I have to admit, I am not entirely sure how Network Render / ORC / RNDR / NVLink all works.
Which are the current paid-for-services, and could they be combined with your own rig?
Like, could you have a (1) GPU rig yourself + (2) slave + (3) paid-for-render service, all rendering your scene at the same time?
What would be the most powerful theoretic structure that can be arranged to render?
Combining Rendering Modules?
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
Good day, Notius,
few explanation of those things You mentioned:
ORC as OctaneRender cloud is a paid service that You can upload .orbx file and render things out. It relies on AWS servers at the moment and is probably the most powerful render farm officially supported by OTOY.
Network rendering allows to combine multiple machines and render things out locally. the limit depends on the version of Your licence (and also for bigger network You need to have 10G).
As for NVLINK, officially for now it's possible to nvlink two cards, but Nvidia seems to be making quad GPU bridges for DGX stations - 70k*$ builds with quad GPUs for deep learning. Also they have NVSWITCH, that allows to interconnect 16 GPUs, but that's a technology inside their 300-400k machine.
So realistically the fastest way to render now is ORC. But RNDR is coming and if it will be even partly as good as OTOY expects.. that might be the largest render farm on this planet.
The interesting thing about RNDR and one of the reasons that personally already built a dedicated machine and will throw it into this service 24/7, is that You can "rent it" storing render power in form of tokens and then reuse them when You actually need. So unlike using single machine when You need and being limited to it's render power (and then leave it dusting until next time) we'll be able to use all it has x multiple times whenever we need. It's like render-power-bank.
Sinc.
tom
few explanation of those things You mentioned:
ORC as OctaneRender cloud is a paid service that You can upload .orbx file and render things out. It relies on AWS servers at the moment and is probably the most powerful render farm officially supported by OTOY.
Network rendering allows to combine multiple machines and render things out locally. the limit depends on the version of Your licence (and also for bigger network You need to have 10G).
As for NVLINK, officially for now it's possible to nvlink two cards, but Nvidia seems to be making quad GPU bridges for DGX stations - 70k*$ builds with quad GPUs for deep learning. Also they have NVSWITCH, that allows to interconnect 16 GPUs, but that's a technology inside their 300-400k machine.
So realistically the fastest way to render now is ORC. But RNDR is coming and if it will be even partly as good as OTOY expects.. that might be the largest render farm on this planet.
The interesting thing about RNDR and one of the reasons that personally already built a dedicated machine and will throw it into this service 24/7, is that You can "rent it" storing render power in form of tokens and then reuse them when You actually need. So unlike using single machine when You need and being limited to it's render power (and then leave it dusting until next time) we'll be able to use all it has x multiple times whenever we need. It's like render-power-bank.
Sinc.
tom