[itsallgoode9]

When using pretesselated displace, if I have the subdivisions set too high the render nodes will just stall at preprocessing and the network reset doesn't fix anything...I have to manually restart each node. This is only on prestesselation, never "on the fly". Is there something on the computer causing this, or is it a bug?

this issue makes it quite time consuming to find a correct level of subdivisions for pretesselation.

[Mihai]

It seems they are running out of RAM? If you render without network with the same setting on one of those nodes, does it render ok?

[itsallgoode9]

yeah, I assumed it was the ram running out, although I have 16 gb in each node, so I kinda hoped that would be enough.

I haven't been able to do a test for sure to definitely determine if that is the case, but regardless, when it hangs during preprocess phase, is it normal to not be able to reset the node from from the system manager? Instead I have to do it manually on each node. Is that a bug, or is that just how it works?

[Mihai]

I'm not sure if it's possible for the node to be reset remotely in all cases. Even locally, it would require Windows to interfere to be able to actually stop a frozen process....so that's why I'm thinking it's not possible for a remote reset command sent from another application to have any effect on a process that's frozen.

[numerobis]

taskkill /s <node name> /u <username> /p <password> /IM mxnetwork.exe


edit: sorry, it doesn't seem to work in all cases... i have to figure out why...

[itsallgoode9]

I'm not sure if it's possible for the node to be reset remotely in all cases. Even locally, it would require Windows to interfere to be able to actually stop a frozen process....so that's why I'm thinking it's not possible for a remote reset command sent from another application to have any effect on a process that's frozen.

but why does running out of ram affect maxwell render nodes SO much compared to other programs? any other programs just start using HDD space if memory runs out but maxwell's render nodes freeze.


thanks for the tip numerobis. I'll look into that. I used to have something similar but it's a pain to get working so i never set it up on my new comp. maybe i'll have better luck with your method.

[Mihai]

It's not really the node, it's Maxwell and as running Maxwell locally when it runs out of RAM it can freeze or crash. It doesn't use the harddrive because that would make the rendering 1000000000% slower

[itsallgoode9]

looking at this more, it's not that maxwell froze or crashed. if i look at the manager, I can see all the nodes still going, albeit VERRRRRYYYY slowly. if this happens, I cannot stop or reset the nodes, i can only do that manually. thoughts about that?

[Mihai]

They actually ARE using the HD to render so running extremely slow and unresponsive?

I'm not sure why you want to set the pretesselation so high that you risk running out of RAM. You need to either lower it a bit (can make a big difference just lowering it a few steps), buy more RAM, or use on the fly (which can be a lot slower, depending on the displacement height most of all).

[itsallgoode9]

oh, it's not that i'm trying to set it that high on purpose. I run into this when trying to figure out the amount of tessellation i need to get the detail i'm looking for. Obviously i'd rather not set it so high that i run out of ram.

I had been using on the fly for everything in the past but the speed improvement is huge for these images using pretesselated, so i'm trying to figure that out.


Is there any way to know how many polys will be created by the tessellation? If there was a triangle counter in the pretesselation window (adjusting to show the amount of triangles created as you adjust the subdivision amount), i think that would be helpful for better estimating this stuff.

Just another one of those things that i'm trying to figure out/understand better.

[numerobis]

taskkill /s <node name> /u <username> /p <password> /IM mxnetwork.exe


edit: sorry, it doesn't seem to work in all cases... i have to figure out why...

I just wanted to add that it was only a stupid mistake... it works as expected with "taskkill.exe" instead of "taskkill"

Code: Select all

taskkill.exe /s <node name> /u <username> /p <password> /im mxnetwork.exe  /t /f

[itsallgoode9]

They actually ARE using the HD to render so running extremely slow and unresponsive?

I'm not sure why you want to set the pretesselation so high that you risk running out of RAM. You need to either lower it a bit (can make a big difference just lowering it a few steps), buy more RAM, or use on the fly (which can be a lot slower, depending on the displacement height most of all).

I'm fine with working within the confines of what my machine can handle, but the problem is, there's no easy way to see what those confines are, when talking about pretesselated displacement.

Why isn't there some sort of fail safe, node auto reset option if it sees the computer has run out of ram? Or an option to allocate how much ram the node is allowed to use (like photoshop) and if it hits that limit, then reset the node and send you an error message.
That is a very simple solution that would make the pretesselated option much more reliable and less error prone.

It's such a huge waste of time have to manually restart nodes just because the displacement setting was set too high when i'm testing a material.

I appreciate the suggestions but they don't necessarily address the issue i'm having