I can't find a better forum channel to post this in, but by all means move it where it's appropriate. I want to give a touch of background before going right to the question. I'm a tech guy who is trying to push through the day job 9 to 5, and has been developing small businesses to see it happen. I paired up with an artist last year and we had a render farm hooked into 3ds with vray. I had between 200-400 xeon cores for him to work with (lots of selling/buying going on) and it was overall good enough to do very high quality renderings-- 8K stills, 360 frames animation in 1080 in 6-12hrs, 360 stereoscopic frames in 30-60hrs. I want to match that performance in octane using GPU.
The farm is gone, we went our separate ways for a while, and I'm reaching back out to him to continue where we left off. I just dropped $1,500 on upgrades to his rig, but think this will only get us ~50% of where we were. He has a 4770, so that's good enough, and I got him a quad sli (accepts 4 dual slot gfx at pcie 3.0 4x) mobo, 3 GTX 980 ti, and he has a 1050ti he can use as primary monitor low priority. We'll get results by end of week (parts will arrive hopefully by Friday or Sat), but I'm already trying to find a solution that fits our needs going forward. If we wind up swamped, then it'll make sense to get another rig and use "the cloud" in the meantime. I need more information to know which way is best on a second rig.
He uses 3ds, so we'll be swapping out vray for octane plugin here. I am concerned about using a slave machine to help render with the plugin. Here are my questions:
1) Is the 3ds plugin working without issue for slave/network rendering?
2) Is there any network bandwidth/latency concerns when using Octane 3 in this setup?
I ask the second question because I have heard some 10-30% reduction can be had when using x1 PCIe (splitters/risers) in Octane 3 vs 2. Will a 1GBe connection see this same limitation, or is the fundamentals different here?
3) What kind of multiplier can I expect from using an identical slave node with Octane 3 on 3ds plugin?
If I double up GPU within same machine, assuming no bandwidth PCIe bottleneck, I should get roughly 95-99% speedup based on the benchmarks I've seen. Does this hold true for a slave as well? Would I see roughly 2x or is it closer to say 1.8 or 1.5x performance?
Overall we're looking for a solution that meets time deadlines without sacrificing quality, and within his lifestyle constraints. He is with family in a house that has not the best internet, so using a remote farm would not be ideal. I think his upload is 1mbps and download closer to 8mbps, which makes for slow exports, difficulty using RDP (latency is around 50ms to my residence some 30miles away, for example). I want him to be able to tap two 15A power circuits at his residence, and have a good workflow, with workable render times.
If we can see a near doubling up of performance with a slave node without issue in 3ds, then I think we will try to make due for 6 months while clientele builds up, then get 4 gtx 2080 next Spring for a second rig.