geforce gtx 590

[Refracty]

JimStar is talking about the output resolution wich will also affect the VRAM consumption. Larger Renderings will consume more GPU RAM.

It is not only about Polycounts but also (especially) about textures. Large textures can consume a lot of RAM because they will be 'unpacked' and uncompressed in Octane.

So I would give you the advice to set up a sample scene with textures and then think about what larger scale projects you might make in the future.
In your position I would go with one GTX 590 + one GTX 580 (3GB) instead of two GTX 590 but that is my personal choice based on the projects I make.
For motion graphics work 1.5 GB could be much more then enough.

[regionfive]

Hi guys,

for our recent project, we bought the ASUS P8Z68 Pro Mainboard with 2 x 590 GTX. Octane sees the 4 cores, Sli is switched off. and for a distinct scene, I render with 50Ms/s.
That sounded good for me at first, but then I compared the same scene on a 5 year old board with one 580 GTX... and: 50Msamples/sec!!
Then, I experimented and turned 3 Cores of the cards off. So octane rendered the same scene with one 590 core. And see there: also 50 Msamples.
How can that be? I tryed all Nvidia configurations like Multi GPU on/off. Always the same result. then I saw: Nvidia Control Centrer says i should plug the 590 into a high speed PCIe.
But as I understand, the ASUS P8Z68 is one of the most up to date boards with quad-sli support. i am a little helpless here. Is the Bus performance a crucial aspect for render performance?
Any ideas?

greets
René

btw: octane is killah

[Refracty]

I am also thinking about adding 590s to my system, so it would be very interesting to clarify this issue.
Is there anybody else arround who uses multiple GTX 590s? What speed increase do you get?

[bepeg4d]

hi regionfive, you must use the latest version of octane build for cuda 4.0, like 1.248c or 1.249, in order to get advantage from multigpu with a 100% scalability.
i've recently change my 2x 460 2GB with 2x 590 and the difference is amazing
here is some food for the zombies out there:
1GPU- ms/sec= 4.31- time 7':17"
2GPU- ms/sec= 8.62 - time 3':38"
3GPU- ms/sec= 12.88 - time 2':25"
4GPU- ms/sec= 17.15 - time 1':50"

ciao beppe

[Refracty]

Hi bepeg4d,
thank you for feeding us
I have just send you a message.
Now we have more clarity about this issue.
Cheers
Refracty

[nehale]

2x 590 and the difference is amazing
here is some food for the zombies out there:
1GPU- ms/sec= 4.31- time 7':17"
2GPU- ms/sec= 8.62 - time 3':38"
3GPU- ms/sec= 12.88 - time 2':25"
4GPU- ms/sec= 17.15 - time 1':50"

ciao beppe

are theses pmc/path tracing render times? wow!

[bepeg4d]

are theses pmc/path tracing render times? wow!

3k poly model in PT 16 at 1280x720
ciao beppe

[Refracty]

Hello bepeg4d,

does 8x PCIe or 16x PCI express make a lot of difference in render time?
Thanks
R

[ROUBAL]

The difference is imho in loading time of the obj or ocs, but once the scene is loaded in the GPU memory, the rendering fps should be equal...

[bepeg4d]

it would change only the time after voxelization and before the render start but i must say that i've noticed an increment of time in the upload into 4 gpu, it seems like it's one at a time, so more gpu more time to wait
ciao beppe