How to diagnose gaming freeze-ups?

May 11, 2011 10:32 PM

My home-built computer seems to not want to undergo any extended load (gaming) without freezing up to extents varying by game. Where do I start?

So I've had this computer for a couple of years, and game quite avidly. It's never been perfectly stable, but never really consistent failures or the like until recently. It's possible you could track it to when I installed the GTX 470 to replace the GTX 260 Core 216, but frankly nothing I've done points to the the video card as a point of failure, nor is my memory good enough to guarantee this wasn't happening before then.

A bit more info: Any game I've played recently seems to suffer from occasional-to-frequent freeze-ups. The type of freeze differs purely by game; some (CS:S, L4D2, Dragon Age: Origins) merely pause for 1-2s every 5, 10 minutes. Others (Witcher 1, Crysis 2, a few others that escape me) simply crash to desktop.

In every case, this only seems to happen when I game. This is the only time I ever stress the video card, but it has happened with every/any version of video card drivers (several versions direct from nvidia, several versions from windows update.)

Stats:
OS: Win 7 Pro
Mobo: GA-P35-DS41
CPU: E6750
RAM: 4x1GB no-name crap
HDD (OS): 1xIntel X-25M 80G
HDD Games usually on: 2TB WD Caviar Blackt

I'm far from computer-illiterate, etc. But I don't troubleshoot gaming freeze-ups for a living, or even really support Windows, so have no idea where to start. Event Viewer shows nothing of note. Here's a sample from a witcher CTD: http://pastebin.com/TKkQdjZK

Help? Even good logging software that would show me which elements are fucking up for 1-2s every 5min in CS:S would be incredibly helpful to tracking this down. I have to believe the CS issues are related to the other game issues, which to my mind rule out HDD troubles: I can play the same server with the same map 24/7 and see freezes absolutely randomly with no texture corruption, which to me indicates the issue is elsewhere.

posted by pahalial to computers & internet (13 answers total)

Have you seen this thread?
http://www.evga.com/forums/tm.aspx?m=310750&mpage=1


posted by cosmicbandito at 10:54 PM on May 11, 2011

I would guess that your video card is either overheating when it ramps up to do 3D, or otherwise it's dying in a more fundamental manner.

If it's overheating because the heatsink is coming loose, you can remove the factory heatsink, clean off the crusty thermal compound, and re-attach it with fresh compound, making sure to seat it tightly.

You may also be able to get an after-market replacement heatsink if you prefer that.

Alternatively, something has gone utterly wrong in the card's guts and you'll have to replace it.

If I'm right about my guess as to the cause, anyway. :)
posted by kavasa at 11:09 PM on May 11, 2011

Also, this thread was posted like 3 hours before yours. Your answer is pretty much contained within it.
posted by kavasa at 11:12 PM on May 11, 2011

From cosmicbandito's link it sounds like there are a bunch of wonky / defective cards out there - can you get it exchanged?
posted by Meatbomb at 11:23 PM on May 11, 2011

My bad, I searched for similar threads and did not actually browse today's threads. I'll take a look at all this tomorrow - for now sleep beckons. A priori it seems to be a different issue, although the whole superfetch thing seems a possible culprit.

That said, I'm really looking for some way to log the card's EVERYTHING so as to be able to troubleshoot this myself - is there really no nvidia --log=DEBUG setting (or equivalent app) for windows here?
posted by pahalial at 11:36 PM on May 11, 2011

Also, cosmicbandito's link is interesting and I'm still paging through it, but I should note that none of the crashes to date have been hard crashes - at worst crash to desktop with persistent audio bugs.

FWIW, I've also dabbled at running the bitcoin GPU-accelerated program overnight, and I *felt* that thing start up and it was still going strong in the morning. Haven't really done it often, but I figure 9 hours of CUDA duty is about as good a GPU stress-test as any these days.

It is an EVGA card though - so I guess I'll try running the apps mentioned in crashbandito's link and seeing if they cause any issue, and if so just refer to evga. Still really looking for anything to give me the user deeper insight though!
posted by pahalial at 11:43 PM on May 11, 2011

This is a long shot, but I recently had a Gigabtye mobo where the stock thermal paste on the north and south bridges had somehow turned into a gritty sand-like substance. The heatsinks on those parts were loose to the touch.

I removed the heatsinks, cleaned off the gunk, and re-applied new high quality thermal paste, and the system went from super-flaky to rock solid.
posted by IvyMike at 1:36 AM on May 12, 2011

Could be overheating. Could be drawing too much power. Could be you're hitting a key combination that switches keyboard languages (English to Spanish, for example).
posted by enamon at 4:11 AM on May 12, 2011

The idea about the chipset heatsinks is a good thing to try for those kinds of issues. Those little fokkers get hot. Especially as you mention that the vid card will pass a CUDA stress-test, and that sometimes you will have audio glitches. Points to possibly an I/O problem.

Another possibility is that it is simply the audio chip on the board, and disabling it and replacing it with an add-on card will solve the problem.

I too am disappointed by the lack of real, deep down troubleshooting/logging ability on desktop PCs. Differential diagnosis sometimes can take you only so far.

(Although memtest 86+ works pretty damn well- I've had it catch memory issues that even manufacturer memory checkers did not. Might want to give that an overnight run.)
posted by gjc at 4:54 AM on May 12, 2011

I have that card, and its a beast. Is it possible you don't have adequate cooling? Overheats can cause freezing.
posted by tremspeed at 7:21 AM on May 12, 2011

Second the memtest reccomendation. Run that for an extended period, 8+ hours. Low quality RAM can show up as the most vexing issues. Also does that motherboard have temp sensors? Set up the temp monitoring software in the background and keep an eye on it through the gaming session until the problem comes up.

Finally, what kind of power supply? That card is rated for 550w. If yours is underspec or low end, it could be either not supplying enough power or experiencing variable voltages causing problems.
posted by T.D. Strange at 7:25 AM on May 12, 2011

Is the freezing coupled with sudden HDD activity? If so it could be SuperFetch moving blocks of memory about. Try disabling that service temporarily in your services console to see its absence helps. (Plenty of threads out there for those of us that want Microsoft to make a "Gaming Mode" config....SuperFetch and Search/indexing services are way up on the list of repeat offenders)
posted by samsara at 7:56 AM on May 12, 2011

Like everyone's said, it sounds like a heat issue. The place to start is to test that theory. Run the game again with the case off, with a fan pointing into the case if you can manage it. Does the game run longer before dying? Then it's almost certainly heat.
posted by Nelson at 10:01 AM on May 12, 2011