GPU crashing computer - completely stumped

[tenks]

Hey Guys!

I have a serious computer issue that is driving me crazy. I consider myself pretty proficient with computers and I haven't really come across an issue I haven't been able to trouble shoot or fix (usually software related). But this one has got me stumped and I was hoping maybe some of you real experts out there can help me solve this or at least confirm my suspicions.

(Wow, I spent about 20 minutes typing out this post all nice and neat and of course right when I'm done previewing it for spelling I get the BSOD I'm having problems with...)


I'll try to make this one shorter and more to the point.

Dell Studio Xps 1647 laptop
Intel Core i7 620m - 2.66ghz (arrandale)
Ati Mobility Radeon HD 4670 1tb
840 Evo 250gig + 500gig 7200rpm momentus

So the issue is this. After gaming regularly for maybe 2-3 hours a day, for 2-3 weeks my computer will randomly get heavy artifacting, lock up and then crash. When I reboot the usual black dell logo is going crazy with artifacts and the screen is primarily yellow. If I try to boot back into the desktop, it locks up after I login and retarts the reboot process. The BSOD flashes too fast for me to see what it is. Upon further inspection in Safe mode, it says my video card is missing as well as the drivers/or it can't load.

The first time this happened, I took my laptop apart. Cleaned out all the dust in the heat pipes and cooling fans. Cleaned the CPU and GPU and reapplied new thermal paste. Nothing. My computer was dead for literally 16 hours or so, with the heavy artifacting appearing on every screen and only being able to boot into safe mode (also artifacting).

Then randomly after trying some more, it boots perfectly normal like nothing ever happened. I can go game, crank up the temps to 80+, completely rock solid stable. Another 3-4x weeks go by of absolutely no issue, 100% stability, then bam, this artifacts creep up and the whole process repeats itself. It's dead for 4-12 hours, I get sad, I think my GPU is dead, then bam! back to life like nothing happened, AGAIN!

This issue has happened about 6/7 times in the last 5 months where it first popped up. Only gaming in Dota 2 (to be specific) has caused this issue to happen. No other game I've played has triggered this.

I've run memory tests, I've run GPU tests and stressed/benched the video card (after it fixes itself) and nothing bad happens! Then again, 3-4 weeks later it happens all over again.


So here are my main questions aside from trying to fix this:

1) If my suspicions are true and my GPU is dying, then how is it coming back to life and able to game rock solid, at high temps, for weeks?

2) if its heat related, again, how am I able to game hard for weeks rock solid?

3) I have read the forums and got some tools but I don't know how to interrupt the data or what to do with it. Bluescreen viewer, event viewer etc. What do I need to post for you all to see whats going on?


I remember coming across some files that appear in the crash reports a lot.

Atimpag.sys npfs.sys - And almost every time, there is always a file that's involved with the BSOD that has to do with a "miniport" be ati or ms related.

Please help, I'm at my wits end!

 

[inachu]

I had the same issues on my very first laptop by dell that used a desktop cpu in a laptop mother board.

Aside from heat issues it would help to know how old the laptop is.
If its only used for gaming then I suspect that you might need to reflow the motherboard.
there are videos on how to reflow motherboards.

What this means is that your solder joints have cracked and could be loose and reflowing a mother board in the oven could reseal them back to new condition.

Other possibilities is static shock damage. video memory damage or as was in my case I had to downgrade the video drive back to factory CD OEM version.

I found out that using the latest video driver made my video card run hot even when just using microsoft office or watching youtube and going back to basics kept the video card cooler a lot!

So my take take the easy route first. You did due diligence and cleaned out the dust so go back to original specs.

Also in system bios if the settings let share as much laptop memory to the video card as possible.

 

[Ketchup]

Hello tenks, this does sound exactly like what you would run into with overheating/throttling. So the first step taken was a good one, dust removal.

I would recommend a closer look at temps as well. CPUID hardware monitor is an excellent tool here. Run a game for a few minutes, and see if you can find some max temps (before a crash occurs).

How many cooling fans are on that unit? (the looking I have done suggests that the CPU and GPU share one fan).

Does the laptop get as loud as it did when you bought it?

What kind of surfaces has the laptop been on since you bought it?

Based on what you have told us so far, you may want to remove the cooling system from the laptop and reapply thermal paste on the CPU and GPU. Aftermarket products usually transfer heat much better than the stock solution anyway.

I don't think we need to start cooking things in the oven just yet

 

[tenks]

Thanks for the responses guys, I appreciate it.

Ketchup, I typed a lot so you may have missed, but I already have removed the cooling assembly and applied new thermal paste, actually many many times. (Can thermal paste evaporate or something? I swear I put a bunch on and weeks later its almost all gone.)


The laptop is as loud as it was when I bought it. I'm very familiar with the heating/throttling design flaws of this line from Dell back when I got it in 2010 and dealt with those issues then. (New bios update, upgraded bigger battery and upgraded to a 130w PSU from 90w).

The laptop is on my wood desk, with the bottom cover removed to improve airflow. 1 Cooling fan is attached to the heatpipe, that covers both the GPU and CPU like you said. I also have a room fan pointed at it when I game.

As far as temps go, they max out at around 84-86°C when Im gaming on the high end, but it usually hovers around 68-74°C. This is normal for my machine, as it hits these temps all the time and nothing happens. As far as inducing max temps before a crash, that's just it. I do it all the time and no crash occurs. I can't recreate the crash at all, its been random up to this point. Im sitting here typing on it right now, with firefox, gpu shark, gpu-z and AIDA64 extreme all running. Idle gpu temps are at 50-55°C but its a nice hot morning out in San Diego...

I gamed for 2 1/2 hours last night, which was right after the most recent crash. No issue.

I really feel like this is an issue that I can fix, I am so puzzled how a GPU can die then come back to life? Is it really dying?

Also, if it matters, IM running Dota 2 at minimum and low settings, low render quality, non-native lower resolution etc etc. Could it be the video memory? My ATI card has hypermemory and I thought that could be causing issues? but I dont know if it's even activated or on?

 

[Ketchup]

What drivers are you running?

 

[tenks]

I had the latest AMD catalyst drivers that supported my card, 13.9 - They stopped supporting/updating drivers for legacy cards after that.

I rolled back the driver to 13.1 in hopes of maybe thinking the new driver was the issue...No dice.

I opened my window and let some air into my room idle temps way lower 46-48

edit: I notice a lot of threads on here and other help forums where people are instructed to download certain programs and then post/paste back here the logs or dump files for others to look at/interpret. Ketchup, is this something that I can do or applies to my situation?

 

[Ketchup]

You could post a screenshot from the temp monitoring program you are using. I like CPUID hardware monitor because it provides so many valuable statics. Maybe something will jump out at us.

 

[tenks]

Ok, so this has nothing to do with like posting the crash dumps to see what files are causing the BSODS?

I downloaded CPUID and took a screenshot for you since you prefer that app.

 

[Ketchup]

Yeah, things look fine to me.

The only other thing I can think of is to try a factory reset on the laptop. Aside from that, looks like a hardware issue from what you have stated.

 

[tenks]

thanks for the help...but there has to be more to this...

 

[D2ultima]

Thanks for the responses guys, I appreciate it.

Ketchup, I typed a lot so you may have missed, but I already have removed the cooling assembly and applied new thermal paste, actually many many times. (Can thermal paste evaporate or something? I swear I put a bunch on and weeks later its almost all gone.)

Thermal paste CAN indeed evaporate if you use low quality paste or pastes known to be susceptible to pumping out. For a mobile system, Gelid GC Extreme or IC Diamond (careful, this scratches dies and such when removing) are good things to use. Arctic Ceramique is also a nice one, though I do not know about its longevity. I've had Arctic Silver MX-2 evaporate on me in quite literally 1 week. I would get 90+ deg on my CPU simply running a virus scan (due to SSDs being able to be checked quickly) and when we removed the heatsink from my CPU there was just a little colour left. You wouldn't have even known it had paste if I didn't tell you, it looked like some paint fell on it by accident.

Also, it is possible something went wrong on the motherboard. If that's the case you'll likely need a new mobo, or as someone else said to reflow the mobo (though I do not know how well that will help as I've never done it before). I wish you luck with fixing it!

Edit: Just to clarify, your problem I am 355% certain is NOT due to overheating. I have never seen an overheating BSOD; even with my CPU at 95 degrees and my GPU at 105 degrees back with my old laptop's 280M. Yes, I know, I should never play something at that temperature, but I didn't realize it was that hot at the time, and I couldn't help it even if I knew back then XD. But still, overheating BSODs --> artifact deadness is nope city.

The fact that it works after 16-24 hours means it's a mobo issue I think. I had one where my PC would randomly "shut down" for no reason at all; even if I was in the BIOS. If it started shutting down, the only way to kill it for a few weeks was to de-power and leave it off for ~24 hours. 12 hours wasn't enough to do the trick. I checked every single hardware on that thing that I could have done, and it would even shut down without an OS installed. So your BSOD & artifacting sounds like a similar problem with a different symptom.