Hi! I bought a GTX650ti to get a few more reverbs working. But i don't really set more CPU usage free.
I got a session with a lot of Nebula EQs an 2 Nebula Reverbs and I can't open a 3rd. switching the Reverbs to Cuda maybe reduces the CPU usage 5% (maybe nothing) and I still can't open a 3rd Reverb, but turning the Reverbs off Reduces it 25%.
Is it right that using Cuda isn't the same as turning the plugin off, is there still a lot CPU needed?
Or am i doing something wrong? I'm working 64bit all the way (win7x64/Cubase7_64/Nebula3_64) 4.0.17 CUDA Toolkit + CudaBridge64 1.1.038 BUT the Cuda 4 drivers don't recognize my card, so I used the new driver an the old cuda 4.0.17 toolkit
When I run TestCuda64.exe it finds the card and says "Device 0 successfully started.", I get no flashing exclamation mark like in the 32 bit version.
Ok, I tested a lot with the DSP buffer size and it's great!! The benefit for the Reverbs is just what I was searching for!
BUT I get the same results increasing the DSP Buffer size on the CPU based Nebula. The Plugin Percent Meter says Cuda is more effective, but in the Windows task manager it's hard to see ANY difference. I Expected to find what I'm used from the UAD = the Card is doing the work and nearly no CPU usage at all.
I've bought the Nvidia Card just for using Cuda with Nebula, so i think I'm sending it back...
Or can anyone tell me whats the really impressive feature in using Cuda?
why did you quit using CUDA?? The Server Version is the better idea!?!?
Actually, hadn't really needed it. My graphics card, GeForce 8600GT 512MB could only handle a few NEB's, and really, using the Server version [on a single machine], along with switching to REAPER with its freeze function, has improved my workflow. The 'Sever' version can share memory [which my current XP-32 bit appreciates]. Although I have 2 multi-tracked projects to mix, Mastering is the focus.
All said, I'm still going to make sure CUDA is ready to go, if ONLY for the reverbs that are coming out.
jpchartrand wrote:I have Neb 3 Pro. Can I use CUDA? How do you guys use your graphics card to load Nebula?? I don't get it.
I'm actually not familiar with CUDA and how it works but I'd love to learn. Can someone give a quick rundown on how to get this done?
The really quick rundown is: you install a particular CUDA SDK, enable CUDA in Nebula, and it offloads some limited processing to the processor on your graphics card. IIRC it only works on 1 kernel programs (stuff like reverbs), I believe that was the case when I stopped using it (and why I stopped, it wasn't affecting the overwhelming majority of the Nebula programs I use frequently).
Ngarjuna break it down for you. The concept is pretty simple.You use "your graphic card power" to release "your computer power".So you can use a couple (not a lot) more Nebula instances on top of what you can load with your actual CPU/RAM.