Graphics card crashes

Pale Shaza · 5 Feb 2012 at 03:30

Hi, my graphics card (HD6850 msi cyclone oc) always crashes. Has since I got it a few months ago. I've ruled everything out including psu, driver, bios, chipsets, os etc...

It doesn't crash during games (It used to, but seems to be alright now that I underclock it. It can usually pass benchmarks no problem. The other day I ran heaven, it crashed the first time immediately but then when I rebooted it ran for half an hour straight.

It does commonly crash when streaming video's or watching downloaded movies etc... Not always, about 20% of the time, also if it doesn't crash the first time, it generally won't crash unless it powers down and powers back up.

Sometimes it'll just crash randomly, I might come back and it'll have died or I might be using ms access or something and I'll click a button and it'll just die.

My old card (HD 4650) worked fine, completely stable.

System specs:
Motherboard: Gigabyte EX58-UD3R
Processor: Intel core i7 920 @ 2.67Ghz
Memory: DDR3 6GB (corsair tri channel)
power: Cool Master gold 1200W (upgraded from corsair TX650W, it didn't solve the problem)
Graphics Card: MSI Radeon hd 6850 cyclone power edition (OC)

Rich43 · 5 Feb 2012 at 03:42

Is it the same as my problem? see: http://forums.overclockers.co.uk/showthread.php?t=18371234

Pale Shaza · 5 Feb 2012 at 14:11

No it's not. When it crashes it looks like this:

I have to force restart the computer to get it back.

Gregster · 5 Feb 2012 at 14:19

I assume you have done a complete reinstall of vga drivers? Another thing to check is the event viewer to see what problems are reported there. That will give a little more info.

Randomface74 · 5 Feb 2012 at 14:21

you were sold a card with an overclock it can't actually run at.

lots of this going around.

Pale Shaza · 5 Feb 2012 at 14:40

gregster said:
I assume you have done a complete reinstall of vga drivers? Another thing to check is the event viewer to see what problems are reported there. That will give a little more info.

I've done all of that and checked the event viewer etc... no help there.

Yes, I think that aswell about the oc, but it still crashes sometimes underclocked.

Randomface74 · 5 Feb 2012 at 14:45

I'd say RMA it, problem is they'll run it through benchmarks and if they don't get the same problem they'll make you pay to get it back. So you've been sold a dodgy product and ur basically boned.

Pale Shaza · 5 Feb 2012 at 14:47

billysielu said:
I'd say RMA it, problem is they'll run it through benchmarks and if they don't get the same problem they'll make you pay to get it back. So you've been sold a dodgy product and ur basically boned.

Yes, it's already happened once with this card a few weeks ago. I've got a new RMA and they're going to send it back to MSI when they get it. I just want to know what the problem probably is.

Alkalinepro5 · 5 Feb 2012 at 15:15

billysielu said:
you were sold a card with an overclock it can't actually run at.

lots of this going around.

All pre-overclocked cards are a complete and utter con IMHO.

Randomface74 · 5 Feb 2012 at 15:28

agreed

Pale Shaza · 5 Feb 2012 at 15:28

How come some can handle the pre oc and others can't i.e. my one? My card runs at a pretty low temperature (< 55C full load I think it was).

ALXAndy · 5 Feb 2012 at 15:36

Pale Shaza said:
Yes, it's already happened once with this card a few weeks ago. I've got a new RMA and they're going to send it back to MSI when they get it. I just want to know what the problem probably is.

I sent mine directly to MSI because of those reasons. I did not want to be charged by OCUK to "test" my card and then pay the return shipping for said broken card.

My advice to you is take more photos. Make videos, do whatever you have to and document every last bit of it.

My 6970 was seemingly fine. Until you spent an hour playing games on it. So, I simply ran Crysis benchmark until it froze and then took photos etc.

Randomface74 · 5 Feb 2012 at 15:39

it's luck. that's why it's a con.

all cards are built to meet stock specs. some come out slightly better, that's where overclocking headroom comes from.

third parties take the stock cards, apply an overclock, and usually a cooler - tactics to make their brand appear different, to get sales. Noobs see a card with a markup and ask why it's more expensive, see that the speeds are quicker, and buy that because it's better. But it's not better, it's running above its rated spec.

Personally I want to see more cards sold at stock speeds with custom coolers and "overclock at your own risk" notices.

ALXAndy · 5 Feb 2012 at 15:54

billysielu said:
Personally I want to see more cards sold at stock speeds with custom coolers and "overclock at your own risk" notices.

That would never work.

Products are made or broken on their overclocking capabilities.

You need to realise that most of the world have the "something for nothing" mentality.

ATI and Nvidia turned "Anti overclocking" a few years ago. For example the Radeon 9800 Pro had a soft blow fuse on it. As soon as you overclocked the fuse blew out letting ATI know the card had been overclocked. They then voided your warranty.

It hurt sales badly.

Then they realised that it was overclocking products people wanted. They wanted something for nothing. That was when BFG turned up selling overclocked cards with lifetime warranties.

Out of all of the cards I've seen I saw more BFG products fail than any other. That was why BFG charged a premium, as the chances were they would be replacing the card sooner or later.

Randomface74 · 5 Feb 2012 at 16:00

And BFG offered a 10 year warranty. I had three BFG cards, two were replacements, they all died, then BFG vanished. Now they make it harder to get a replacement.

Rich43 · 5 Feb 2012 at 18:19

My 4850 did that, and when I got fed up with that happening all the time, I bought a 5850, but the problem morphed itself into the one I have in my thread (http://forums.overclockers.co.uk/showthread.php?t=18371234) and never really went away since.

Loads of people get it with all sorts of cards and motherboards and I have no real fix

mmj_uk · 5 Feb 2012 at 18:24

That looks exactly like the grey screen of death.

Monkian · 5 Feb 2012 at 19:09

It could be that your temperatures are too low rather than too hot.

if you google "cold bug" you will find some info about it but my card suffers from it (6870)

something to do with the vrm's not liking low temps, mine has an accelero extreme plus 2 on it and runs very cool, if i power on or resume from sleep and dont let it warm up a little before doing anything gpu intensive, i will get display driver crashes and/or blue screens.

i tried lots of thing to fix it once i found out what the problem was

increasing 2d clocks made things worse
increasing voltages made it worse
i tried a few different manufacturers bios and changing from xfx black edition to a standard clock bios helped some

in the end i set set a fan profile in afterburner to turn off the fans until the gpu was over 45c, combined with a sapphire 6870 bios, now it is cured except for if i do something like resume from sleep and immediately run kombustor while the card is still cold.

Pale Shaza · 5 Feb 2012 at 20:00

Yea, that's interesting. Using speedfan my gpu when not under load is running at 33C and that's with the computer being on for the last few hours.

I'll try doing what you've done with the fan to see if that solves it.

These are the temperatures speedfan is giving me, I don't know what most of them mean, hopefully somebody can enlighten me:

Temp1: 25C
Temp2: 44C
Temp3: 43C
Temp: 2C
HD0: 27C
GPU: 33C
Core 0: 49C
Core 1: 47C
Core 2: 48C

It also says that Fan1 is running at 1116 RPM and GPU fan is running at 1309RPM. It also says Speed 01, 02 and 03 are running at 100% and 04 is at 0%.