[Bug - v0.6.1.9e] repeated CTD, openGL driver reports "TDR" error

[arisian]

I'm getting regular crash-to-desktop in the current build, with a "TDR" message. I'm running Windows 7 x64, most recent graphics driver, on an Nvidia GTX 580. It seems to happen pretty randomly; doesn't matter what I'm doing, and isn't triggered by doing a specific thing; I've seen it happen when I'm not even touching the controls. But it happens pretty frequently, it's not just a one-time thing.

According to my reading, TDR indicates that the GPU wasn't responding to the scheduler for more than 2 seconds; it's basically a timeout mechanism used to detect when the graphics stack crashes, since a CTD is preferable to a freeze. Based on the theory that it could mean you're overworking your GPU, I've tried lowering all my graphics settings to the lowest values, but it doesn't seem to make any difference (I've never had serious performance issues with Exanima on this machine, even with high settings). Also, this happens pretty frequently in the main "story" mode (time-to-crash is anywhere from half an hour or more at the high end, down to less than a minute after starting at the low end), but doesn't seem to happen much in the "arena" mode (I can play for several hours, with only a small chance of seeing the crash). Not sure why; could be amount of assets loaded, could be some kind of script or shader or something that isn't used in arena mode.

Anyway, it renders the current build pretty much unplayable for me (especially given that you can't just quick-save every 30 seconds; the auto-saves aren't always as up-to-date as I'd like). Now, I totally get that this is a beta, so I'm not here to complain, but rather to report the bug, and try to get a sense if it was unique to me or something that other people are seeing. Anyone else gotten this error? If so, did you find a fix or workaround?

 

[The Witcher]

ok bud gona message a moderator to look at your bug

 

[Murf]

Here Madoc replied to someone else with this issue on teh Steam forums

This is an OpenGL specific issue and it's hard to say exactly what causes it, I once started having it in all OpenGL programs and it didn't go away until I replaced my graphics card. It's also possible that it results from other software also running on your system. There's some info on what it is here, though it's the symptom rather than the cause:

http://nvidia.custhelp.com/app/answers/detail/a_id/3633/~/opengl-message-“tdr-detected”

You could try increasing the timeout in the registry, when I had this issue it did not help though, and I think 2 seconds is unrealistically long already. There may be some other settings that might affect this, such as vertical sync and triple buffering.

I wouldn't worry about turning hardaware antialiasing off if you're using supersampling. Antialiasing is more of a crappy alternative to supersampling, which does the job much better and other things, even though they do work together.

And here is another one about it as well:

http://steamcommunity.com/app/362490/discussions/1/458607699622605442/



It's not something that has been able to be pinned down specifically yet, but it does seem to be system side and OpenGL related

 

[arisian]

Okay, I'll keep messing with it and try to figure out whats going on. It doesn't seem to be a generic OpenGL problem, since I don't get this crash in other OpenGL games (just tested it out this morning, and was able to run a different OpenGL game for several hours with no problems; afterward, I fired up Exanima, and it crashed within minutes). It's possible that some other OpenGL things will cause this crash, but clearly not all of them, and so far I've only seen it in Exanima. I don't think it's an issue with the GPU overheating; my graphics card is well enough cooled that even furmark can't get it above ~45C. I tried messing with graphics options and limiting framerate with vsync, but none of it made a difference. I also tried doing a "clean" uninstall/reinstall of the nvidia drivers, but that didn't make a difference either.

What's really strange is that the TDR mechanism is supposed to be a freeze-detection safeguard, but the TDR error is not preceded by a freeze. Visually, the game is smooth as butter, right up until the OS decides to kill it. I have, on rare occasions, seen the game slow down or stutter momentarily, but it's never a "freeze", and it definitely never lasts for 2 seconds (which is the default timeout for the TDR). It's like the OS thinks the graphics stack is unresponsive, even though it seems visually to be working just fine.

 

[arisian]

Success!

After a bit of poking, I determined that (at least for me) this seems to be some sort of strange "overclocking" related error. I put "overclocking" in quotes because I've never actively overclocked this card, it's always been run at it's default settings, but this card came from the factory with a mild overclock built into those defaults. I've never had any problem with it in any other game, and still don't (including openGL games that push the card much harder than Exanima does), so I don't really think this is an issue of the card showing generalized signs of stress. However, by using an overclocking tool to underclock my card (relative to its defaults) to get it back down to what nvidia lists as the baseline specs for this GPU, the TDR errors seem to go away, so it's clear that the "overclock" is causing the problem.

I still think that Exanima must be doing something unusual, since it's the only program I've ever seen have this problem, but it's also clearly got something to do with the overclocking, which is obviously not the fault of the Exanima devs. It might be worth mentioning the issue in some sort of FAQ, though, since factory overclocks are not all that uncommon. Given that others have clearly experienced this problem already, it's something that people should be made aware of. Especially given that there are likely other people out there who don't think of their cards as being overclocked (and therefore it doesn't occur to us as a possible cause of the problem), even though they are technically over the official spec.

 

[Tony]

@arisian thank you for sharing these results.

 

[Murf]

{Hell, this was sitting as a draft all night, meant to post last night I swear}

Wow very interesting for sure. No issue with OC here, but I am not using factory default OC I am OC beyond that. Maybe it is an issue with older cards, haven't had a lot of these issues to make a definite determination if that is the case or not. Thank you for letting us know though

 

[Faelivrin]

I am also having CTDs with this error in my ancient 560TI OC (overclocked by factory). It worth pointing out that i starting having in recent versions of the game. I reported the first one and the second one here, but i have suffered a couple more times since then:

http://www.baremettle.com/forums/index.php?threads/0-6-1-2d-nvidia-opengl-driver-crash.3145/

 

[arisian]

It worth pointing out that i starting having in recent versions of the game.

This was the case for me as well, but unfortunately I'm not sure this makes it any easier to pin down. This kind of thing is almost impossible to debug unless the devs have access to a system on which it occurs (and even that's no guarantee). But if it's a problem specific to OC'd cards (not ALL OC'd cards, but ONLY OC'd cards), and there's a fix (reduce clocks to spec, even if that's below card defaults), then it may be something the devs can afford to leave as it is. Though that will require work to make sure people are informed of the potential problem and what the fix is, so the impact can be minimized.

If this is happening to multiple people in the "alpha" player base, it's safe to assume it will happen to a non-trivial number of users once the game hits mainstream release, so if it can't be tracked down/fixed, it's important that it be dealt with properly.

 

[Faelivrin]

there's a fix (reduce clocks to spec, even if that's below card defaults), then it may be something the devs can afford to leave as it is. Though that will require work to make sure people are informed of the potential problem and what the fix is, so the impact can be minimized.

If this is happening to multiple people in the "alpha" player base, it's safe to assume it will happen to a non-trivial number of users once the game hits mainstream release, so if it can't be tracked down/fixed, it's important that it be dealt with properly.

Yeah i hope they can locate and fix the issue before v1.0. It is a showstopper bug, and the most reliable workaround can be very dangerous thing to mess with for not advanced users.

 

[Fanzeo]

Yeah i'm running with a gtx 580 on windows 7 x64 as well, and recently my driver has been crashing consistently within 5 minutes of playing. I tried a driver update with a clean install but to to no avail, so i'm going to try that overclocking method you proposed. Hopefully I don't need to replace my card, but I was interested in upgrading fairly soon.

 

[Fanzeo]

Arisian you're a genuis! Just underclocked my gpu and i've been playing a steady 2 hours without any driver issues. Thanks for the help!