Hey all,
I’m on day 7 of trying to render a scene in high detail and it’s just not working. My computer resets about 25 hours into the render, right before it is finished, and restarts with nada as far as what went wrong.
I have tried stopping the process and saving it. I have monitored my computer for hours during a go, and there are no indicators that anything else is at fault (CPU/GPUs/RAM/Disk - everything checks out, even when it crashes).
I’ve given up on rendering the scene in high at the moment, which really sucks as I think it needs it.
Anyone have any suggestions? I’ve gone over my rig a hundred times, turned off everything for RC to run, and it just keeps on jamming up and I lose my work. Doesn’t happen in any other program. Anyone else ever have these issues with long recon runs? Any tips on getting to high? Does rendering in Normal get me somewhat of the way there or does hitting recon-high start from the very beginning?
Any help is appreciated - if I can’t render this model, I can’t imagine rendering others that I am going to be paid for…
Hi Castlenock
you mentioned restarts, you mean Windows is ( W10 or ? ) restarting after some time or ?
Wishgranter,
Thanks for the quick response. It restarts sans any dialogue or warning, I’ll just turn on the computer after letting it run and find the log-in screen. Again, I’ve caught it crashing a few times while actively looking at a detailed perf monitor and there is nothing hinting at the system approaching any limits. I have done, in the past, some extensive load testing that passed and I dialed all my overclocking to ‘off’ when I started having issues with this. Only happens in RC.
Here are my specs:
Windows 10 Pro 64GB of RAM (just upgraded from 32GB a few minutes ago, between crashes).
1 TB raided M.2 memory (where I/RC works off of)
4 Ghz i6700K
2 x Nvidia 980 ti
EDIT: In answer to one of your questions: restarting after some time. Usually after 15 hours, around the 20 to 23 hour mark.
Hi Castlenock
take a look at the EVENT VIEWER - there must be a log info regarding the crash-restart, with some numerical code regarding the restarts. Or are the Windows 10 forced updates in work again??
It may be easier to use the reliability graph.
Its a somewhat hidden option in windows.
Just press the windows key and start typing reliability (zuverlässigkeit on german windows)
This view will give you information what happened on any given day.
it also shows if any applications misbehaved or windows itself crashed.
Double-click such an entry to get more detailed information.
I prefer it over the event viewer as it prevents you from having to search through sometimes thousands of log entries.
Thanks everyone, I really appreciate it. I went through reliability graph (didn’t know about that, great little tool!) and event viewer and only saw errors coming form applications as they came back up ungracefully.
In trying to troubleshoot stuff I added a second GPU card so maybe the PSU is overwhelmed at 650 Watts. I was having the same exact error before on one GPU card (which the 650 watts can handle easily) - but maybe I fixed something and overwhelmed the PSU on the other end. Time to overnight some equipment.
Hi Castlenock
Look only for CRITICAL ERRORS, you wil find Kernel erro code 41. that mean PSU issues in most cases… Try thsi http://www.coolermaster.com/power-supply-calculator/ if you PSU is capable to delvier with at least 20 % margin…
650watt is probably a bit low for 2 video cards.
dose it happen in the gpu stage or cpu?
It is error 41 - as soon as I saw it I disabled a GPU (my bios has a switch to turn off PCIe entirely) - Wishgranter, thanks for confirming. For one GPU and all the trimmings, calculators are saying 500 watts is fine and I have 650. I monitored it for a solid 2-3 hours and watched it throttle both GPUs and the CPU before it became more CPU dependent, but who knows, maybe it smashes the GPU and CPU again at the very end.
I don’t know if it’ll fix the problem, however - I only stuck the 2nd GPU in there about 48 hours or so ago because RC was crashing towards the end of the render and I was getting desperate and wanted to lesson the recon time to see if it would work (as again, it seems to be crashing at the end of the reconstruction). I may have fixed a couple of things and introduced the load error in the hopes to speed up my attempts to see if I fixed the issue or not.
So I’m on hour seven of another render with one GPU, and should have a 1k PSU coming in tomorrow (they’ve gotten smaller! 650 was the max I could find that would fit in the the absolutely magical Lian Li TU300 a few months ago). Regardless of RC I’ll need the new PSU to run anything in SLI.
Again, thanks for the help, I’ve been chomping at the bit to see the results of the shoot and under a lot of timing pressure so was getting a bit despondent - just having people respond goes a long way.
the main thing I did to stop crashes towards the end of “creating model” stage, was to put something in the “minimal Distance between two vertices”
this helped a bit. but i wasn’t i was getting pc restarting like you have.
but also keep an eye on the pagefile and ram usage.
chris wrote:
the main thing I did to stop crashes towards the end of “creating model” stage, was to put something in the “minimal Distance between two vertices”
this helped a bit. but i wasn’t i was getting pc restarting like you have.
but also keep an eye on the pagefile and ram usage.
Chris,
Thanks - where do I find that? Another thing that could be messing everything up is that I forgot to set reconstruction bounds (I generally like the extra stuff way back and leave it mostly in with a super-crazy recon box).
Well this is bittersweet! I got the high recon done, installed the new PSU, got both GPUs working, everything is grand…
…and then I lose my production AND backup drive at the same ■■■■ time. I’m livid. Didn’t lose my current project files, but I’m out in the boonies and can’t do cloud backups like normal. Just going to run a few recovery programs on my backup drive, hope for the best, and slam into this model now that it’s ready for some post work.
Thanks everyone for the help even though it was obviously my rig and not RC - really appreciate it!