As a recent and enthousiastic user of Nebula (Nebula Pro 3 here), I now have a need to increase the processing power of my PC (Quad 6600). I would like to avoid building a new DAW with a new proc so I was wondering which is -to your experience- the most cost efficient way to run more instances of the program. Would that be using Cuda (and if so what would be your recommended fanless video card for that), or by running Nebula Server on an other (old) PC equipped with an Intel 2160 which is on the same private LAN than my main DAW?
Sorry if this has been addressed before in a previous thread..
Unfortunately Classik, to run the server well, you need another computer that is at least as powerful as the one you are using since it handles the CPU load of Nebula. I haven't used CUDA yet, but I think that the technology is getting there. But I am learning that the only way to fully maximize a Nebula experience is to have a very powerful DAW. I have a Q8400 CPU (Quad, but not an "i") and pre-Nebula this thing was a monster. But as I use Nebula for almost all of my plugin needs now, I have adjusted my work-flow to render stems down to free up CPU power..... So in the long run, your best bet is to build a more powerful DAW..... sorry.
Thanks for the swift reply. Too bad if the second PC needs to be as powerful, as I would have only needed maybe 4-5 additional instances on it, the rest would have been handled by the main DAW (I assumed I can run both Standard & Server versions in parallel to balance the load, which I guess is feasible, right?). Regarding building a new DAW, I don't have the time to do it now unfortunately, nor the budget as it would mean buying PCIe MADI+AES cards (I currently use PCI cards).
Any other user would like to comment on Cuda? What would be the processing gain?
It has been a while since I checked in with CUDA (so it's possible that this info is out of date, though I haven't seen anything to the contrary on the forums in the interim here) but...last I did it was pretty limited. Basically it only really helps on programs that have really long kernels (like Reverbs). So if you were running a whole boatload of different reverbs CUDA might help a fair amount; but for a whole boatload of consoles/tapes/EQs, not so much.
I think Server will likely be what you want (not sure if that's supported yet, though, I think it's just Local Server right now).
well im not sure if nebula is effcient in cuda 2 or been coded with the latest 4.0 toolkit, but the (femi) Telsa cards which is like 448/512 cores per processor on the GPU, but im pretty sure as cuda gets updated the graphics cards will become more like UAD'S, so for now i would recommend server, you dont need to run it on a serperate computer you can just run it on a single machine and it is still effcient
My i7 930 system, 4ghz, 12 gb ram, has an Gigabyte Nvidia 570GTX OC gamer video card in it. Slightly faster than a vanilla 570GTX, which is one of the faster cards around now - I also use it for playing games Windows 7 64 bit
I just tried the Cuda thing - installed the v4 toolkit, installed the x86 Cuda bridge, since most of my apps are 32 bit
Fired up Wavelab 6. Now the hall reverb from the CDsoundmaster producer pack 2 demo, used to take 37% of my CPU according to Nebula 3 Pro display - a single core I assume, since the actual load on the Windows is much lower.
Anyway it now reads 0-2%
More testing to come.
I had to create a 2nd folder for the Pro version, since I also have a Nebula Free install, and they conflict. And I also copied and renamed the Nebula Pro Reverb VST dll for the Cuda version to make things easier in my apps, with different settings in the master page
EDIT - does anyone know the difference between these different CUDA options in the "opt freqd" setting in the master page?
Ok tried AlexB's TFT console demo which is also super CPU hungry, and at first it didn't work with a DSP buffer setting of 512 or under, but once set to 1024 it works
CPU load also 0-2%
The reverb worked fine with a buffer of 512 and under, but not the console one
Can't tell you how many instances I can run yet, I'll test that later today. But I did try insert 2 instances in Wavelabs master FX chain and then it stopped working... until I further increased the DSP buffer size...
Last edited by CoolColJ on Mon Sep 19, 2011 6:19 pm, edited 1 time in total.