Black Screen Crashes - GPU??

[RAD]

When I run up graphics intensive games (rig in signature, built about 4 years ago) (diablo 3, call of duty, SC2, etc) it is not uncommon for me to get crashes where the screen goes black and the sound starts looping in the background. The only way to break out of the crash is to hold down the power button on the tower until the computer resets. It used to be that these crashes might only happen once every month or two. More recently they might happen a couple of times in a few hour span.

I've extensively stress tested the ram and cpu (memtest and p95) and gotten no errors and the computer never ever crashes in day to day operations (word processing, spreadsheets, photoshop, web browsing, etc, etc). There is no overclocking at all on the system right now.

I am not sure what to blame but my primary suspect are the GPUS. My current rig is running on 2x gtx 285s. I've tried all of the solutions offered up by the game forums to deal with crashes of this nature (updated drivers, enforcing lower frame rates, compatibility settings, etc) no luck. I've also tried downclocking the GPUs and greatly increasing the reactive fan speeds to keep them cooler. Temperatures never climb out of the 60s when the crashes are happening. I am stumped.

Is it time to just move on to a better GPU solution? If so, any thoughts on which direction I should go (or should I ask in the graphics card sub forum)? On the other hand, I might start pouring gasoline on the rig if it keeps crashing after putting a new GPU in :/

I am not biased towards AMD or NVIDIA. What price point am I looking at to get a good jump in performance over my current set up? I'd be perfectly happy with these GPUs as far as all other aspects of performance go, if they weren't crashing (if it is in fact the GPUs at fault).

With the D3 xpack around the corner and my preference being to play on hardcore, you can imagine full system lock ups are not a great thing to have happening.

 

[Dangman]

Have you tried just trying out the GPUs individually with the second card physically removed?

 

[RAD]

Have you tried just trying out the GPUs individually with the second card physically removed?

Since the crashes are so erratic, it has been tough to nail down what exactly is causing them. I will give that a try this weekend, though, since the crashes are happening more frequently now. The switch to being SLI was actually something I did about a year ago, the first GPU was in there since I built the machine. I will try running it on just the second 285 for a couple days and see if I get any crashes.

Honestly, if it is that first 285, at this point I will probably go with a new GPU solution regardless rather than buying another 285 to SLI, but it would be good to know for sure.

 

[RAD]

So I had time to sit down with the machine for 4 or 5 hours today and put it through it's paces on just the second 285, with the original one pulled out. Got no crashes for the duration. Is there any way beyond just playtesting the box for another week or so to definitively confirm it? Also, any residual value to a 4 year old gpu that throws crashes, or is it just scrap silicon at this point?

 

[Dangman]

So I had time to sit down with the machine for 4 or 5 hours today and put it through it's paces on just the second 285, with the original one pulled out. Got no crashes for the duration. Is there any way beyond just playtesting the box for another week or so to definitively confirm it?

Run furtest on it for a few hours. Though I highly highly recommend the playtesting route since it's more than likely not going to kill your GPU in the process.

 

[cortexodus]

So I had time to sit down with the machine for 4 or 5 hours today and put it through it's paces on just the second 285, with the original one pulled out. Got no crashes for the duration. Is there any way beyond just playtesting the box for another week or so to definitively confirm it? Also, any residual value to a 4 year old gpu that throws crashes, or is it just scrap silicon at this point?

Having tested only one of the cards, I wouldn't immediately put the finger on the card you didn't test yet. I would definitely suggest you follow dangman's suggestion and continue play-testing. However, if you're unable to replicate a failure on the card you have in place after a few more hours, try putting the other in and see if goes belly up on you.

I am always loathe to call out a component failure without a positive diagnosis. That sucks that your system is giving you the dreaded "inconsistent issue"

 

[RAD]

Well another week of testing and not a crash to be had since pulling the elder of the two 285s out of the system. I guess tomorrow if I find time I will open it back up on my bench and swap which GPU is in and see if I start getting some crashes.

 

[Dangman]

Good luck.

 

[RAD]

Never been so happy to see a computer crash. 2 hours into a session after the GPU swap back to the old GPU got a crash. Huzzah, a positive result. Now to decide to buy another cheap 285 (looks like used at about $60 on ebay is not unusual) and go back to SLI'ing them, or to find an modern generation gpu in the $200 ish range and modernize the system.

Thoughts?

 

[cortexodus]

Never been so happy to see a computer crash. 2 hours into a session after the GPU swap back to the old GPU got a crash. Huzzah, a positive result. Now to decide to buy another cheap 285 (looks like used at about $60 on ebay is not unusual) and go back to SLI'ing them, or to find an modern generation gpu in the $200 ish range and modernize the system.

Thoughts?

Nice to have the issue nailed down with a fair degree of certainty As for replace vs. upgrade *shrug* that's pretty much a budget call. I definitely wouldn't spend $60 on a card that old at this point, but that's just imo.

 

[Dangman]

Never been so happy to see a computer crash. 2 hours into a session after the GPU swap back to the old GPU got a crash. Huzzah, a positive result. Now to decide to buy another cheap 285 (looks like used at about $60 on ebay is not unusual) and go back to SLI'ing them, or to find an modern generation gpu in the $200 ish range and modernize the system.

Thoughts?

I'd just buy a new GTX 760 for $240 and call it day.

 

[RAD]

Should add that it doesn't have to be Nvidia, if AMD is winning the P/Q war this time around, haven't really been paying attention.

 

[Dangman]

Should add that it doesn't have to be Nvidia, if AMD is winning the P/Q war this time around, haven't really been paying attention.

No, at around that price point, your best bet is the Nvidia GTX 760. The crypto-currency craze is screwing with AMD card pricing. Should provide enough of a performance increase over your GTX 280 SLI setup.

 

[RAD]

So interesting new update. I decided to contact EVGA's support regading the problems I have been having with this card, thinking that maybe they would offer some sort of small credit towards the purchase of a new card. Their customer support instead said the card was eligible to be RMA'd (apparently it was under their old lifetime warranty system since it was bought in 2009).

In 2011 they switched to a limited year warranty system (usually 3 year) but since I bought this one in 2009 it is under the old "lifetime" system.

They are starting the RMA process since I can provide proof of when I bought it (thanks Newegg's invoice system). Will be interesting to see what they send back out to me (assuming they actually approve the RMA, which they said can take 1-2 days), I can't imagine they have a bunch of 285s sitting in a store room somewhere.

 

[cortexodus]

Sweet. Sit tight, run that RMA through and see what arrives!

 

[RAD]

So the saga continues...

• 
  Sent the card back in (had to pay shipping, ~$15).

• Hear nothing from EVGA for a few days, log back into their support site and see that they are having problems with microsoft based emails so can't contact me.
  Change my email with them to a gmail email.

• Get a message saying my card was too dirty and they want to charge me $35 for cleaning my bad card.

• Note - Apparently they will be sending out a refurb 650 ti 2gb as a replacement. Not to be ungrateful, but according to quick research this is basically equivalent to the SLI'd 285s I was running but sucks much less power, so basically side grading for lower power consumption on a refurb gamble at this point.

• I balk at the fee because at that point that would put me $50 into a card that I know is bad, after paying full MSRP back in the day (also puts me around 20% towards a 760).

• On principle alone this is objectionable to me so I decide to commence the run around to dispute.

• Email tech support, they say to call on Monday. Call Monday, get bounced from base level CS, to escalated CS, to a message to the RMA manager. RMA manager has escalated CS get back to me and let me know they are going to waive the fee this one time (with a cautionary word about cleaning my case out more often) .

Will be interested to see how the 650 ti 2gb performs and holds up. I've never had a lot of faith in refurbed anything, but fingers crossed it is stable.

Here is the picture they sent me to indicate the card was too dirty:

Now there is no doubt this card had accumulation of dust and grime for being in use for quite a few years, but it still seems wrong to me that they were trying to stack another charge onto a bad card that has been bad since day one (I have proof in their support system that it was crashing even when immaculately brand new. At that time they claimed it was drivers and/or software and wouldn't execute a RMA.)

It also seems crazy to me that in order to execute their warranty without hidden fees the card has to be in brand new clean condition. Am I off base on that thinking and did I just turn into that jerkwad customer no one wants to hear from? Am I being an unreasonable cheapskate at this point? To me it really isn't about the money, or at least that is what I am telling myself right now.

 

[cortexodus]

It also seems crazy to me that in order to execute their warranty without hidden fees the card has to be in brand new clean condition. Am I off base on that thinking and did I just turn into that jerkwad customer no one wants to hear from? Am I being an unreasonable cheapskate at this point? To me it really isn't about the money, or at least that is what I am telling myself right now.

I'm not trying to sound like a dick, I truly am not. That having been said, I cannot even fathom how you could send the card to EVGA looking like that You're lucky they didn't just deny the RMA outright. The 650Ti (refurb or not) on its way to you for the price of $15 is a boon and you should be thrilled.

 

[RAD]

I'm not trying to sound like a dick, I truly am not. That having been said, I cannot even fathom how you could send the card to EVGA looking like that You're lucky they didn't just deny the RMA outright. The 650Ti (refurb or not) on its way to you for the price of $15 is a boon and you should be thrilled.

You don't sound like a dick at all, I appreciate the feedback from an impartial third party. I will definitely do a better job of keeping dirt accumulation down going forward. Looking at it in person when I had it, it never struck me as that heinously dirty. That picture is undeniably ugly, though, I guess. At the end of the day I think I was just frustrated with the long term problems I've had with the card.

 

[chibe]

It almost looks like it's rusted onto the card. Im guessing cigarette smoke?

 

[RAD]

It almost looks like it's rusted onto the card. Im guessing cigarette smoke?

No, completely smoke free house. That picture with the heavy flash makes it look worse than it was. I take the rig apart and blow it out once every couple of months in addition to that. Strange part is, the second 285, which has been in the rig for almost two years now as well, has none of the accumulation that one had.

 

[cortexodus]

No, completely smoke free house. That picture with the heavy flash makes it look worse than it was. I take the rig apart and blow it out once every couple of months in addition to that. Strange part is, the second 285, which has been in the rig for almost two years now as well, has none of the accumulation that one had.

Based on the dust pattern, I'm guessing that card was directly underneath a very large air-cooler.

 

[RAD]

Based on the dust pattern, I'm guessing that card was directly underneath a very large air-cooler.

That is indeed the case. On all future clean outs I will pay careful attention to making sure an accumulation like that does not occur again. Thanks again for the reality check. I owe EVGA an apology email for coming in hot on the phone call to CS today.

 

[RAD]

Had a pretty package waiting for me on my doorstop today. A shiny, in like new condition, EVGA GTX 580, 1.5 gb. More than twice the horse power of the bad card it was replacing under warranty. Yeah it is technically still a few year old technology, but a big upgrade over the card it was replacing.

Consider me an EVGA guy for life, that is truly epic warranty service. I am pretty blown away at there actually being a company out there that honors their warranty to this degree.

 

[Dangman]

Well now the big question: Are you still seeing black screen crashes?

But damn man, CLEAN YOUR PC! If you haven't already, get a DataVAC like this:
http://www.amazon.com/dp/B001J4ZOAW/?tag=extension-kb-20

Best damn purchase I've ever made for cleaning my PC.

 

[RAD]

Well now the big question: Are you still seeing black screen crashes?

But damn man, CLEAN YOUR PC! If you haven't already, get a DataVAC like this:
http://www.amazon.com/dp/B001J4ZOAW/?tag=extension-kb-20

Best damn purchase I've ever made for cleaning my PC.

No black screen crashes since pulling out the bad card. I tested with the other 285 in place and got no crashes, put the bad card back in and got crashes, then RMA'd the bad one back to evga and got the 580 in return.

I didn't have any actual heavy use impact for the 580 tonight, not a ton of free time. Tomorrow I am really going to give it a warm up lap or two, but so far its behaving like a dream.

I actually own that exact datavac. I blew out every last speck of dust before mounting the 580 tonight. I sincerely have no idea how such an accumulation got onto the back of that 285 but I will be doing a bi monthly tear down and complete blow out to prevent it every happening again. I have the computer up off the floor on a stand and repalced the carpet in that room with tile, so dust is much less of a factor than it was prior.

 

[RAD]

A question: I now have a fully functional 285 on hand (the one in the SLI config that wasn't bad). Would it at all be a good idea to put that into the rig as a dedicated physx card or is that a pointless waste of power?

EDIT: to answer my own question, a quick google seems to indicate a possible consensus that it not only would be a waste of power, it might actually be worse than just letting the 580 do all the work itself.

 

[Dangman]

A question: I now have a fully functional 285 on hand (the one in the SLI config that wasn't bad). Would it at all be a good idea to put that into the rig as a dedicated physx card or is that a pointless waste of power?

Pointless waste of power IMO.

 

[RAD]

On an amusing aside (probably sadly speaks to how out of touch I have gotten), I was legitimately confused about what to do with the SPDIF wire from the mobo, had to google to discover that it was no longer a necessary component.