Page 3 of 3

Re: Network Rendering - Performance observations

Posted: Thu Jun 12, 2014 11:00 pm
by kavorka
ASyme1 wrote:Well this distributed rendering is a very big deal for us here. Been waiting for this. We've managed to get it working somewhat. It seems to cap out at 11 cards though. Has anyone else noticed this? Anyone else get it to render with more than 11 GPU's? BTW, this is using the C4D plugin.

Thanks!

alec
They say it goes to 12.
are you including the GPU in your master?
also, are you counting any dual GPU cards?

Re: Network Rendering - Performance observations

Posted: Fri Jun 13, 2014 1:42 am
by ASyme1
I heard about the 12 limit. Why 12 by the way?

Yes, I'm counting my main computer.

Basically, I can't get more than one other computer to load. My main has 7 and my secondary has 4 more. Even if other computers are checked in the SA they don't show up in render. Just stops at 11.

Thanks
Alec

Re: Network Rendering - Performance observations

Posted: Fri Jun 13, 2014 3:00 am
by abstrax
ASyme1 wrote:I heard about the 12 limit. Why 12 by the way?

Yes, I'm counting my main computer.

Basically, I can't get more than one other computer to load. My main has 7 and my secondary has 4 more. Even if other computers are checked in the SA they don't show up in render. Just stops at 11.

Thanks
Alec
The way how network rendering is currently implemented it accepts a slave only if the total GPU count is 12 or less. It can't enable a sub-set of the GPUs a slave provides. To make it work you have to set up the second daemon to use only one GPU. That's what the GPU selection is for, when you install the daemon (see http://render.otoy.com/manuals/Standalo ... ge_id=1269 under "Daemons"). If you want to select the first and the third GPU for example, you would enter "1,3" during the daemon installation.

Re: Network Rendering - Performance observations

Posted: Fri Jun 13, 2014 4:21 am
by Rikk The Gaijin
What I don't understand is why there is a GPU limit in the network rendering? I mean, the purpose of that feature is to be able to use as many GPUs as possible, so why cut it's legs like that? Is it a marketing reason, or there is a technical limitation?

Re: Network Rendering - Performance observations

Posted: Fri Jun 13, 2014 12:48 pm
by kavorka
I am imagining it is marketing since they are releasing a version that can access any number of GPUs. I am curious about the price for that though.

Re: Network Rendering - Performance observations

Posted: Fri Jun 13, 2014 3:01 pm
by ASyme1
kavorka wrote:I am imagining it is marketing since they are releasing a version that can access any number of GPUs. I am curious about the price for that though.
Is this true? Has the network GPU limit of 12 been set for marketing reasons? Is there a way around this? I've got a room full of quad GPU workstations. I would LOVE to use as many as possible on these 6K renders I'm working on.

Thanks!

alec

Re: Network Rendering - Performance observations

Posted: Mon Jun 16, 2014 12:14 pm
by geo_n
Rikk The Gaijin wrote:What I don't understand is why there is a GPU limit in the network rendering? I mean, the purpose of that feature is to be able to use as many GPUs as possible, so why cut it's legs like that? Is it a marketing reason, or there is a technical limitation?
Since they will bring out a "pro" version and a "cloud" version its clear that the current version is "semi-pro" version :mrgreen:
I don't mind. I'm sure the pro and cloud will be way higher in price. Maybe more expensive than vray and arnold licenses.
It would be nice if otoy finally decides to implement what they originally said which was limited network nodes without having to buy extras. No standalone license for each node, no cheaper render node license. Thats what they planned before anyway and I'm pretty sure its a matter of decision about money now since network works already and its easy enough to just lock the standalone version to make it simply a render node version.
For this current version it would be nice to have extra 2 render nodes with each having 5 gpu limit. That's probably fair to both parties. Octane has a good price now but not with network rendering.