8 GPU system Benchmarking

[joe.renderstream]

Hello All,

In about a week or so I'll be getting an 8 GPU system for internal testing and benchmarking here at RenderStream. I plan to benchmark most of the main GPU renderers including octane. I plan to see how the renderer performs in an 8 card configuration, as well as looking at how well it scales from 1-8 GPUs.

Also, I'm hoping to get some input from you guys on what else you'd like to see tested. Obviously I have access to the demo scene files but I'd really like to throw something more substantial at this machine. If someone has a type of scene in mind that they would like to see tested, or a scene you can share, please let me know.

The machine can hold 8 cards. Listed below are the cards I'll have for testing in various configurations:

8 GTX 580
2 GTX 580 Phantom
2 GTX 590
8 HD6970
8 Tesla C2070

Thanks!

[radiance]

I suggest you use our new demo version which is coming out later this week, (after the licensed version is released and tested for a few days),
It will also include a more recent demoscene suite with more scenes to test with.

Radiance

[suhail_spa]

8 Tesla C2070

oh man!
i reeeaally look forward to those benchmarks...

btw... can you please let us know some specs/pics of the 8 GPU system...it would be a good info..(of course if its not confidential stuff )


EDIT:
just found the info i needed...
http://blog.renderstream.com/?p=959

[joe.renderstream]

thanks for the info. i'll plan to benchmark with the demo suite and post back when i get some results. looking forward to it as well.

more info on these systems can be found here http://www.renderstream.com/VDAC.pdf

[Maryus3D]

Octane Render doesn't work with HD6970 . I'm also interested about the tests.

[gpu-renderer]

oh my god... please release the benchmarks.... drooll.... sometime next year I will get one of these... that will soooooooooo cut down on rendering time.... by any chance can this be linked up for batch rendering across multiple tower / rankmount units...

I'm not sure if octane would ever use network clustering in the future if it does... its a hollywood production tool... awsome stuff.. I will have 16 of those badboys please lol... 65536 cuda cores 3gb gpu ram or if its possible 1 x nuclear powerplant on standby 8 x gtx 590's x 16 units (1.5gb is just right) 128k cuda cores .... i am normal i am not a megalomaniac mmmmk!

near realtime rendering :0)

please send the benchmarks and a video of such a setup... the pros out there will line up for them!

[abstrax]

Sorry, but we had to take the multi-GPU functionality out of the current demo version. We did this to reduce the confusion about the different CUDA versions. This allowed us to release CUDA 3.2 builds only. When the new multi-GPU framework is set up and runs properly on CUDA 3.2 or 4, we will release another demo with multi-GPU enabled again.

Cheers,
Marcus