I did expect a somewhat slow render but combiner just seems to hang at high res while it’s no performance culprit at low resolution.
As you can see i’m rendering at 64K & about 1000 seconds were spend on the pre combine devices, meaning around 40K seconds of the remaining time were spent hung on the combiner (over 10hours stuck on it at 49%).
Additional details :
The combiner is in “detail” mode
As you can see at even relatively high resolution the problem doesn’t appear (at 16K on the same graph combiner takes only 1 second, i expected 65K to take around 16 seconds at the usual 4X slowdown per resolution doubling, and stoped at over 40K sec)
Very bizarre. There is absolutely no reason for the combiner and that mode in particular to be special/take a long time at high res.
How much physical memory is available? It looks like WM is not managing swapping, is it possible that this is a swap-induced slowdown? Even in that case I find it unlikely.
If this happens again, go to the Task Manager, right click on the World Machine instance, select “Create Dump File”, and then send me that resulting dump file – I would like to see what is happening here internally.
It can’t be swaping (256 gig of ram and only 32 or 64 gig were in use during the build), CPU usage was at 0% when i checked so not sure what was happening. I’ll try to repro this and send you a dump.
I was trying to repro this however this time combiner also stoped at 49% but instead of hanging world machine crashed and generated this dump, please find it attached and let me know if it helps.
Would really like it if you could also give a look at my 32 thread limit bug report from a few days ago as that one is actually a lot more critical to me.
Thanks!