|
|
|
|
View Forum Thread
Anyone can view the forums, but you need to log in in order to post messages.
> Forum Home >
Feature Requests >
Separate priorities per machine for GPU renderers
|
|
Tue, 04/Feb/2014 6:31 AM |
Marc
11 Posts
|
Hey everyone!
Most of our machines are used for both, CPU- and GPU rendering and we rely heavily on the "variable packet size" feature, which is calculated per machine via the machine priority setting.
The problem we have at the moment is that there is only a single priority value per machine, which does not really reflect the various capabilities of that machine. For example, some of our machines have much CPU Power, but limited to none GPU Power and vice versa.
So ideally, there would be separate priority values per machine for both, CPU and GPU capabilities, to optimize frame distribution across a non-homogeneous machine pool for all kind of render engines.
Any thoughts?
|
| |
|
|
|
Thu, 06/Feb/2014 8:39 PM |
Robin
1138 Posts
|
Hi
That's an interesting idea, and I can see how it would be useful for some scenarios. I'll do some thinking on a way to accomplish this. One idea I am working on already is separate job queues, and maybe this could be tied in to that system. An alternate idea would be being able to set engine priority separately for different Pools or different Products.
Thanks,
-robin
|
| |
|
|
|
Thu, 13/Feb/2014 6:56 AM |
Marc
11 Posts
|
Hi Robin,
separate job queues sound interesting, though I have troubles visualizing the big picture here. Would they be able to run in parallel? e.g. "GPU" queue jobs don't block the machine for jobs in the "CPU" queue and vice versa. That would be a real game changer, because it reflects exactly the available resources per machine.
The alternate ideas sound equally good for us, but might be a bit too much for customers with lots of pools/active products?
Thanks for considering
|
| |
|
|
|
|
. |
