Questions about network processing

[bendead]

Hi,

I am currently working in Ivory Coast for a drone training academy. We also do Photoscan training.

I have a few question about the network possessing, because I will soon receive a very good laptop 9900k with 2 GTX 1080, and would like to test it with our 9980XE with double 1080Ti.


- Will it share the load, based on the machines performances, like 30% on an old computer and 70% on a new system? can I configure that?

- If on normal processing "single computer" some treatment like dense cloud or mesh takes 40go of RAM for a 128GO computer, will it use 20-20 on both computer in the network?

- If the answer to the questions above is positive, does that make any sense to go to AMD epyc or Intel Xeon with a network setup?


I just would like to expend my knowledge a bit on that


Thanks

Benjamin

[bendead]

Hi again,

I can see that my topic is not really successful.

I just would like a single answer on this question :

If on normal processing "single computer" some treatment like dense cloud or mesh takes 40go of RAM for a 128GO computer, will it use 20-20 on both computer in the network?

Thanks

Benjamin

[Alexey Pasumansky]

Hello Benjamin,

The network processing feature doesn't take into account the specifications of the nodes. The server just splits the available task into subjobs and send them to the available nodes.
If one node is much faster then another, it would likely be able to complete considerable amount of subtasks, while the slower one will only finish a few.

It's hard to predict the memory consumption that will be required for the certain subtasks, as it depend on the complexity of the scene. But it wouldn't depend on the number of available nodes.

Actually, you can try to run the network processing using the single node only and check the memory consumption peak for it. Most likely it would be the same, if you process the same project using multiple nodes.