AMD's Failure Rates, Part Two (2015 Edition)

TaintedSquirrel

[H]F Junkie
Joined
Aug 5, 2013
Messages
12,678
Previous installment: AMD Failure rates

We're back again, this time April 2015 to October 2015.
Cartes graphiques - Les taux de retour des composants (14) - HardWare.fr

The numbers:
  • R9 270 2.83%
  • 3.13% R9 270X
  • 0.00% R7 370
  • R9 280 4.62%
  • 2.25% R9 280X
  • R9 285 3.39%
  • R9 380 2.06%
  • R9 290 8.71%
  • 8.88% R9 290x
  • R9 390 3.60%
  • GeForce GTX 660 0.00%
  • GeForce GTX 760 1.36%
  • GeForce GTX 950 0.54%
  • GeForce GTX 960 1.53%
  • GeForce GTX 970 2.30%
  • GeForce GTX 980 1.71%
  • 3.53% GeForce GTX 980 Ti
  • 3.63% GeForce Titan X
AMD: 3.947% avg
Nvidia: 1.825% avg

Once again, AMD's failure rates top double Nvidia's. Notably, the numbers are down for both brands.

A poor showing from the 290 and 290X, which is unsurprising.
You can see MASSIVE improvement on Grenada thanks to improved board/cooler designs from the AIBs.
 
Last edited:
Interesting, but I note that most of the sample data that gives AMD a bad rep is made up of Sapphire cards. My experience with Sapphire cards has been disgusting, far worse than indicated by the hardware.fr results, whereas other AMD brands haven't been nearly as bad. I wonder what their failure rates would look like with Sapphire excluded...
Also interesting to see Inno3D perform so poorly, and Gainward so well (my Gainward card is a very tacky low quality product).
 
they lack some info there. The 290X shows only 2 listings for failures which inflates their avg greatly, where as the 980Ti has 5 where although still less than the 290X they weren't that far behind for their 2 worst. Also Not sure how anyone can blame AMD or Nvidia when some manufacturers list 0-very small issues (ASUS) to other that have issues across the board (Sapphire and Inno).

Maybe they should get definitive info on what failed. Sure a great deal of those are users ocing too heavily or in poorly ventilated cases.

I guess it makes it easy to tell which vendors to stay away from: Sapphire, Inno, MSI (debatable).
 
hmm how strange. Failure rates at my place are AMD: 0.00% avg. NVidia 80% avg. Statistics = awesome!
 
hmm how strange. Failure rates at my place are AMD: 0.00% avg. NVidia 80% avg. Statistics = awesome!
It's based on minimum 500 units sold per GPU. The numbers are also very similar to last year's, which suggests a trend. The mining excuse won't fly this year; AMD has a problem.
~8% failure rate on Hawaii is a travesty.
 
Too bad all the EVGA GTX970's with heat sinks that didn't fit or the 3.5 GB fiasco don't count as failures. Might even out the numbers a little. /fanboyrant :D

Either way, under 4% for either brand is still pretty good. Means you've got a >96% chance of getting a solid card from either brand. Pretty good odds if ya ask me.
 
Too bad all the EVGA GTX970's with heat sinks that didn't fit or the 3.5 GB fiasco don't count as failures. Might even out the numbers a little. /fanboyrant :D

Either way, under 4% for either brand is still pretty good. Means you've got a >96% chance of getting a solid card from either brand. Pretty good odds if ya ask me.
AMD's numbers would inflate if the Fury X were included (pump issues).
 
better reliability than hdd going by backblaze which seem to be in the 5-30% range other than hitachi which is mostly 1-3% with some models reaching 8%

seem to see a lot more gigabyte 290\390 cards go back for rma on au forums than other brands
another thing that would contribute to 290 rma rates is the number of them used for mining
 
Pretty sure these results only cover the 12 month warranty period rather than lifetime which may explain Backblaze's stats. It also doesn't cover faulty hardware that goes unreported and isn't returned to the retailer, which would inflate all the statistics by a certain amount. My experience would suggest that 2% average motherboard failures <12 months is on the low side, but it'd certainly grow sharply from there. I'd probably put the industry average in double digits if you went as far as, say, 4 years.
 
AMD's numbers would inflate if the Fury X were included (pump issues).

Not by much. The EVGA issues were on all their cards and the 3.5 issue was on all 970's. That'll far outweigh some bad pumps on the Fury X that wasn't that widespread. My fanboy-fu is stronger than yours. :p

I was mostly kidding anyway. I dont think failure rates are a problem with either manufacturer really. Like somebody else posted, its a hell of a lot lower than hard drives and probably motherboards too.
 
Motherboards and graphics cards are the two components I find most likely to be faulty on a new build and I'd also say are the most likely demises for older PCs. The average stats assuming all brands equal (which of course they won't be) from the stats posted in that article put them in this order:
Motherboards 1.99%
Graphics cards 1.91%
Power supplies 1.22%
Mechanical storage 0.86%
Solid state storage 0.65%
RAM 0.60%

I imagine CPUs don't make the list because they are so rarely returned.
 
Even though I know it kinda pointless to say this here but I haven't had a single AMD product fail that I didn't kill by pushing an OC too hard or by crushing a die...
 
Of the 18 AMD cards I've owned, 4 have been faulty, 3 of which I returned, one I didn't bother. (Two DOA Sapphire 4GB HD5970s, a DOA Sapphire HD3870 and the card I didn't return was an Asus HD3870 which had an odd software issue where it corrupted the OS of any machine it was installed in). I also returned the 290X not because it was definitively faulty but because it was incompatible with my UP3214Q. The Sapphire cards were the more exciting ones, one of the 5970s produced an intense 'electronics burning smell' as soon as the PC was powered up, and the HD3870 worked for a couple of days before sparks started coming out of it.
I do also have to concede the one working HD3870 was destroyed by having its die crushed trying to fit an Accelero S1 cooler to it, so that made 0 for 3!
 
Bad drivers, 290x black screens that a vbios update can fix, wake from sleep issues, sapphire's infamous coilwhine are the key ingredients why these are being called failures. XFX had some serious VRM heat issues on the DD with the 290, reference cards are louder than a jet engine. I can go on and on. What AMD needs to do (like Nvidia already does) is set a more strict set of guidelines to what components can be used to build them. I know this sounds stupid but take a look at Nvidia's box art and tell me what Nvidia AIB retail boxes have in common. The same logo branding. AMD would let the AIBs draw smiley faces with marker that's how lenient they are.

They need to take a page from Nvidia's book and maybe this new 10 year Nvidia veteran AMD hired will help with that.
 
nvidia gave the green light for them to make the non-founder as shitty looking as possible. XD
Wouldn't surprise me if the non-founder edition have loose screws and unplugged fan pin. Haha
 
As others have pointed out I have a feeling the high failure rate on the 290s in particular was because they were used for mining. Running a card 100% 24/7 is bound to cause more failures than a card that sits idle most of the time.

It's anecdotal but I've had one graphics card ever that failed within 1-2 years, and that was the one card I did folding on (running it 100% 24/7). That could just be a coincidence but it was enough that I won't ever do any GPU folding/distributed computing again. Meanwhile all my cards that I didn't fold on tend to last 5+ years (in secondary PCs, given to family members, etc). My old X850XT (AGP) is still running in someones PC and I bought that card more than 10 years ago.
 
been using AMD for 20 years and I have had only 1 card fail during that time. Those statistics cannot even be remotely close to accurate as they don't differentiate between the GPU and the rest of the more crap components used to put it together.
 
It's based on minimum 500 units sold per GPU. The numbers are also very similar to last year's, which suggests a trend. The mining excuse won't fly this year; AMD has a problem.
~8% failure rate on Hawaii is a travesty.
Why won't the mining excuse fly? Ether mining is going strong on AMD cards. On top of that you have huge number of second hand cards being sold from the mining craze of 2013-2015
 
XFX is also called Pine... and no they were not included...

I have a sapphire 270x works fine i like it better than my msi which has 2 dying fans on it...

my XFX 280x was DOA and killed a mobo when i put it in... xfx replaced the card asus replaced the mobo replacement xfx card works fine...

my amd r9 m375 in my laptop is a mess with drivers...
 
I'm more interested in seeing failure rates for the AIO on the Fury cards. The fact that it's AIO is a huge problem for me, I prefer air over AIO/CLC any day
 
I mean shit, my 7870's are still kicking in CF. They're basically R9 series lolol.
 
Could it be since Nvidia does not have the long term driver support that AMD has, their cards are retired from service at a faster rate and not reported as faulty? Not trying to start a flame war, just posting a legitimate possibility.
 
nV and AMD driver support time lines for older gens are about the same.....

Now if you are talking about driver optimizations for specific applications that is different....
 
The only card that died on my was my 2900 pro but I blame it on EU hipsters and their RoHS ecoterrorism
 
  • Like
Reactions: N4CR
like this
All of this fear over anything AIO water...

"Oh nyo! The Fury X would have much higher failure rates because of the AIO!"

not really. Thats like saying a graphics card with 3 fans would have a higher failure rate than a graphics card with 2 fans. It has a higher failure POTENTIAL, meaning it has more points of failure than the others, but potential does not translate to real-world failure rates.

I've built literally hundreds of systems with AIOs and I've never seen a pump fail. I've seen a skrewdriver put through a radiator by an idiot customer, does that count?
 
There were lots of issues with the pumps when the Fury X was first released, to my recollection, so they're probably alluding to that. That aside, however, I have seen a few stories of leaks in AIO systems through no fault of the user. A faulty graphics card is one thing, but a water leak onto other components is quite another.
 
Under a 5% failure rate for most mass manufacturing is considered a good thing and a goal that is strived for but often is not hit. I have had both a nvidia card die on me and a amd card die, but I had both for many years of use and it was time for a upgrade anyway.
 
Previous installment: AMD Failure rates

We're back again, this time April 2015 to October 2015.
Cartes graphiques - Les taux de retour des composants (14) - HardWare.fr

The numbers:

AMD: 3.947% avg
Nvidia: 1.825% avg

Once again, AMD's failure rates top double Nvidia's. Notably, the numbers are down for both brands.

A poor showing from the 290 and 290X, which is unsurprising.
You can see MASSIVE improvement on Grenada thanks to improved board/cooler designs from the AIBs.

Wow, just in time for the release of the newest Nvidia cards, go figure. :eek::rolleyes: The real question is, how is the customer treated when the card fails due to manufacturer defect. Also, how many of the "failures" are based on someone overclocking to far? Sorry but, I am not buying into your narrative but, I suppose this thread is entertaining. :D
 
As others have pointed out I have a feeling the high failure rate on the 290s in particular was because they were used for mining. Running a card 100% 24/7 is bound to cause more failures than a card that sits idle most of the time.

It's anecdotal but I've had one graphics card ever that failed within 1-2 years, and that was the one card I did folding on (running it 100% 24/7). That could just be a coincidence but it was enough that I won't ever do any GPU folding/distributed computing again. Meanwhile all my cards that I didn't fold on tend to last 5+ years (in secondary PCs, given to family members, etc). My old X850XT (AGP) is still running in someones PC and I bought that card more than 10 years ago.

I have mined bitcoins, litecoins and now ethereum -- sure mining is wear and tear but none of my cards failed. Miners bought up tons of 5870, 5970s, 7970s, at one point i had 5 290s. I bought an open box 390x from newegg that was doa, i bricked an XFX 6870 that i pushed OC too far. Other then that, ive been pretty lucky. Ive never had a card die on me past warranty.
 
I remember a few years ago during the heat of the Xbox 360 and PS3 race, the failure rate of the Xbox 360 was something like 25% and the PS3 was 10%?. I know it's a whole system, not just a graphics card, but that's a staggering percentage that people generally shrugged off (obviously some were not happy). I don't care about a 3% failure rate. I don't really even care about a 10% failure rate. Buy a card through a reputable seller that honors their warranty and there are no issues. Sometimes you'll even come away with something better in return.
 
Hmmm, over the past 20 or so years, the only AMD/ATI card I ever had trouble with was a refurb 4870x2.

And a family member had a 4850 die on them about a year or so ago that I purchased new.

And I think we had a passively cooled x1300 go bad in a Dell work computer.

Pretty sure a lot of it has to do with other crappy components (power supplies) and poorly ventilated cases.

Now if they did a controlled study with a huge number of computers that had all the same high quality components, you would see a lot lower failure rate across the board.

Just doing the "studies" on specific pieces of hardware with no care for anything else such as other parts, usage, or environment basically invalidates any and all results they come up with.
 
One thing that might happen alot with AMD users, is that they return their 290's and 390's because of the coil noise. Something that is not exactly 100% qualifying for RMA, only on extreme situations. So maybe they didn't take that into account and many partners they don't accept that to be eligible for replacement as such they get returned to the user, unless is a severe coil noise.
 
One thing that might happen alot with AMD users, is that they return their 290's and 390's because of the coil noise. Something that is not exactly 100% qualifying for RMA, only on extreme situations. So maybe they didn't take that into account and many partners they don't accept that to be eligible for replacement as such they get returned to the user, unless is a severe coil noise.

8800GTX, Crysis 1 intro videos, holy shit that thing was loud
 
I been running video cards since 2002 and still have my very first AGP card MX 420 .. I have ran Nvidia and AMD for the last 14 years and have never had a card failure as my current line up is both Sapphires R9-280 and 290X.
 
I would love to dispute this, but I believe it. I never had a faulty NVIDIA card, and I've dealt with used ones. I have only had a few AMD GPU's, almost all new, and one that was new failed within a couple days. However, that is just my experience, and it was never reported to AMD as far as I know.
 
Had failures from both manufacturers in mid-high to high end designs and use more AMD cards, so technically, Nvidia is over represented in my data.. would take this all with a grain of salt, there are so many variables.
 
Back
Top