Inconsistent render time - troubleshooting

[Neilg]

I'm getting some hugely inconsistent render times with an image i'm trying to render. it's 5.5k wide - i'm doing it in strips, all materials have a reflection depth of 1 and it's using a pre-calculated IR map/LC.

Both of these were rendered on the same machine:
This strip took 21 minutes. as expected.


This strip took 4 hours and 55 minutes. If anything, i'd have expected it to be much quicker than the one before it.


ram usage is around 4/5 gb while rendering, regardless of rendertimes. This machine has 16gb...


Any ideas? low resolution tests are consistent all the way through the image, it's only when I ramp the image size up the time goes crazy sometimes.
Last night when i left it going on all machines I seemed to get 2 quick strips for every one slow one, give or take one on some machines.

Thanks in advance for any help!

[vlado]

Is the 3ds Max bitmap pager enabled or disabled?

Best regards,
Vlado

[Neilg]

It's disabled.

[vlado]

Well, you could try to generate a dump file for the 3dsmax.exe process from one of the machines while it is stuck on a slow strip so that I can check what it is that it's doing.

Other than that, are you using any proxies, VRayHDRI tiled textures, displacement, fur?

Best regards,
Vlado

[vlado]

Also, if you generate a sampling rate render element, is it consistent between the fast and the slow strips?

Best regards,
Vlado

[Neilg]

I did the sampling rate element at a lower res and it's pretty much consistent.

No proxies, a 2k wide hdri in a domelight stored with IRmap/used for reflection, displacement off, no fur. 5mil polys in total.

I generated a dump from a strip which has an eta of over 100 hours but it's 5gb... I dont know what to do with it!

I'm checking some geometry at the moment and i'll do a couple of tests with various things switched off - reflections/bumps/etc.

[Morne]

please keep us updated with your findings so we know what to look for if we encounter a similar situation...

[vlado]

I'm checking some geometry at the moment and i'll do a couple of tests with various things switched off - reflections/bumps/etc.

Also do a test without maps.

Best regards,
Vlado

[Neilg]

Ran through all those things and some overrides and I think I may have fixed it - i'll only know for sure once it gets stuck in overnight though. There was some overlapping geometry in the frame which didnt show in the render but a dirt test had the huge weird black squares littered over it. Xrefs being merged in to a scene with hidden objects and layers...
I've optimised it some more too from my run through so it's also going much faster.

I'm also rendering it without the extra tex now to save it some work, I'm going to do that separately at half res so it doesn't slow down my main renders. Update to follow tomorrow!

[Neilg]

Hit the main section of wall and after watching it for a minute, some machines started the strips crawling along with an eta of 2hours (which was done in 20 mins) but a couple have since dropped that and are going pretty quickly now.

Could it be an issue with maps transferring over the network? my materials are all layered so they look good at a distance but also hold fine detail when you zoom in 20x - there are quite a few big maps involved in this...

[Neilg]

Guess that was just how it started - the double faces thing fixed it.