Digistruct Posted November 5, 2016 Share Posted November 5, 2016 (edited) Hey guys! I've had some serious issues these past days with insufficient RAM where Windows 10 force closes Houdini, so it's not even real crashes it just kills Houdin mid-simulation. Since it's a very troublesome process to try and detect when the force quit has happened and restart it from the last simulated frame I'm turning to you guys for help. 1. Is there some way I can limit Houdini so that it sims slower but I won't have to worry about it crashing? 2. (Probably a bad idea) but is there some way to stop Windows from killing Houdini if the problem should arise again? 3. Am I simply in over my head and calculating something too heavy that won't be possible to write out? (Feels pretty weird though since I can simulate at least around 10-12 frames between every crash) Any and all answers are greatly appreciated! Cheers, Jack Edited November 5, 2016 by Digistruct Quote Link to comment Share on other sites More sharing options...
catchyid Posted November 5, 2016 Share Posted November 5, 2016 Have you tried limiting your cache limit to something "small" (depending on your machine), In my case I set it to 5GB (DOP Network --> Cache --> Cache Memory = 5000, I think this is the default anyways). Also, maybe you should make it really small like 1 GB and enable "Allow Caching to Disk". I see also "Save Checkpoints", the documentation says that it will save the entire state of the simulation so you won't start from scratch each time (I think in your case, if the simulation crashes at frame 100, then next time it starts it will start at frame 100, I have not test myself take this with a grain of salt ). Finally, have you tried diving inside the DOP network, go to Output node and just click Save to Disk? Quote Link to comment Share on other sites More sharing options...
Digistruct Posted November 5, 2016 Author Share Posted November 5, 2016 Hey Catchyid! Thanks for the reply, I can see now that I should have been way more explicit in my previous post haha. I am not simulating per se but writing it to disk through an ROP Output node and that's when the crashes happen, the DOP Network is switched off completely! Quote Link to comment Share on other sites More sharing options...
catchyid Posted November 5, 2016 Share Posted November 5, 2016 Aha yep, whatever I said has nothing with your problem I am not expert here, but have you tried looking at the console (i.e. run houdini from dos prompt and see it it has written any useful hint before it crashes?), also maybe you could try using File Cache or simply File node and switch to output (i.e. instead of ROP output)...Other than that, I really don't know! Hope you can figure it out Quote Link to comment Share on other sites More sharing options...
Atom Posted November 5, 2016 Share Posted November 5, 2016 I just started using Houdini on Windows 10, on a low end 14GB system, and my experience is that Houdini is stable on that platform so I would not jump to the conclusion that it is the OS causing the problem. I would investigate the hardware first. How much memory do you have? Are all the memory chips running at the same speed? I know we rely on the motherboard to help keep mis-matched memory speeds in check but I find that Houdini is sensitive to this and will often crash if chips are not running at the same speed. Pull down CPU-Z and examine the RAM chip speeds. Also, are you overclocking your motherboard or GPU? Is your system stable with a 2 hour prime95 stress test? Quote Link to comment Share on other sites More sharing options...
mosssi Posted November 5, 2016 Share Posted November 5, 2016 I had this problem some months ago and after lots of try and error I found one of my eight memory modules was corrupted so I changed it. This is probably different to your situation but if you are facing a new machine so it's better to check each ram modules first. Quote Link to comment Share on other sites More sharing options...
Digistruct Posted November 6, 2016 Author Share Posted November 6, 2016 (edited) Thanks for the replies you guys! I'll follow along with the steps you recommended Atom and then I'll report back to you guys on my findings! EDIT: Due to my lacking tech wizardness I do not think I'm fully qualified to indentify an issue should I find it, but anyways to answer your earlier questions. 1. I just upgraded my computer to 64 GB of RAM, which (I hope) should be plenty for now. 2. According to CPU-Z they are all running at 1067 MHz in frequency (please let me know I am looking at the wrong thing here). 3. The CPU is overclocked, yes! But it's been for a while now and this issue is recent. 4. Will run the prime95 now and check back! EDIT: I ran prime95 for 2 h 14 m and all of the workers reported zero errors or warnings - so unless I am missing something there is no problem there. Edited November 6, 2016 by Digistruct Quote Link to comment Share on other sites More sharing options...
Atom Posted November 6, 2016 Share Posted November 6, 2016 (edited) Well that's good news it looks like your hardware is stable. Is there any way you can post the problem scene? What are your power settings on Windows 10? Maybe it is something silly like the screen saver is kicking in and that messes things up? Or power save mode is turning off the disk when your are trying to write to it? Is your disk a SSD? If not, you may want to schedule a surface scan to see if there are any problems with the disk drive itself. There is also the dreaded TDR which can cause things to time-out or error out. If you have OpenCL enabled on the FLIP solver it may be contributing to instability. Basically it is a mechanism to prevent blue screen crashes by safely crashing the video driver instead. Are you monitoring heat? Perhaps things are getting too hot? You can monitor heat with the companion to CPU-Z called HWMonitor. Edited November 6, 2016 by Atom Quote Link to comment Share on other sites More sharing options...
Digistruct Posted November 6, 2016 Author Share Posted November 6, 2016 (edited) Sure, I'll post the scene! Also this message keeps popping up in the console: "OpenCL Exception: Failed to create compute grid. (-4)" Might that be connected to the crashing? I am running Houdini from an SSD but not writing to one, if you think that could be the problem I'll start simming to the SSD instantly. I'll download HWMonitor and see if there might be any heat related issues, but the computer is equipped with liquid cooling so I doubt that'd be the issue - but I'll check it out! P.S. Managed to sim a few more frames so I don't think the issue was with a specific frame in the simulation or so, this time it proper crashed my computer though! EDIT: For some reason I can't upload the file, I'm only getting a notification about it failing. https://www.dropbox.com/s/1d88ru5mp4fhm0s/gasCanister_test.rar?dl=0 Try this link and let me know if it's working as intended. Edited November 6, 2016 by Digistruct Quote Link to comment Share on other sites More sharing options...
Atom Posted November 6, 2016 Share Posted November 6, 2016 (edited) I took a look at the file and I don't think I can re-create on my side due to the complexity and multiple caches that are needed as well as missing OBJ files. But you may want to try a larger division size on the smoke object itself just to sim quickly through more frames. Also try turning off Use OpenCL on the solver to see if that makes a difference. If you have not seen this thread on smoke grenades, check it out, there is a complex example file posted. Edited November 6, 2016 by Atom Quote Link to comment Share on other sites More sharing options...
dimovfx Posted November 7, 2016 Share Posted November 7, 2016 Using OpenCL means that the simulation is limited to VRAM only. So if you don't have enough RAM on your GPU this might be the problem. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.