AMD Failure rates

TaintedSquirrel

TaintedSquirrel

[H]F Junkie
Joined
Aug 5, 2013
Messages
12,396
Some of this data is old (2014) and from a small sample size (two stores) but since this topic rarely gets discussed, I thought I'd bring it up again... Or for the first time, as the case may be.

https://www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/
The current generation - GTX 7xx and Titans - has so far been better than any other generation to date. The .64% in the field failure rate is like a little lower than it eventually will be since the cards have not been out for as long as the other generations, but the 1.61% failure rate during our initial testing is very, very good.

Unlike NVIDIA, our failure logs show that the latest generation of AMD cards is currently seeing an increase in failure rates. The good news is that most of the Radeon R7 and R9 cards that are failing are doing so in-house, so our customers are mostly isolated from the problems. Historically, AMD has about a 50-50 DOA to field failure rate, so we are actually very happy to see that ratio shifting away from failures in the field. However, a 13.46% failure rate during our testing is really, really high and indicates that there is a problem with the latest AMD video cards.
Click to expand...

http://linustechtips.com/main/topic/255466-huge-list-of-failure-rates-for-all-pc-components/
- 2,53% Radeon HD 7850
- 1,66% Radeon HD 7870
- 10,28% Radeon HD 7950
- 7,63% Radeon HD 7970

- 2,81% Radeon R9 270
- 5,79% Radeon R9 270X
- 8,81% Radeon R9 280X
- 6,63% Radeon R9 290
- 5,58% Radeon R9 290X

- 1,57% GeForce GTX 660
- 2,57% GeForce GTX 760
- 3,16% GeForce GTX 770
- 4,75% GeForce GTX 780
- 2,91% GeForce GTX 780 Ti
- 1,33% GeForce GTX TITAN/BLACK
Click to expand...

Averages:
Nvidia: 2.715%
AMD: 5.746%

AMD's failure numbers are roughly twice as high as Nvidia's. Then we have more anecdotal sources like these, reflecting specific problems with one particular card (280X):

https://rog.asus.com/forum/showthread.php?43057-R9-280x-DirectCU-II-Top-Artifacts

https://www.sapphireforum.com/showt...-Sapphire-Toxic-Problem!-Artifacts!-New-VBios

Could this be related to the mining craze of early 2014, perhaps? GPU lifespans shortened, increasing their RMA rates later in the year and leading into 2015. Or does it represent a bigger problem with AMD's QA?

And I guess the biggest question of all, why are these numbers not more prevalent? Shouldn't there be some research done into this by bigger hardware sites with the resources to find answers. If we take the numbers at face value then it appears buying an AMD GPU doubles your chance over Nvidia of getting a failed card and that's something consumers should be aware of... and AMD needs to be held accountable for.

How much of this blame belongs on the AIB partners themselves? In the 280X's case, it looks like they were pushing the GPU's too hard (1100+ Core, 1600+ Memory) and it caused reliability problems. How did those cards reach the market in the first place?
 
AMD allows users to overclock without a real limiter on voltage. I roasted the memory on my HD 7950 trying for 1900 speeds. I think the stock speed was 1250? It was stable at 1800 so I went for broke and broke it. :) At any rate it would show as a failure and I'll be the first one to say I fucked up. I didn't send it in for warranty as I believe in being truthful.
 
In support of AMD "hardware", I bought a 290, blew a 290x BIOS and fitted a Xtreme III cooler on it with glued ram/VRM sinks, it hasnt put a foot wrong and clocks like mad.
If ever there was a time it could go wrong and when going wrong would be more devastating, that was it.

I'm not such a fan of their drivers though.
 
I would say mining craze, and it wasn't the GPUs that failed, it was the cooling solution, I had 2 MSI cards, which the fan died on within a month of being run 24/7

(On the other hand, I had gone through 7 290/290x cards that were all reference, without a single failure while mining)

I did purchase a 290x, which crashed in BF4, sapphire did RMA it without a problem though, and I didn't really mine on that card either :)
 
I had a Club 3D 280x Royal Queen card explode in my computer (technically the VRMs burst). I was not overclocking and the case had plenty of ventilation. It was being used in Crossfire. Its replacement hasn't had any issues, neither has the other card.

But that's the first video card I've had to RMA (it was less than a week so the retailer immediately replaced the card, the issue was very obvious as the card smelled very strongly of burned PCB) in many years and many, many cards.
 
This is from the same people that brought you, roughly, the same skewed numbers in 2011/2012 and said they had moved away from buying/selling AMD products...
 
Too many factors are involved to explain the card failure rate. Not all factors can be blamed on AMD, and neither the AIB vendors.
 
LordEC911 said:
This is from the same people that brought you, roughly, the same skewed numbers in 2011/2012 and said they had moved away from buying/selling AMD products...
Click to expand...

And this is all old news, its been discussed to death already, why has this topic been re posted, there is no new info.
 
I would have to say it is not important to me. Now, if companies were denying legitimate RMA's like Asus, then I would be bothered. However, this is all just conjecture anyways.
 
Puget's results are in their own test beds, Hardware.fr's results are from a French retailer which is a country that didn't suffer the Litecoin/Dogecoin craze. So coin miners had no effect on these results.
 
crusty_juggler said:
I certainly think so. Look at the failure rates of the most popular mining cards (7950 and 7970) compared to the rates of AMDs other models.
Click to expand...

Yep this, 280x as well.

I my personal thought on the higher rates on 290/x is that these are more enthusiast level cards so people OC'd these guys harder than others. I could be wrong though.
 
Pretty sure that a lot of the shitty board manufacturers contribute to this issue. (HIS and Sapphire having the highest returns of any AMD manufacturer...)

Not to mention that even the venerable XFX started using discount parts on it's boards after switching from AMD to Nvidia.
 
Well, I remember this been discussed before on overclockers uk. Gibbo stated the returns rate of both AMD and Nvidia cards were roughly the same, under 5%

That stopped the arguments. The owner of one of the biggest stores in Europe would know better than any of us.
 
click bait I tell you

from my own personal experience I have had one card of each fail
Here is the list of ATi/AMD cards
ATi Mach 32
ATi Mach 64
Rage 128
Radeon LE
Radeon 8500 (blew a capacitor)
Radeon 9500 Pro
Radeon x1800x
Radeon x1950 Pro
Radeon 3780
Radeon 4780
Radeon 7670
Radeon r280
Radeon 3870

Nividia cards

8400GS-M (Nvidia forgot how to make a video card) in a Dell Vostro 1400. The replacement has been going strong for nearly 8 years. This was the one and only Nvidia card I owned.

Other cards

Trio 64
Virge
Savage 3D
Savage 4
Savage 2000

Rendition 2100

none of which failed.
 
My GTX 780 caught on fire (capacitor) after 1 year. The replacement RMA was shooting out sparks in less then 2 months. Nvidia FTW :D

But to be fare I have an 8800gt that still works.

No failures with AMD cards that weren't caused by me. Mining 24/7 with 7950's for 4 months after buying them used and gaming with them for about a year and they were still going strong when I sold them on e-bay for a quick change up to a GTX 780 and two 7850's. Man that was the best flip ever :D (expect for the fact that the 780 has died twice so far)
 
TaintedSquirrel said:
Puget's results are in their own test beds, Hardware.fr's results are from a French retailer which is a country that didn't suffer the Litecoin/Dogecoin craze. So coin miners had no effect on these results.
Click to expand...

and why would France not suffer the crypto-craze like every other country in the world?
 
stealthy123 said:
Glad you called him out on that. It was a ridiculous statement.
Click to expand...
It's a true statement.
The cost of power in EU outweighs the profits from coins. There was no price inflation on AMD cards in the EU markets since there was no mining demand.

I'm sure some people did it but certainly not enough to make an impact on RMA figures like these.
 
TaintedSquirrel said:
It's a true statement.
The cost of power in EU outweighs the profits from coins. There was no price inflation on AMD cards in the EU markets since there was no mining demand.

I'm sure some people did it but certainly not enough to make an impact on RMA figures like these.
Click to expand...

at the height of the mining craze, you could mine back the value of a card in 2 weeks, pretty sure they can offset the price of power too.

Here are some #s, http://shrinkthatfootprint.com/average-electricity-prices-kwh

France doesn't seem that high, I'm paying currently 16c/KW (Average), so France @ 19 is not too bad, definitely not "not worth mining".

as for pre-built systems, that doesn't mean they didn't get mined on :) I mined on pretty much everything I could find.'

They don't really mention much about number of cards sold, just an overall (That's from Puget)

We on average sell just under 2000 video cards per year so we have a pretty large sample size with which to base our results, but this data may not be 100% accurate to the industry as a whole. In addition, one factor that you must take into consideration is that we tend to sell more mid and high end cards than the average PC company which likely makes our failure rates a bit higher than normal. Also, we put all of our systems through a very rigorous testing process which is specifically designed to make any card that is close to failing actually fail while it is still at our facility. Because of these two factors, our failure rate is likely higher than the industry average.
Click to expand...
 
Digital Viper-X- said:
at the height of the mining craze, you could mine back the value of a card in 2 weeks, pretty sure they can offset the price of power too.

Here are some #s, http://shrinkthatfootprint.com/average-electricity-prices-kwh

France doesn't seem that high, I'm paying currently 16c/KW (Average), so France @ 19 is not too bad, definitely not "not worth mining".
Click to expand...
A few weeks in December won't off-set 12 months of regular RMA's. And that's certainly not enough for people to assemble GPU farms like we saw in the states (100+ GPU's).

I just wonder what it will take for people to acknowledge that either AMD Is producing low-quality GPU's with poor reliability, or their partners are producing shoddy cards? I realize we're talking about the difference between about 3% and 5% on average, which isn't really enough to cause community outrage. We'll probably revisit this topic next year, citing 2014's figures, and I'm curious to see the new batches of excuses.

Just consider how much it sucks to get a GPU failure (up to 12 months after purchase) and now consider your odds are doubled or more with an AMD card. To me, that's unacceptable. Especially as someone who suffered two faulty ASUS 280X's in a row in 2013.

Hopefully we'll see better numbers moving into Fiji / 14nm next year.

The author of the Puget article responded to my Reddit post, you can find his posts on his profile here:
http://www.reddit.com/user/PS-Matt
 
I'd say a large portion of the failures are user related. Those AMD 7000 cards going up in flames? 99% sure those were due to high clock voltages.
 
TaintedSquirrel said:
A few weeks in December won't off-set 12 months of regular RMA's. And that's certainly not enough for people to assemble GPU farms like we saw in the states (100+ GPU's).

I just wonder what it will take for people to acknowledge that either AMD Is producing low-quality GPU's with poor reliability, or their partners are producing shoddy cards? I realize we're talking about the difference between about 3% and 5% on average, which isn't really enough to cause community outrage. We'll probably revisit this topic next year, citing 2014's figures, and I'm curious to see the new batches of excuses.

Just consider how much it sucks to get a GPU failure (up to 12 months after purchase) and now consider your odds are doubled or more with an AMD card. To me, that's unacceptable. Especially as someone who suffered two faulty ASUS 280X's in a row in 2013.

Hopefully we'll see better numbers moving into Fiji / 14nm next year.

The author of the Puget article responded to my Reddit post, you can find his posts on his profile here:
http://www.reddit.com/user/PS-Matt
Click to expand...

I didn't build a farm, but I still went through 9 gpus in 4 months, which is 8 more than i would normally go through in 2 years, if that doesn't skew numbers, I don't know what would :p

and it's not just 2 weeks, people see the prices, and buy in, stores were sold out of AMD cards(280x/7970/7950 and then 290/290x) for nearly 4 months! even at the jacked up prices, people were still buying them
 
TaintedSquirrel said:
A few weeks in December won't off-set 12 months of regular RMA's. And that's certainly not enough for people to assemble GPU farms like we saw in the states (100+ GPU's).

I just wonder what it will take for people to acknowledge that either AMD Is producing low-quality GPU's with poor reliability, or their partners are producing shoddy cards? I realize we're talking about the difference between about 3% and 5% on average, which isn't really enough to cause community outrage. We'll probably revisit this topic next year, citing 2014's figures, and I'm curious to see the new batches of excuses.

Just consider how much it sucks to get a GPU failure (up to 12 months after purchase) and now consider your odds are doubled or more with an AMD card. To me, that's unacceptable. Especially as someone who suffered two faulty ASUS 280X's in a row in 2013.

Hopefully we'll see better numbers moving into Fiji / 14nm next year.

The author of the Puget article responded to my Reddit post, you can find his posts on his profile here:
http://www.reddit.com/user/PS-Matt
Click to expand...

Except they don't describe or explain their "rigorous testing," they don't investigate why or what failed on the cards. There are far too many uncontrolled variables in their "study."

The fact of the matter is AMD's build quality on reference cards is just as good if not better than Nvidia's. AIBs always tend to cut corners on both sides.
 
i had received a new buggy 280x from xfx that i just got around to rma the replacement works great... overall it was a good experience i am looking at a amd card for the future... i had a geforce 1 gts 250 a geforce 2 mx4 a geforce 4200 ti and a geforce 6800 xt a gtx 250

i had on ati/amd
rage 2
hd 3870
r7 260x
r9 270x msi
r9 270x sapphire
r9 280x xfx

mobile
ati
hd 6310

nvidia
7800
 
Maybe it's just me but I don't see this as a big deal. Reverse the numbers and you've got a 94.5% chance of your AMD card not failing vs a 97.5% chance of your Nvidia card being good. Throw in the mining craze and I don't really see anything out of whack here. That's a better rate than hard drives, monitors or motherboards I'd wager.
 
It could be argued that, all things being equal, a company that caters for a lower budget market will see greater failure rates due to poorer quality of installations (cheap/weak PSU, poorly ventilated case, etc).

In my experience, however, Nvidia and AMD have been equal. Infact I've had none die on me from either company, but I've certainly KILLED many! :D

Let me see now, starting from first card to ever fail me, and omitting cards that lived...

Geforce2 GTS > Superglued heatsinks onto ram, and then pulled off ram chip when trying to remove.
9700Pro > Over-volted to death.
9800Pro > Superglued heatsinks onto mosfets and then pulled them off when trying to remove.
7800GS > Over-volted to death.
7800GS #2 > Over-volted to death.
8800GT > Over-volted to death.
780ti > Over-volted to death.
980 > Will be volt modding this weekend, so we'll see how long it lasts... :p
 
killed a few in the past
ati x800xl bad vram ran at stock heavy artifacts until no boot
msi geforce3 ti200 sudden death at stock while idling
gigabyte and bfg 2x 8800gt bad lead free solder
gigabyte 3850 knocked sm resistor loose for pci express bus would only work in 8x slot
 
spine said:
It could be argued that, all things being equal, a company that caters for a lower budget market will see greater failure rates due to poorer quality of installations (cheap/weak PSU, poorly ventilated case, etc).

In my experience, however, Nvidia and AMD have been equal. Infact I've had none die on me from either company, but I've certainly KILLED many! :D

Let me see now, starting from first card to ever fail me, and omitting cards that lived...

Geforce2 GTS > Superglued heatsinks onto ram, and then pulled off ram chip when trying to remove.
9700Pro > Over-volted to death.
9800Pro > Superglued heatsinks onto mosfets and then pulled them off when trying to remove.
7800GS > Over-volted to death.
7800GS #2 > Over-volted to death.
8800GT > Over-volted to death.
780ti > Over-volted to death.
980 > Will be volt modding this weekend, so we'll see how long it lasts... :p
Click to expand...

Jeez you're dangerous. Don't come near my cards :D
 
The % for the 7950 doesn't surprise me. Had a stock / non-overclocked one die on me. No mining with it either.
 
LordEC911 said:
Except they don't describe or explain their "rigorous testing," they don't investigate why or what failed on the cards. There are far too many uncontrolled variables in their "study."

The fact of the matter is AMD's build quality on reference cards is just as good if not better than Nvidia's. AIBs always tend to cut corners on both sides.
Click to expand...

I've always been very impressed with AMD's reference designs. They honestly seem to scream quality on their top tier cards.
 
TaintedSquirrel said:
In most cases it does seem directly related to the factory OC's.
Sapphire had some huge problems with VRM's on their 280X cards last year.
Click to expand...


Yeah I remember nV had similar issues when they first started letting vendors overclock, they latter started being a bit more strict with vendor oc's and of course the vendors learned too to have been QC.
 
i don't know about AMD but the only gpus that died on me were al NV, i switched to AMD in 2010 and all is ok now, no crashes or problems whatsoever.
 
xarin said:
Even the 290/x?
Click to expand...

Takes a lickin' and keeps on tickin'. Really impressive hardware that matches the speed of the Nvidia offerings across the board. Only the 980ti has a slight advantage over the AMD offerings in their respective price range. I was playing Just Cause 3 a few minutes ago. Frame rates in the 80's for much of my playtime. It did dip into the 55 - 60 range when I blew up a ton of stuff at once then went right back to 80. I bet it wouldn't even do that if I were running a 5820k like I am considering upgrading to. I'm using settings that are above what the game recommends.
 
BababooeyHTJ said:
I've always been very impressed with AMD's reference designs. They honestly seem to scream quality on their top tier cards.
Click to expand...

Same, that's why I always go for them. Plus the most OC/modding collaboration as they're fresh meat on the block at that point.

Only had an 8800GT fail, usual BGA/dry solder failure.

Perhaps AMD failures in more recent years (post 79X0), are more related to the lower sales volume, thus a failure is over represented?
 
You must log in or register to reply here.
Back
Top