Random crashes when slicing big models (Intel i9-13900K / 14900K issue ?) #11845
Comments
|
Hello, Same issue here since prusaslicer 2.7.1 update. |
|
I have the same issue |
|
Having same issue, When I try to slice a complex model it crashes on 20%. |
|
Guys, we are thankful for your reports, but we need at least minimal cooperation in order to be able to find and fix them. Please, enclose your 3MFs, please. Please. We want to fix it! |
|
In my case the crashes happen with all the models I download. This is not related to a particular model, it is for all models, and even as soon as I had already printed with prusaslicer 2.6. The crash is relatively instantaneous, I click on Slice Now, and PS closes by itself, without an error message. As I couldn't find where the error was coming from, I uninstalled PS and all my profiles (although I use that of Artillery X1 and FLSUN V400), and since then it seems to stabilize. I've had 1 or 2 other crashes, but it's no longer instantaneous. I wonder if it is not linked to profiles... There should be an option in PrusaSlicer, when you install the program, it is to "empty PrusaSlicer of all profiles to start with a clean installation" I'm going to do some tests again, select old profiles, and STLs and I'm going to add that to the ticket. What is certain is that I am not the only one and that this happens to quite a few people since version 2.7. Maybe it's related to a hardware conf too? I don't know. |
|
COLLAPSING-KATANA.zip |
|
I confirm, i crash too when I slice this model :) |
|
@lukasmatena it probably not caused by specific models, for me every bigger model crashes the Slicer the same. If I scale it down it slices just fine. Is there a way for me to enable a debug mode or access additional logs? |
You are on Windows, meaning you can enable local crash dump collection. The crash dump would show us where the application crashes. If you want to spend time with it, it would be very helpful. See the respective wiki page for instructions.
Thanks, but I am not able to reproduce it with your 3MF either. |
|
prusa-slicer.exe.44728.dmp |
|
STL + Dump file |
|
Thanks. The backtraces from @KenneyDe and @SckyzO are not the same. @KenneyDe This is from your dump: @SckyzO Your dump: |
|
It might also be worth looking if you are not running out of memory (by watching system monitor during slicing). But that should still be reported more gracefully than this. |
|
Hello @lukasmatena Thank you for your feedback. My computer is an Intel I9 14900K, 32 GB DDR5 6800. So I think that's not the problem :) How can I help provide you with more data? |
I’m also running 14900k, 64gb ddr5 on my end, rtx4090. |
|
Just an update, I have tried disabling Arachne perimeter generator (switch to classic) and still get the same crash at 20%. Also, I noticed if i zoom the camera to where the 3d object is no longer visible, it will progress past 20% and crash on 90% generating Gcode. Here are both dumps. (dump203706 is the 90% dump). |
|
Thanks for all the data. Is it ok in 2.6.2? Does it work in the 2.7.0-alpha? Can you guys also test which is the first PrusaSlicer version where it crashes? |
|
Could you please try out the following build? It is very wild guess and it will probably not solve the problem, but it is worth trying. PrusaSlicer-2.7.1+3-lm-mysterious-clipper-crashes-win64-g945677b92 Thanks. |
NoSuchKey
The specified key does not exist.
No such object: prusa3d-content-prod-14e8-slicerbuilds/PrusaSlicer-2.7.1+3-lm-mysterious-clipper-crashes-win64-g945677b92-202312221129.zip
|
|
|
@Maywtf12 Sorry, my bad. I have fixed the link in my previous post. |
|
I'm running 2.7.1 on Windows 11 on a 13900k, 32gb DDR5, RTX4090. I think it may be important to call this out as it seems at least 2 others in this thread are using new high-end rigs and initially made me wonder if there could be a hardware component to the issue. I'm experiencing crashes when slicing that sound like the same symptoms as OP with the app crashing shortly after the slicing starts. I am able to slice the exact same Prusa 3mf file on a Win10 virtual machine on a different server as well as an older laptop running Windows 10. I got an inkling to try a CPU limit using my motherboard utility since I did notice an quick fan ramp on my CPU AIO before crashing. I went into my BIOS and the configuration for the CPU power was essentially set to unlimited and after specifying the Intel POR limits (~253 Watts) it hasn't crashed since. I've repeated the test 3 times with and without limits and it continues to crash when power is not limited (even though I never see the charts go above 255 watts, but that could be a sampling issue). I did try running with the integrated graphics disabled earlier and it had no effect. I'm not sure what the Slicer could be doing differently than other intensive applications... maybe using a different/new instruction set... Either way... it's working crash-free now and I'm hoping this helps someone else's issue as well! Take care! |
|
Wow, that's funny, today I did the same thing. I used XTU to reset overclocking and add CPU limit, and after that no crashes. I actually think we are making progress on the bug. |
This version also crashes the same way, uploaded a fresh dump. |
|
Thanks.
It could be, but if you are all having problem with PrusaSlicer and not with any other application, it may as well be some threading bug in our code, which just incidentally crashes on the specific processor models. Could you please test in which version it started? Was it ok in 2.6.1? Was it ok in 2.7.0-alpha1? |
|
Also, I have two builds with two separate attempts to fix it. Could you please test them both and tell me whether it helps or not? |
|
I've been having the same exact issues and was able to fix it. Like @newlandk said above I think the issue is related to using "high-end rigs". This is my new work computer and I don't know much about it but the IT guy who built it was able to fix the issue. He said went into BIOS and "turned off sync all core overclock and switched to 'auto'." I hope this helps. If you need more info I'll try my best to find it. |
2.7.1+3-lm-mysterious-clipper-crashes-win64-g945677b92 build - I have had five consecutive successful slices when loading the sample .3mf project I uploaded in the other thread. However, when I load the source stl into a blank project and enable organic supports it still crashes silently. Both behaviors have been fully repeatable over about 5 attempts 2.7.1+1-lm-updated-tbb-win64-g1f6f352df build - one successful slice from the 3mf project, but then three consecutive silent crashes after that. As far as I know, I am not having issues with any other applications. |
Interestingly, If I add the source stl, enable organic supports, save a new .3mf project, close PS, reopen PS and load the saved file, the crash so far has not occured over 3 attempts and the slice was successful. It's not conclusive, but the "mysterious-clipper" build seems to work with a project loaded from a .3mf, but not if I build the project manually and slice without a .3mf save and a reload. |
|
After reading through reddit:i913900k_instability_crashes_and_svid_behavior and others, I am starting to think that it really could be a hardware issue. The common denominator in all these strange bug reports are the i9-13900K / i9-14900K processors. The crash dumps that some of you have posted lead to various places without an obvious connection - they all end somewhere in the Clipper library, but that can be a coincidence, given that Clipper is frequently used during the slicing process. This would indicate that they occur more or less randomly, when the CPU is under heavy load. I am not even slightly a hardware expert, but reading though the online discussions would indicate that the default BIOS settings are set at the edge of where the hardware is stable, maybe in an attempt to make the high-end CPUs look like even higher-end. Who knows. These are the things that some people report as solutions in similar cases (BIOS settings):
Could you all please try that and report whether it helps? And I am still looking forward to when someone will share the information when the problems started. If it is ok in e.g. 2.6.1 (on the same computer), it would indicate that the bug is really in PrusaSlicer, or that we at least started doing something that provokes it. Thanks.
This is quite confusing. Some defects in the model may have been repaired during the save-load, so it is possible that the model from the 3MF is (slightly) different, but I don't see the connection with the changes in the "mysterious-clipper" build. |
lukasmatena
changed the title
|
the issue happens in prusa slicer, bambu studio, superslicer, so seems that whatever it is, likely is underlying from slic3r. |
|
prusa-slicer.exe.2540.dmp |
|
Good morning, the best way I found to stabilize PrusaSlicer is to tune the CPU a little thanks to XTU (then if you want to put it in the bios to save the parameters hard, that's no problem).
|
|
Hey everyone! so Im not actively using Prusaslicer, I use Orcaslicer but I am having the same issues as described here. I hopped over to Prusaslicer to see if the same crash occured and copied as many settings as I could over from Orcaslicer. In a file that the slicer states has more then 1M triangles even when simplified it crashes. https://www.thingiverse.com/thing:6382941 I have to cut the mask in half in order to fit my build plate. I recently upgraded from a I7-10700k to a I7 13700k both with 128gb of RAM. Both systems would crash but only when it had any type of Overclock, I mess around with some BIOS settings when overclocking, but if I just leave AI Overclocking with the 13700k the programs do not crash. I believe its definitely some sort of Hardware issue but its a spike that occurs when the walls are being generated. It feels like that activates a spike of some kind that the programs do not like. My 10700k used to BSOD the PC, while my 13700k now just crashes both prusa and orca. Im going to try the suggestion above from [lukasmatena] and report back on prusaslicer. |
|
I tried the suggestion and it did not work out for me. I did get BSOD every time i tried a different SVID behavior option. Unless I left on Auto which I found odd. I think the spike might not be on the CPU side, but on the RAM side. if I leave it on XMP profile base it crashes . but recently I starting messing around with basic overclocking of my RAM. Im running 64gb Trident Z RGB My XMP profile is 3600 mhz at 1.35v it crashes. but if I set it at 1.45v-1.5v and increase it to 4000mhz or 4266mhz in the BIOS, it does not crash and slices quite quickly. idk if anyone else wants to try that out. |
|
I opened an issue for the crashing as I was unaware of this thread. Thank you @lukasmatena for your recommendations to test. I only had to make the one change to SVID behavior. Motherboard: ASUS ROG STRIX Z790E GAM WF D5 OS: Windows 11 Pro Version 22H2, OS build 22621.3155, Windows Feature Experience Pack 1000.22684.1000.0 PrusaSlicer: Version 2.7.1+win64 In the BIOS I changed: EXPERT -> AI TWEAKER -> SVID BEHAVIOR -> INTEL'S FAIL SAFE PrusaSlicer now works as expected. |
|
Well, it did work. I upgraded PrusaSlicer from 2.7.1 to 2.7.2 and can no longer slice again. I downgraded back to 2.7.1 but no change, still cannot slice. When I try to slice the program just closes with no errors. Back to the drawing board. |
|
I had a similar experience @jcleek. My solution that is still working is to set the core voltage to a fixed value manually and lowered the Turbo Boost maximum power draw. In my case I chose 1.425 volts and 200 watts. I have a 300mm AIO and don't see exorbitant temperatures... so I wouldn't expect whatever is happening to be thermal related. I'm wondering if something is hiccuping with power delivery when the load hits the CPU. What's odd is the regularity of being able to trigger this with a slicer... I have had one or two instances where I've seen similar behavior when under other loads, but Prusaslicer seems to trigger without fail when slicing a large file. Are there particular instructions that are being used that would behave differently than a standard benchmark or stressing utilities? |
|
Replaced my processor today with a i9-14900K and immediately tried to slice the same model I have been having issues with this week. Sliced without issue. I put in more and more complexity in the slicing and it didn't fail or crash one time. Guess it is time to reach out to Intel to get a replacement or refund. |
13900K on an Asus Z790 ProArt. Same issue and this suggestion fixes the problem. I picked SVID > Worst Case, which also works on my config. |
|
I was able to fix this by limiting the Program to my 14900K cores 0-7. My steps are as follows: open task manager, go to the details pane, scroll down to find Prusaslicer, right-click Prusaslicer and select set affinity, select only cores 0-7 or the respective amount for lower-end LGA 1700 chips, then confirm the selection and Prusaslicer works as expected. |
|
@ke19lly that only works per session. If I close Prusa Slicer and reopen I have to do that again. Any chance you found a more permanent "fix"? |
|
No, I haven't found a more permanent solution yet. I've submitted a feature
request asking for a compatibility option to be added, but for now, it
seems we're stuck with this temporary fix.
…On Mon, Mar 11, 2024 at 8:11 AM Jim ***@***.***> wrote:
@ke19lly <https://github.com/ke19lly> that only works per session. If I
close Prusa Slicer and reopen I have to do that again. Any chance you found
a more permanent "fix"?
—
Reply to this email directly, view it on GitHub
<#11845 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AQRPX4ER37LMPGVLAU5QNH3YXXCR5AVCNFSM6AAAAABAHVCN32VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSOBYGUZTOMJZGI>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
I have a more permanent "fix".
This isn't a fix, just a good workaround for PrusaSlicer. |
|
I attempted that workaround yesterday but encountered an issue where the
shortcut wouldn't launch the slicer. I'll give it another try, as it would
indeed be convenient to avoid manually setting the affinity each time.
Thank you for the detailed instructions.
…On Mon, Mar 11, 2024 at 8:47 AM Jim ***@***.***> wrote:
I have a more permanent "fix".
1. Locate your prusa-slicer.exe file. This is the executable and is
likely located here: C:\Program Files\Prusa3D\PrusaSlicer
2. Right-click it and select CREATE CHORTCUT. If you don't see CREATE
SHORTCUT you are on WIN11 and you should first click SHOW MORE OPTIONS ->
CREATE SHORTCUT
3. You will probably see an error that you can't create a shortcut
there and ask to create one on the desktop. Answer YES.
4. Find the shortcut on your desktop. Right-click it and select
PROPERTIES.
5. In the TARGET box add this in front of what is there:
%ComSpec% /C Start "PrusaSlicer" /Normal /Affinity FF
So the TARGET box should look something like this:
%ComSpec% /C Start "PrusaSlicer" /Normal /Affinity FF "C:\Program
Files\Prusa3D\PrusaSlicer\prusa-slicer.exe"
6. Click CHANGE ICON
7. In the top box LOOK FOR ICONS IN THIS FILE put in the path to your
prusa-slicer.exe file like this: C:\Program
Files\Prusa3D\PrusaSlicer\prusa-slicer.exe and press ENTER
8. You should see the icon for PrusaSlicer. Make sure it is selected
and click OK
9. Click OK
10. Now double-click the shortcut. Check the affinity by looking at
the details in Task Manager as @ke19lly <https://github.com/ke19lly>
indicated above.
This isn't a fix, just a good workaround for PrusaSlicer.
—
Reply to this email directly, view it on GitHub
<#11845 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AQRPX4GFI2R25WDWGLGTOCTYXXGXZAVCNFSM6AAAAABAHVCN32VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSOBYGYYTOMJZGA>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Same result here. with a slightly different motherboard, memory, and GPU: Motherboard: ASUS ROG MAXIMUS Z790 FORMULA |
This fixed (or at least bypassed) the issue for me, thanks! |
Description of the bug
Since updating the PrusaSlicer to 2.7 I'm having problems slicing big models. I press "Slice Now", it goes to 20% maybe and then crashes without any message. It works just fine if I scale the model down substantially. Is there a way I can enable additional logging? I attached the event from windows event viewer.
Also full spec:
https://pastebin.com/bQXTByYB
Project file & How to reproduce
RG351P_Epic_ultra_Release_candidate_SD1.zip
Checklist of files included above
Version of PrusaSlicer
PrusaSlicer-2.7.0+win64-202311231457
Operating system
Windows 11 (build 22631), 64-bit edition
Printer model
Prusa MK4
The text was updated successfully, but these errors were encountered: