LiveKernelEvent Lockup on an AMD GPU with official drivers.

[SoftwareSlicer]

Expected behavior: The game scales and renders across different resolutions without interfering with system stability.

Current behavior: Scaling the game across multiple screens causes my computer to lock up with graphical artifacts and multiple LiveKernelEvent error codes logged clicking the reset button. In other words, My computer is absolutely not vibing.

Failure Information: I noticed when running the game yesterday with the mod that there were issues with textures and models glitching out like crazy (That's a whole other can of worms that I might want to document in another bug report.) and then I decided to move the game to another screen and then back to my main screen which shortly led to the entire system locking up with graphical artifacts on both displays. Checking reliability monitor afterwards revealed to me that my graphics card experienced four different LiveKernelEvent issues. The codes for each LiveKernelEvent include 141, 1a1, and two 117 errors. It's possible but unlikely that this might have to do with the status of my video card given it's age and use. Other users on AMD graphics cards have experienced issues with performance and stability although I haven't seen anything like this before.

Steps to reproduce: Load up the game with the modifications and hardware mentioned and hop into a world. Move the game in windowed mode from a 1440p display to a 4k display. Move the game back to the 1440p display. Wait a few seconds and there you go.

Context: Graphics Drivers: Radeon Software Version 21.5.2. (Latest currently available for my graphics card.) Operating System: Windows 10 Home 20H2. Build 19042.985. KB5003173. Hardware: Intel core i5-10400f, Msi Mag Mortar B460, 32 gigs of ram, Rx 470 video card, 2 terabyte nvme ssd. Displays: 1080p 144hz screen (Set to 1440p with VSR so I could record lighter games at a higher resolution while still playing at 144hz.) with a 4k 60hz second display. Mod Versions: fp2-0.0.1.390-1.12.2 (AMD Performance Workaround branch.), ForgeRocks-6.13.3, CubicWorldGen-0.0.146.0-SNAPSHOT-all, and CubicChunks-0.0.1192.0-SNAPSHOT-all.

Here's a screenshot of one of the hardware errors that Reliability Monitor documented:

Here is another screenshot showcasing other events that popped up during testing:

This is actually the first issue thread I've made on the platform so hopefully the information I provided helps out with development and that I formatted this issue thread appropriately.

[DaMatrix]

thanks for the detailed report! i'll definitely look into this as soon as possible.

is this reliably reproducible, or does it only happen sometimes when moving the game between monitors?

[SoftwareSlicer]

At this point, I am not entirely sure as I've only tested the modification once as after my system locked up, I felt pretty concerned about the reliability of my graphics card. (Especially given how expensive new cards are right now.) I will test out the modification again tomorrow and I'll see if I can replicate this issue and also provide more details regarding how well the mod is working and stuff as I noticed during that point that there were also other kinds of graphical issues with chunks loading in and out depending on where I looked and things generally being pretty strange. If possible, I'll see if I can record footage of this stuff happening although that will depend on if the LiveKernelEvent errors are replicated or not. I believe it might have to do with how video memory is allocated for the game but I honestly have no idea. Like I mentioned earlier, It also might be possible that this has to do with my graphics card acting up as I've started to notice odd behavior present in other applications like Minecraft Bedrock as shown in the attached screenshots below.

[DaMatrix]

ah. if it's happening in other applications then you could very well be right in assuming it's a hardware issue. maybe someone else would be able to reproduce this?

[SoftwareSlicer]

Disclaimer: The videos I uploaded have flashing lights and colors due to the game really freaking out with rendering. Especially with the end of the second clip where the graphics card kind of just called it quits. Also both clips are still uploading and processing so it might be quite a bit before both clips are available.

So I did some further testing and during my first test where I didn't change the size of the window, My computer did not lock up but the game itself experienced a bunch of weird oddities. One issue included how it was attempting to render outside of full-screen onto the second screen. Another issue I had includes how textures and video memory just kind of flickered all over the place if I looked in a certain direction or moved to a specific position in-game. During this test, As I was loading it up, I figured it would be a good idea to record what was happening with my phone as well as with my computer as my recording software was acting up. During editing, I synchronized both clips I recorded to the best I could when given how OBS Studio was not vibing. Here is that first comparison clip: https://www.youtube.com/watch?v=KzHXeuzpp8Q

Afterwards, I rebooted and I did some hardware testing to make sure my graphics card was not acting up and I ran Furmark for about half an hour and then Unigine Heaven for two hours and both tests ran without any issues. As a result, I am reasonably sure my graphics card is working as intended. I believe the graphical artifacts present in the screenshots I sent earlier might of had to do with how the mod was interacting with video memory as I took that screenshot shortly after the first test before rebooting and checking the status of my graphics card more throughly. I realized though that I haven't tried to resize or move the window like I did initially so I tried that for the second testing session and I replicated the exact LiveKernelEvent issue that I experienced during my initial testing although I only experienced a single LiveKernelEvent error instead of four. After just a few minutes, I had the entire system lock up as shown here: https://www.youtube.com/watch?v=XFvx4ZGrfJk (It says it will take nearly three hours to upload at the time of me writing this.)

Hopefully these clips help out with development and describe the problems I experienced in more detail as I saw a popular video on this mod a few days ago and I wanted to try this out for myself.

[SoftwareSlicer]

These two screenshots I took from the second clip is a pretty good summary of what I noticed during the second testing session:

The second screenshot was from my phone's camera as there wasn't any way I could record with OBS if that was happening to the computer I was recording with. Figured this would help while I wait for the clips to upload and process as I have pretty slow upload speeds.

[DaMatrix]

oh wow, this is the most amazingly detailed report i've ever seen. thank you so much for actually putting in significant effort and not being just another person saying "why game crash pls help"!

well, assuming it isn't a hardware issue, this is almost certainly just another bug in AMD's shoddy opengl implementation. not much i can do other than tell you that i plan on circumventing the use of opengl altogether on AMD systems, assuming all goes according to plan, which is likely to to help. other people have reported similar graphical glitches on AMD (although none of them went outside of the game window...), so hopefully avoiding the broken part of the AMD driver entirely will be enough to fix all of the visual and performance issues at once.

i'll keep you updated with any further developments regarding the AMD situation, but you may want to follow along with the progress of #19 as well.

[blix5]

The same thing happens to me as well using the latest AMD workaround fix, but it only happens on voxel mode for me. Changing it to heightmap mode completely removes the glitching out.

[DaMatrix]

could i ask you to try this again on the latest build and see if this still occurs, or at the very least, whether the framerate is any better than in your previous tests?

[DaMatrix]

SoftwareSlicer reported over discord that this has been resolved.