Vray 3.6 numa

[savvadesogle]

Hello, Can anybody help me to setting distributed render on client machine with 4xNUMA?

I have 4 x NUMA on client and 4 x NUMA on server, like this:
Logical Processor to Socket Map:
********************************-------------------------------- Socket 0
--------------------------------******************************** Socket 1

Logical Processor to NUMA Node Map:
****************------------------------------------------------ NUMA Node 0
----------------****************-------------------------------- NUMA Node 1
--------------------------------****************---------------- NUMA Node 2
------------------------------------------------**************** NUMA Node 3

How can i start vray render from 3ds max window (vray render setup) without render in this 3ds max?
Don't understand what i am describing? Well..

If i start distributed render from 3rd PC (Intel 3770K, 8 threads) to 1st PC (spawnerXXX.exe 4 x IP 20204-20207) + 2nd PC(spawnerXXX.exe 4 x IP 20204-20207) - they are used 100% of theirs CPUS.
If I start distributed render from 1st PC (2 x Xeon 64 Theads) to 2nd PC (AMD 32 threads) - 2nd PC - used 100%, but 1st PC used sometimes 60%, sometimes 40%... BAD Utilization.
I know about ...
I found information on the site of autodesk about the solution of such a problem:
- virtual machines
- disabling virtualization (HT)
There was still no official way to disable 2-3 cores.
But I want to try my way: setup render scene in 3ds max -> click button 'Render' - > But the VRAY will not start as usual, but will send the job to the distribution render for the same machine(-s) without Backburner (I want both machines to work on one picture. Or can this task be carried out by a Backburner?)

Or is this a stupid idea?

[savvadesogle]

Admins, please move this topic to https://forums.chaosgroup.com/forum/...s-max-problems channal. Sorry.

[mihail_kuzev]

Hello,

"If I start distributed render from 1st PC (2 x Xeon 64 Theads) to 2nd PC (AMD 32 threads) - 2nd PC - used 100%, but 1st PC used sometimes 60%, sometimes 40%... BAD Utilization."

In case you want to start the distributed rendering from a machine with NUMA nodes (or if you want to render locally on it) then you will need to exclude the local machine from the rendering (remove the check from "Use local host)":
https://docs.chaosgroup.com/display/...deringSettings

Afterwards you will need to point the machine to use the 4x Spawners that are running locally (you can use the 127.0.0.1 address).

Similar configuration could be used with Backburner - you can send the job to one of the machines and set it to start distributed rendering in the same way.