OTOY Forums

so if i have a couple of boxes all on the same gigabit LAN with CUDA GPUs in them
could octanerender see and use them over the network, providing that octanerender is on the other machines as well?

currently as far as I'm aware, all GPU's have to be connected directly to the mainboard of the machine ( either through an extend case, and connector, or in the main machine directly ). Network rendering with GPU's in Octane is not currently available ( I have no idea if it will be in the future either ).

It's currentlty not possible. The redesign of our CUDA framework has "high-latency" devices in mind though and hopefully allows us to do these things in the future.

Cheers,
Marcus

oh goodie, goodie, goodie
i would love to throw all of my extra GPUs in a box and really have fun makin movies
so when will multiple GPUs (in the same box) be back in use?
ive read alot of things on here that says it will only use one gpu atm
any idea on when that issue will be fixed?

disturbed13 wrote:oh goodie, goodie, goodie
i would love to throw all of my extra GPUs in a box and really have fun makin movies
so when will multiple GPUs (in the same box) be back in use?
ive read alot of things on here that says it will only use one gpu atm
any idea on when that issue will be fixed?

We currently provide two builds per platform: A CUDA 3.2 build and a CUDA 3.0 build. The CUDA 3.0 builds use all GPUs correctly.

Cheers,
Marcus

abstrax wrote:It's currentlty not possible. The redesign of our CUDA framework has "high-latency" devices in mind though and hopefully allows us to do these things in the future.

Cheers,
Marcus

i think you would be better off if you just 'served' parts of the animation to the networked GPUs
by using some sort of GPU rating system, or a way to measure the GPUs speed, RAM, and bus transfer speed to rate it with an appropriate number
then you must also consider the latency due to the network
WiFi (which is a joke), Cat5 100Mb Lan, Cat5e 1Gb lan, fiber optic lan, ect
the network speed would also have a rating system
and then the two should be averaged to provide an accurate number of how much that GPU can do over the given network
i really dont see a use for it to try and help a single frame render, but mainly in animations
just my thoughts

Single frame renders are indeed silly with octane to do over a network,
That is why we have not added the functionaty to date.

Note that with out new plugin products, like the 3ds max plugin product and the others coming soon, you can use standard network queue software compatible with your host app to render frames of animations on a network on computers with CUDA GPUs...

With octanerender for 3ds max, you can use the backburner queue software for example...

Radiance