Trine reboots my PC. Software or hardware issue?

[TyraaRane]

Hi guys--

So, this is a weird issue, and in the process of chasing it down I figured I might as well post here to talk directly with the devs.

First of all...I want to say that the game is absolutely gorgeous and I've been enjoying it. Especially the little bits of humor. I was playing it on the current system (posted below) with an ATI 4870x2 and everything ran great at max settings. Then I upgraded to an ATI 6970 and the weirdness began:

At random intervals, for no discernible reason, while playing Trine--and so far I've only been able to replace this with Trine--the PC would shut down and then reboot. Event Viewer's no help--it just displays a generic "oh noes I shut down unexpectedly" critical error with no useful error codes inside. I've verified that the CPU and GPU aren't overheating (last time I was logging when the crash happened, CPU was at ~56C and GPU was at 75C, respectively). This has happened at 30 seconds into the game (from clicking "Continue"), a minute into the game, and as far in as 30-40 minutes.

Here's the system in question:

CPU: Intel Core 2 Quad Q9650 @ 3.0Ghz
RAM: 8GB DDR2 (4 x 2GB)
Motherboard: Asus P5Q Pro
GPU: Sapphire ATI Radeon 6970 2GB
GPU drivers: Catalyst 11.8
PSU: Thermaltake Toughpower 750W
HDD: Hitachi 1TB SATA
OS: Windows 7 Professional x64

Now, I'm totally baffled by this. I know random shutting down is more indicative of a hardware failure than anything, but the fact that this only happens in Trine so far is puzzling me. (Note: not blaming the game here! Just trying to figure out the problem.) I've hauled the system into the shop, where they're currently running tests on it overnight. Shop guy says PSU is fine, HDD might be on the way out but we're not sure if that's the root cause of the issue yet.

So. While he's chasing possible hardware issues down I thought I might post here and see if the devs themselves had any ideas. It's the Steam version, so it's up to date, and I verified the game cache and it's all good. Settings (off the top of my head) are: windowed mode @ 1366x768 resolution, very high graphics quality settings, anti-aliasing and antistropic both set to max. Cannot remember all the little ticky boxes off the top of my head but I do believe all the graphics-related ones are checked. EAX is not, and there's one other that isn't but I can't remember what it is. I think it's sound related. These are the same settings I used with the 4870x2 which, again, ran the game without errors.

Anyway, I've rambled enough, but if someone at Frozenbyte has any ideas--please share! If it turns out that this isn't a hardware-related issue, once I get the PC back from the shop I'd like to have an idea of where to start on my own troubleshooting if it turns out to be a problem specifically with Trine. Thanks!

[fb_joel]

Hmm, sounds very strange. The game tends to reveal any GPU-related issues (could be drivers too) because it's "GPU-bound", meaning that the CPU is not really doing all that much and the GPU can do things as fast as it wants... So typically something like this would indicate a hardware issue with the GPU, RAM or the PSU, and normally I suggest limiting the framerate (open game console with F8 and write "/show_fps 1" and "/render_max_fps 30") and changing the graphics settings a bit (lower), but that might not apply here.

So if no hardware fault is found, this can get really tricky. The solution might be something simple like some program interfering in the background (or even Steam). One thing we can do is to run the game without Steam - I'll email you some thoughts on this particular point.

Let's see how this develops...

[TyraaRane]

Thanks for the prompt reply, Joel!

You're right--this is really strange (not to mention the cause of much hair-pulling). I'm actually sort of hoping this turns out to be a hardware issue...it would make figuring out the problem that much easier. We'll see how the tests turn out tomorrow.

I thought about driver issues, too, but the odd thing there is I've always kept my drivers up to date, so I was running the 4870x2 on the 11.8 set already. Plus, on the off chance I'd screwed something up by uninstalling/reinstalling the drivers when I put in the new card, I did a system restore back to before I did all that--no dice; still crashing. Could always have something to do with the combination of the newer hardware + drivers, I suppose.

Once I get the PC in question back, assuming we haven't discovered a hardware fault in the meantime, I'll try running the game sans Steam as you suggested and see what that does. May try limiting the framerate and toning the graphics down as well. I'd really like to get to the bottom of this.

[TyraaRane]

All right...after chasing some software gremlins around for a bit, I'm back.

First of all, good and/or bad news is that a teetering hardrive aside, my PC passed all the stress and diagnostic tests with flying colors. GPU, RAM, CPU, PSU are all fine. Temps are also fine, again.

This pretty much brings it down to either an issue of software gremlins or--by totally random chance--Trine has for some reason chosen to highlight my HDD's imminent demise. So, I brought the PC back home to do some testing of my own.

First off was to delete the Steam edition of Trine off my drive and reinstall it from scratch. I also installed a non-Steam version of the game (thank you again, Joel) just in case the issue happened to be with Steam, and simply to have another copy of the game running from another location on the drive.

The second thing I did, on a crackpot theory, was to uninstall PhysX (v.9.09.0814, according to my control panel), reboot, and then reinstall using the 9.09.0408 version included in the Trine Steam folder.

Also--in theory--Windows had been instructed not to automatically restart in the event of an error in hopes that it would spit out an enlightening error message. So the goal of each test I ran was to either get a useful error message out of Windows or successfully run the game for 45 minutes to an hour. (Since the longest it's gone without crashing so far has been 30-40 minutes, I figured that was a good goalpost to aim for.)

Test #1:
Steam version, exact same settings as prior:
- windowed mode @ 1366x768
- very high graphics settings
- AA and antistropic to max
- everything but EAX & multiple controllers enabled
- vertical sync on

Result: computer shut down & restarted at three minutes into the game, just after beginning the wizard solo sequence. No error message (curses!).

Then the non-Steam version, exact same settings as above.

Result: seemed to happen faster this time, if anything...restart/shut down again, no error message, just after starting the thief solo sequence.

Ruh-roh.

Test #2:
Steam version, same settings as above but using the console options Joel suggested to limit FPS.

Result: ...success? I successfully made it to my stated goal of an hour of continuous play--from the intro up to Crystal Caverns.

I did notice at a few points--seemed to happen mostly when lots of the wizard's objects were on the screen, if that makes any difference in terms of code--where the game seemed to "hang" momentarily. A split second at most, just enough to be noticeable. Also, with FPS limited my GPU stayed much cooler--around 65C with the fans going at 45% (compared to 75C, fans at the same speed, without limiting FPS).

So...here's my current theory. I honestly don't know what kind of FPS the 4870x2 was pulling in, but I'm willing to assume it was less than the 6970. Something about this game and uncapped FPS, when combined with a newer card, causes a hard shutdown. Why? ...Absolutely no idea whatsoever.

Am I willing to declare "victory!" on this? In all honesty, no. I've thought I had the solution to this thorny problem far too many times at this point to say that I've hit on the answer. I will say this, though...current plan is to take the PC back into the shop tomorrow to get it a new HDD. I'm going to try and get through the rest of the game before then.

If I can make it all the way to the end credits without the game shutting down (I've just hit "Forsaken Dungeons" now), then I think I'll stamp a "victory!" on this and move on. We'll see how it goes!

Edit: Yup...spoke too soon. Mid-way through "Forsaken Dungeons" with the FPS capped, boom, shutdown and then restart. AAAAARGH.

[TyraaRane]

On second thought...I just had the computer do a hard shutdown and reset with another game (FEAR 3).

I officially don't know what to do and am at my wit's end, but it's not an issue with Trine alone. I'm sorry to bother you and make you think your game was frying my computer.

[AnnularVoid]

Disable the network card in BIOS and disconnect all USB devices. Since you have a non-Steam version, that shouldn't be a problem. See if it makes any difference.

[fb_joel]

Sounds very annoying/severe. These can be very difficult to solve - I had similar kind of issues with my home computer, it would randomly reboot when doing certain things, e.g. playing Borderlands. Tests tended to go fine, although 3DMark06 did hang a couple of times (out of a dozen or so tries). After a few days of trying to figure out what was wrong, changing the PSU and trying different RAM configs and removing the GPU and using the integrated one, and having no success, I took the parts to the store and they ran heavier tests for two days (they had trouble getting any problems at first). Ultimately it turned out to be a problem with the motherboard (Asus M4A785TD-V EVO) and the RAM - apparently some incompatibility. I had some OCZ-1600 DDR3 RAM in before and when it was switched to a generic brand, the problems stopped.

So if you have any different RAM around, that's worth a shot (or just using one stick or so). The P5Q Pro doesn't seem to have an integrated GPU but you could plug in the old 4870 and see if it is still able to run the games flawlessly (just to see if the problem has become generic).

But yeah these are really difficult. In the case of Trine it's not so bad because you can complete the game quite quickly, but of course if it happens with other games then I guess a proper solution is needed...

[TyraaRane]

Very annoying/severe indeed!

I don't have any spare RAM or PSUs to throw at it, unfortunately. What I have done is pop the 4870x2 back in, and while I'm not entirely willing to say "that fixed it" (since every time I say that my PC apparently takes it as a challenge ), I was able to run Trine at the same settings as always for about an hour until I had to quit for RL reasons. No shutdowns.

So it appears that my PC is for some reason happier with a GPU that runs hotter and draws more power than the new one. Go figure! I'm still going back and forth with the shop; with the 4870x2 still in the system I'm running Prime95 for a few hours to see how that does. (Although in the interests of making this as hard to diagnose as possible, every test we've thrown at it so far it's passed. I ran FurMark last night with the 6970 for over an hour without anything happening.)

Sigh. We'll figure this out eventually, but I foresee it taking a while.

[fb_joel]

Yeah sounds like it's one of those mystery things... I suggest checking out 3DMark06 btw, it's still a good test for these purposes even if it's a bit old. Not much else I can suggest anymore.

[AnnularVoid]

Don't ignore my advice. An Ethernet adapter can be doing this, especially when chimney offloading is enabled. Some poorly written USB device drivers can cause it as well, and they can introduce a lot of latency, too.

[TyraaRane]

Don't ignore my advice. An Ethernet adapter can be doing this, especially when chimney offloading is enabled. Some poorly written USB device drivers can cause it as well, and they can introduce a lot of latency, too.

I'm not ignoring your advice--it's just that I've had nothing to add to this thread, really. The shop's been running the tower with just a monitor, keyboard and mouse attached--their own, not mine, which means none of my USB devices are at fault. (I don't have that much plugged in, USB-wise, anyway--keyboard & mouse and a set of speaker sticks. That's it. Latency has never been an issue for me until the HDD started to fail, either.)

No Ethernet cards are involved here, btw. I have a wireless adapter and that's it. I assume the shop still has it on, but probably not connected to any network.

@Joel - the shop has run 3DMark at least once that I know of, in the initial round of tests. It passed, since apparently my computer is still okay with diagnostics. (Prime95 passed too, although with some high core temps that suggest it's time for new thermal paste...) Games? Not so much. I tried running 3DMark here at home, but my monitor won't support the minimum resolution for it, so it refused to run.

The good news is we are making progress...sort of. I talked them into running the non-Steam copy of the game (which I decided to hang onto for this purpose--it's a better option than asking them to log onto my Steam account!), and they were able to duplicate the shutdown/restart behavior. When last heard from they were testing the PSU again, since this still seems like PSU-related behavior.

I asked them to run some more diagnostics on the GPU, too--since we've so far only been able to duplicate this with the 6970 installed, that sort of points to it as the culprit. (Particularly since it's the only new thing that's been introduced to the system.)

Hopefully we'll have more to go on once Monday rolls around.

[fb_joel]

Well at least it's good they've managed to replace it. Hopefully they replace it with other games/tests too...

[TyraaRane]

Good news, everyone! */Professor Farnsworth voice*

While I'm still afraid of jinxing it again, I think we can consider this issue finally nipped in the bud! The shop stuck a new PSU in (a Corsair 650W--replacing the 750W Thermaltake), and neither they nor I have been able to get the computer to spontaneously shut down/restart since. Not for want of trying, either...I played Trine for over four hours last night to see if it would shut down. (Which means that I finally got to finish the game, too! Added bonus!) So...it looks like the PSU may have been the culprit all along.

We're still not sure why the old PSU developed a sudden case of the herp derps. One theory is that the new GPU has a digital power management feature that the old one doesn't, and this might have somehow unearthed a pre-existing or pending issue. Other theories involve voodoo hexes and poltergeists. Not sure which is more likely.

Anyway...next and final step is a new hardrive to replace the failing one, and then we'll be all set. Thanks again for all your help, Joel.

Edit: Oh, and I should add...the shop guys really liked the game, too!

[fb_joel]

That's great to hear (I hope it didn't make a sudden return in this time though!). These are always difficult issues but it's nice to hear success stories every once in a while too.