With the announcement of Octane 3, I'm building a new machine optimized for Octane. I know load times are small in both 16x and 8x configurations but I render animations with motion blur and the load times can have a significant effect on decent quality DL previews when the scene must be resent between each frame. I want a build that can push out a lot of preview frames fast. Going from 20 sec send times to 10 sec sent times will make a big difference in this situation. I'm wondering what the differences will be between having one machine with 7 GPUs @ 8x each and having 1 machine with 4 GPUs @ 16x each and just adding a node for the rest of the GPUs. The node would allow frames to start rendering fast and the node can jump in when it's ready.
I have already ordered an ASUS X99 E Workstation Mobo w/ a Xeon 1650
Has anyone run some of their own or does anyone know of information on improving Octane load times?
New Build: 16x vs Load Times
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
What you call "load times" is probably mostly the geometry compilation time, which is dependent of course on the amount of geometry and the speed of the CPU. The actual upload to a GPU usually takes only a fraction of a second.garytyler wrote:With the announcement of Octane 3, I'm building a new machine optimized for Octane. I know load times are small in both 16x and 8x configurations but I render animations with motion blur and the load times can have a significant effect on decent quality DL previews when the scene must be resent between each frame. I want a build that can push out a lot of preview frames fast. Going from 20 sec send times to 10 sec sent times will make a big difference in this situation. I'm wondering what the differences will be between having one machine with 7 GPUs @ 8x each and having 1 machine with 4 GPUs @ 16x each and just adding a node for the rest of the GPUs. The node would allow frames to start rendering fast and the node can jump in when it's ready.
I have already ordered an ASUS X99 E Workstation Mobo w/ a Xeon 1650
Has anyone run some of their own or does anyone know of information on improving Octane load times?
In theory there is no difference between theory and practice. In practice there is. - Yogi Berra
That makes sense and the delay I experience is dependent on the size and geometry of a scene. I am trying to figure out how to speed it up and I have read about these times being improved by going from 4x lanes to 8x lanes which would suggest the delay is related to PCIe lanes. It would be nice to how PCIe lanes effect Octane but I guess that is going to vary so much between systems. Thanks for the info!
Win 8.1 / C4D R16 / Maya '16 / 5x GTX 980 Ti
As I said, PCIe speed is not very critical at the moment. If you want to shorten the frame update time, you probably want to optimize your geometry (by using instances for example) and maybe get a faster CPU. PCIe speed may be important for out-of-core textures, but we haven't done any proper testing so we can't say anything more specific.garytyler wrote:That makes sense and the delay I experience is dependent on the size and geometry of a scene. I am trying to figure out how to speed it up and I have read about these times being improved by going from 4x lanes to 8x lanes which would suggest the delay is related to PCIe lanes. It would be nice to how PCIe lanes effect Octane but I guess that is going to vary so much between systems. Thanks for the info!
In theory there is no difference between theory and practice. In practice there is. - Yogi Berra