Page 2 of 3
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 1:12 am
by pixym
Ah Yoon, at last you are able to test octane with your 4 GPUs monster!
Very thanks to Rad. and of course to you.
I am still waiting for my gtx 480, but I am very tempted to buy another one! (Sorry If I tend to be out of topic
)
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 1:51 am
by YoonKyung
0.5 GTX295 = 1.48 msps (+1.48, 1.00x, 100%)
1.0 GTX295 = 2.77 msps (+1.29, 1.87x, 93.5%)
1.5 GTX295 = 4.14 msps (+1.37, 2.79x, 93.0%)
2.0 GTX295 = 5.41 msps (+1.27, 3.65x, 91.2%)
2.5 GTX295 = 6.71 msps (+1.30, 4.53x, 90.6%)
3.0 GTX295 = 7.81 msps (+1.10, 5.27x, 87.8%)
3.5 GTX295 = 8.61 msps (+0.80, 5.81x, 83.0%)
4.0 GTX295 = 9.33 msps (+0.72, 6.30x, 78.7%)
Very nice scalability !!
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 2:14 am
by YoonKyung
pixym wrote:Ah Yoon, at last you are able to test octane with your 4 GPUs monster!
Very thanks to Rad. and of course to you.
I am still waiting for my gtx 480, but I am very tempted to buy another one! (Sorry If I tend to be out of topic
)
I have uninstalled one of them, because of the heat.
Now, it is calm and warm(not hot) and fast.
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 2:17 am
by pixym
Thanks,
I have noticed 3 instead of 4, but did not dare to ask why.
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 3:37 am
by YoonKyung
Two bugs regarding video buffers,
1. Connections between video buffers are not continuous.
2. The first video buffer (from the left) is dimmed.
And, when I switch between 'daylight' and 'texture environment', Octane is crashed.
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 5:58 am
by YoonKyung
'Motion blur' does not work except for the first video buffer.
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 1:36 pm
by radiance
Hi guys,
thanks for testing.
AS said in the first post, mixing slow with fast GPUs will drag the fast gpu down.
we still need to divide the workload based on the speed of each GPU.
We also need to parallelize the tonemapping, to make the speedup even more.
I'm already quite happy with these result, especially the ones by yoonkyiung (or how do you spell it again ?!)
We'll continue working on it and we'll get something more solid and with better performance out soon.
Thanks,
Radiance
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 1:41 pm
by radiance
andrian wrote:There is something weird, or I do something wrong, but seems I get higher samples with first single GPU than with both...
And what about memory usage?? It's double in multi GPU mode. I know that have to be loaded in both GPU-s.. but that way seems i have left less memory than with one GPU scenario. Also some weird line in the middle, separating both GPU calculations in multi GPU mode. Renders was done with pathtracing kernel, scene is saved in Octane beta 2.1 non multi GPU version. No material applied to any object ( uses the kilaD scipt to export ).
Hey,
This is indeed very strange.
we'll look into this and we might ask you to test again with another executable soon.
Radiance
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 7:47 pm
by acc24ex
radiance wrote:andrian wrote:There is something weird, or I do something wrong, but seems I get higher samples with first single GPU than with both...
And what about memory usage?? It's double in multi GPU mode. I know that have to be loaded in both GPU-s.. but that way seems i have left less memory than with one GPU scenario. Also some weird line in the middle, separating both GPU calculations in multi GPU mode. Renders was done with pathtracing kernel, scene is saved in Octane beta 2.1 non multi GPU version. No material applied to any object ( uses the kilaD scipt to export ).
Hey,
This is indeed very strange.
we'll look into this and we might ask you to test again with another executable soon.
Radiance
As I said before, my system behaves in the same way as his.. 8800 gtx and 260 gtx running together is slower than just using 260 gtx,
using latest nvidia drivers on win7 64 bit
Re: Intermediate multi-gpu test version for windows x64
Posted: Sat May 01, 2010 9:16 pm
by radiance
acc24ex wrote:radiance wrote:andrian wrote:There is something weird, or I do something wrong, but seems I get higher samples with first single GPU than with both...
And what about memory usage?? It's double in multi GPU mode. I know that have to be loaded in both GPU-s.. but that way seems i have left less memory than with one GPU scenario. Also some weird line in the middle, separating both GPU calculations in multi GPU mode. Renders was done with pathtracing kernel, scene is saved in Octane beta 2.1 non multi GPU version. No material applied to any object ( uses the kilaD scipt to export ).
Hey,
This is indeed very strange.
we'll look into this and we might ask you to test again with another executable soon.
Radiance
As I said before, my system behaves in the same way as his.. 8800 gtx and 260 gtx running together is slower than just using 260 gtx,
using latest nvidia drivers on win7 64 bit
that would be normal.
as i explained, this version does'nt yet take into account the power of the gpus,
so if you use a slow and a fast card together, they will both get half the work,
and the GTX260 will have to wait until the slow 8800 is finished.
our next test version will be a lot optimized so you should try again with that one.
Radiance