia graphic cards?
Makes it sence to buy a computer with such an graphic card or is only the cpu power important for SL?
SL 11 supports GPUs for AI processing
1 Like
SL11 can even use integrated GPU, and not just discrete graphics cards, and not just NVIDIA.
1 Like
@Detgyver For a detailed discussion of some of the subtleties of using the GPU for processing in SL11 see the following topic: Unmix song with extreme settings.... slow Zzzzzz Currently, for the best performance when using the GPU for unmixing (especially Unmix Song in Extreme mode), it would seem that a graphics card with at least 8GB VRAM is recommended (and if you have the choice NVIDIA is probably preferable).
1 Like
Thank you all for the quick help.
It seems that I need a new Computer with a good graphic card on board (gaming pc here I come 
Best
Det
One last question here:
Can someone explain where I find this configuration?
I guess you are using the german interface, in that case it is
Bearbeiten → Optionen
and then here
1 Like
Many people in the source separation community both training and separating use 2nd hand Nvidia 1080 Ti (11GB version) . $150-175. They really stand up well still and cheap. I know someone running 6 servers in US and 6 in Norway running these 24/7 for source separation service, he swears by them.
1 Like
Hi @JuergenP , yes you are right, I use a german system.
Thanks for the hint - I know now where to make adjustments.
I will configure a new PC in the next days because my laptop does not longer match my needs. I think it will be a Nvidia 4070 that will empower SpectraLayers in the future
1 Like
I got the 4070, and this latest update significantly improved the times to unmix a song. A 4 min song, using the extreme settings took around 2 minutes to complete. It was taking 3-4 times longer with 11.0.
2 Likes
FWIW, the PALIT GeForce RTX 3050 6GB KalmX is running 11.0.10 quite well here.
Not a gamer card by any means, but no fans, no special power connectors and only 70W power consumption.
3 Likes
I have SpectraLayers Pro 11 and do have a NVIDIA Quadro RTX 3000 Max-Q , 6GB memory size, GDDR6, but when I select it in AI processing device, it says : Less than 8GB of VRAM.
What should I do? Here is my system:
Asus ProArt StudioBook Pro 17 (W700G3T), Intel® Xeon® E-2276M processor; 32GB Ram, 2TB (PCIe® NVMe Gen 3 x4, M.2 SSD x 2); NVIDIA® Quadro® RTX 3000 Max-Q 6GB GDDR6 VRAM; Sabrent Rocket XTRM-Q Thunderbolt 250G External SSD; Steinberg UR-RT 4 Audio Interface; Cubase Pro 13 Pro; Windows Pro 11; Dorico Pro 5; Spectralayers Pro 11; Spectrasonics Keyscape; Kawai MP-10; Nektar Panorama T4; Roland GR-1.
Thank you for your help.
Joss
@Joss_Scarlatti You don’t really have to do anything. That is a warning message that performance may not be optimal if you are using GPU acceleration for unmixing (Unmix Song for example). It doesn’t necessarily mean that it won’t work, it means that performance may be slower. 6GB VRAM should be OK as witnessed by @MrSoundman above. For more info see also: Unmix song with extreme settings.... slow Zzzzzz - #56 by Magnus_N
Odd, I didn’t get that warning.
It’s on the GUI next to the chosen AI Processing Device.
Yes, give it a try, 6GB may work - it’s just not officially supported, but it’s worth giving it a try, you might still have some nice acceleration from it 
1 Like
Merci Robin! Content de ne pas avoir à remplacer mon coûteux portable à 4 000 boules! SpectraLayers est fantastique!
3 Likes
Must be new in 11.0.10 … I didn’t see that when I installed 11.0.0, just did an upgrade from SL10.
I have seen very significant acceleration, for example a 7:30 song in 24-bit 192kHz, everything selected, set to Extreme, took 1hr 12min to complete in SL11.0.0, and only a little over 10min in SL11.0.10. Task Manager showed it using about 5.8 GB video RAM.
This is more than I will ever need, and has the advantages (for me) of no fan, no special connectors and relatively little power consumption. I think it’s actually a laptop chip on this card, with a massive heat sink.
1 Like
11.0.10 now uses my RTX2060 Super better, but I can see in Taskmanager only up to 36%. Why not more?
Might be due to memory transfers between successive models, as well as a few remaining CPU data processing happening between models. Could potentially be optimized, but not high priority for now, maybe patch 3.
2 Likes