How Often Do Cards Blow Up?

[carborundum]

Just had my first 5770 die on me.... so sad (and annoying) I spent ages trying to work out if the spangle on the screen and dramatic hash drop was a software glitch. Turns out the thing is dead.

Does this happen much I wonder? And what do you do with the dead cards?

[vdragon]

Just had my first 5770 die on me.... so sad (and annoying) I spent ages trying to work out if the spangle on the screen and dramatic hash drop was a software glitch. Turns out the thing is dead.

Does this happen much I wonder? And what do you do with the dead cards?

 You might be able to reflow, but seems to me its a memory problem

[mokahless]

Just had my first 5770 die on me.... so sad (and annoying) I spent ages trying to work out if the spangle on the screen and dramatic hash drop was a software glitch. Turns out the thing is dead.

Does this happen much I wonder? And what do you do with the dead cards?

I've personally only had one card die. It was an 4870 and it died because of me, not because of itself. I tried to do the voltage mod on the memory and fried it.
My only other dying experience was the fan. XFX was nice about replacing it. I now regularly run the thing above 100*C with no issues.
I've never had a card die on me with normal use, or Bitcoin mining.

In my experience, and from hearing about video cards dying from friends, it is pretty damn rare.

The definitive statisics, though, shall come from the first Bitcoin miner to post in this thread who has at least a cluster of 40 cards.

My suspicions are lots of dead fans and maybe 1 dead card but we'll see.

[Gator-hex]

5770 and 6770 are very hot and stressed out cards. I tend to avoid xx70s and use xx30s/xx50s

The only card I've ever had die was a XFX 6770. Cooler is too small IMHO. Then one hot summer night it died.

Get some nice 5830s instead, with a big fan and big cooler, similar price but 260MH/s.

[mufa23]

Been running a bunch of 5830s (Sapphire and XFX) for a year or so now. No problems yet. Never had a card burn out in my desktop/gaming computers either.

The only thing I have ever had die on me was some crappy G.Skill RAM that started acting up more and more as time progressed (I think it was defective from the start). And I also had a cheap PSU that went on one of my first mining rigs when I first started mining. Other then that, I have never had any components ever die on me. Even have a couple of old computer running Win95 that works fine. Guess I've been lucky.

So to answer your question, I'm going to say that it's not often that they fail. Probably varies by brand though.

Get some nice 5830s instead, with a big fan and big cooler, similar price but 260MH/s.

I'm running a bunch of 5830s at 291Mh/s 70'C using these settings in BAMT:

Code:

  core_speed_2: 900
  mem_speed_0: 300
  mem_speed_1: 300
  mem_speed_2: 300
  core_voltage_0: 1.163
  core_voltage_1: 1.163
  core_voltage_2: 1.163

  fan_speed: 70

  kernel: phatk2
  kernel_params: BFI_INT WORKSIZE=256 VECTORS FASTLOOP=false AGGRESSION=11

Used to do ~310Mh/s with the core at 950. But stuff seems to crash after a given time if it's higher than 900. You can get a little more out of those 5830s without them crashing and keeping them cool. Unless you're downclocking them or something to save electricity.

[johnyj]

My experience is that in 1-2 years VRM will die before GPU dies

[superfastkyle]

I've had a few vrm's fail in my 5830. I believe they only have a 3 phase vs 4 phase with 5850/5870 so I stay away from 5830's now

[Gator-hex]

I've never had any problems with my Sapphire 5830s but I never mess with voltages to get them to 300MH/s.
I use Sapphire ones with a huge heatsing, 5 heatpipes, and a large fan.
Keeping the VRMs cool is important and some cards with cheap coolers don't do a good enough job.
Saving money using a cheap ring cooler "XFX" is a false economy IMHO.

[mokahless]

I've never had any problems with my Sapphire 5830s but I never mess with voltages to get them to 300MH/s.

Really? I've never had to touch the voltages to get to 1GHz (321MH/s). Thing's been stable 24-7 for 24 days of uptime as of right now. My outage 24 days ago was due to a power outage.

[superfastkyle]

Well the way they usually go is they stop being stable at 1gh+ and you have to lower it to 950 then it will be stable a week or two more. You might be able to lower it then to 850-900 for a couple weeks. Then they die.

Don't buy used 5830's!

I've never had any problems with my Sapphire 5830s but I never mess with voltages to get them to 300MH/s.

Really? I've never had to touch the voltages to get to 1GHz (321MH/s). Thing's been stable 24-7 for 24 days of uptime as of right now. My outage 24 days ago was due to a power outage.

[GernMiester]

2-6870's (XFX and HIS) died a water cooled death about 6 months ago after 2 years or so of gaming and mining. Left on an old piece of tubing and it had a brittle end I never had to disconnect and forgot to replace. Dripped tiny drops from CPU onto the GPU's. It ran for I don't know how long until the non conductive coolant became conductive from contamination.  Lost about 2 ounces of coolant and 2 cards.
Around the same time a Gigabyte 5870 just stopped working.
Just today one of the replacement 6870 cards just stopped working for no apparent reason.

In different systems with different model PSU and UPS.. Paid for themselves a few times over back in the day and still scraped a few btc out until this afternoon.. New HD7970 coming to play a few games while the other cards will run till they die, profit or not.

[mokahless]

Well the way they usually go is they stop being stable at 1gh+ and you have to lower it to 950 then it will be stable a week or two more. You might be able to lower it then to 850-900 for a couple weeks. Then they die.

Don't buy used 5830's!

Mine is the sapphire that was so popular back in the day. Got it used around Feb. last year. Set it up and have been running it 24-7 at 1GHz with no overvoltage ever since with probably no more than 3 weeks total downtime. linux, no GUI.

[crash_override]

I've been overclocking video cards since Riva128's were around... never had a single one die. I guess I should consider myself lucky..? 

Seriously, it seems very, very rare to me, with people I know as well... although I admit all the non-mining cards I've owned over the years are NVIDIA cards (which have always been of superior quality IMO).

[Crystallas]

I have had a few die, and revived one by changing the caps. But none died because of mining, and I have been building systems since the mid 80s.  That's a large sample size, and I would say that killing a card because of usage is rare. Killing a card because of mild OC/+volting is rare too(because you're not going to start OC'ing something just for the hell of it, you'll likely research the results of those with deeper pockets first LOL.)

[dogie]

Ive owned >1000 cards, so a larger sample.

I've had up to 10 DOA by natural causes
I've DOA'ed up to 10 by non natural causes :p
I've had up to 10 need repair due to fan failure
I've had a handful die by overheating not due to cooling system, which weren't DOA.
The other 970 either still work perfectly or are so far behind the curve they've retired to the shelf retirement home.

[carborundum]

I found out what killed the 5770, it was...... THE CHEAP CHINESE PCIE EXTENDER!
Yes, it had some seriously sloppy soldering, loads of lines shorted on it. Had to fix it with a knife and hacksaw.

[Gomeler]

The only cards that have given me troubles are my HD 5970s. Invariably one of the cores will just die. Considering how old they are and how irrelevant they have become I just pull the card and add it to the stack of dead 5970s. I'm ready for some no maintenance ASIC machines.