V-ray Distributed rendering which PC to use as Host?

[Digital Magic]

I have 4 PCs connected for distributed renders. One of them (Threadripper with 128GB RAM and 32 cores) is much more powerful than the others which are i7 with 64GB, 64GB and 32GB RAM.
Right now I am launching 3ds Max with the Threadripper and checked Use local host to include it in the render. The other 3 PCs are running Vray spawner as render nodes.
Is this the correct way to do this? I am wondering if the most powerful PC should be hosting it or should be a render slave? One of the other 3 PCs has full license of Max if needed to launch from it instead.

It seems like there may not be a speed incease in the Threadripper rendering by itself vs the Threadripper rendering with 2 or 3 nodes helping.
3 nodes seem no faster than 2 nodes.
I have yet to do a long test to compare times but I am questioning if the 3 nodes actually reduce render time or not....

[Karol.Osinski]

It seems you are doing everything correctly but I highly doubt that other machines would not reduce the render time.

Are you able to "ping" the machines from the Threadripper one? What I mean is you can open DR Check Tool and do vraydr_check -host=(IP number of the machine) -port=20207, it should display the current status of slave machine and number of cores it has.
Another way to check if all machines are rendering simultaneously is to check the names of Buckets, which you should use for DR rendering instead of Progressive mode by the way.

[Digital Magic]

Thank you very much.

I did not understand how to do this "Are you able to "ping" the machines from the Threadripper one? What I mean is you can open DR Check Tool and do vraydr_check -host=(IP number of the machine) -port=20207, it should display the current status of slave machine and number of cores it has."

But I did some tests on a short TGA sequence. Render times listed per frame

Threadripper only
Frame 0 - 2:54
Frame 1 - 2:45
Frame 2 - 2:45

GTX desktop i7
Frame 0 - 6:06
Frame 1 - 6:05
Frame 2 - 6:05

Threadripper and GTX desktop i7 DR
Frame 0 - 2:45
Frame 1 - 2:45
Frame 2 - 2:42

All 4 PCs DR
Frame 0 - 2:48
Frame 1 - 2:35
Frame 2 - 2:38

All 4 PCs DR (set to bucket instead of progressive)
Frame 0 - 2:04
Frame 1 - 1:59
Frame 2 - 1:57

Threadripper only (set to bucket instead of progressive)
Frame 0 - 2:36
Frame 1 - 2:36
Frame 2 - 2:36

Does this sound about right? Seems you found a major thing by telling me to switch to bucket. That makes a difference

[Karol.Osinski]

DR Check Tool is a little command line tool used to check the status of individual slave machines. Buckets in distributed rendering utilize the network better thus render faster.

I still wonder if all of your i7 are utilized and if yes, after what after how much time since the frame starts they arrive. In some of our scenes it might take a minute or two for the slave machine to receive all assets and start to render. There is a tick box "Use cached assets", it is under "Transfer missing assets", perhaps you can try it and this way slave machine won't have to wait for the assets at each frame.

Although sometimes it can cause dark buckets and idk why.

[Karol.Osinski]

[Digital Magic]

Wow thank you very much for the help