Hi, just got my setup ready. 1 Cubix box, with 1 redRocket and 3 GTX570 2.5GB cards.
On a MacPro 24Cores(HT). OS X.7.4, FCPX10.0.5
Running the optimize performance I get:
Radius = 0 => 21.7 fps (04 CPU cores, and 2 x GTX 570)
Radius = 1 => 11.6 fps (10 CPU cores, and 1 x GTX 570)
Radius = 2 => 09.9 fps (10 CPU cores, and 1 x GTX 570)
Radius = 3 => 08.4 fps (07 CPU cores, and 3 x GTX 570)
Radius = 4 => 07.5 fps (12 CPU cores, and 3 x GTX 570)
Radius = 5 => 06.5 fps (15 CPU cores, and 3 x GTX 570)
So does the CUDA acceleration scale well? I don't know, what should we expect...
Looking at the graph of it, shows a big dump from 0 to 1 radius, and therefrom, only a low decrease...
So how important is the radius parameter, is it different on very different shots?
Codecnoise VS. old, scruffy VHS noise i.e. ???
cheers Helge
NV3.2 and multiple CUDA cards, results!
Please kindly post the complete log from Optimize for some typical parameters like: 1920x1080p, rad=1, 32-bit.
If I understand correctly, when running in an expander box, all the cards are still connected through a single PCIe slot of the Mac, and that may be the bottleneck in this case. If they were individually connected to separate PCIe slots of the Mac, without the expander, perhaps that would allow to achieve better speed with 2 and 3 GPUs. You may want to try that for a test.
Thank you,
Vlad
If I understand correctly, when running in an expander box, all the cards are still connected through a single PCIe slot of the Mac, and that may be the bottleneck in this case. If they were individually connected to separate PCIe slots of the Mac, without the expander, perhaps that would allow to achieve better speed with 2 and 3 GPUs. You may want to try that for a test.
Thank you,
Vlad