[Misleading] Data from large French retailer show AMD GPUs less reliable than Nvidia

[Cloudfire777]

Hardware.fr published these data about a month ago. The data is from 2012-2013.

If there is an unknown at the board and purchasing equipment, it is the reliability thereof. Certainly, the reputation of a manufacturer is to reassure us, but one model to another reliability can vary greatly, even the well-known manufacturers are not immune to a blunder.

Although, as we like to report in financial circles, past performance is no guarantee of future performance, we publish today a part of the statistics of return rates for failure to us. This type of statistical needs a certain distance, so that much of the products have become obsolete. There remains that these statistics are informative and allow finger pointing see products manufacturers which we hope they will improve things in the future.

The first question we must answer is of course the origin of these statistics. They come from a large French e-retailer, in which we could have direct access to databases. So we were able to directly extract the statistics we needed.

- 12.67% Sapphire Radeon HD 7850
- 7.44% Sapphire Radeon HD 7870 OC V2
- 7.41% Sapphire Radeon HD 7870 OC V1
- 7.02% Sapphire HD 7950 With Boost (11196-16)
- 6.09% ASUS HD7750-DCSL-1GD5
- 5.82% Sapphire Radeon HD 7870 V1
- 5.65% Sapphire Radeon HD 7870 V2
- Gainward GeForce GTX 5.30% 670

If we look at the figures GPU we get:

- Radeon HD 7850: 3.74%
- Radeon HD 7870: 5.48%
- Radeon HD 7870 XT: 4.25%
- Radeon HD 7950: 5.75%
- Radeon HD 7970: 5.31%

- GeForce GTX 660: 1.01%
- GeForce GTX 660 Ti: 2.81%
- GeForce GTX 670: 2.87%
- GeForce GTX 680: 1.99%

*google translated* It is easy to conclude that the cards based on AMD GPUs are less reliable than maps based on NVIDIA GPU

Source: Hardware.fr
Hardware.fr translated

 

[broken_pixel]

It is still up to the silicon gods to decide after your purchase, the fait of the GPU.

Hahaha cloudfart you must be PR for Nvidia, lulz!

Warning issued for thread crapping and personal attack.
-- stahlhart

 

[BallaTheFeared]

The Radeon 7850 is an overclocking beast.

It also seems to not enjoy being overclocked so much, and it has low resistance to soup.

No real shock value here, despite my objections kepler voltage locking made sense on every level to Nvidia except selling me one of their cards.

 

[chimaxi83]

Oh, it's this thread again.

 

[Cloudfire777]

Look at GTX 680 and 7970 which was their first flagships this generation.
7970 have over twice as much failures as GTX 680.

Sapphire cards seems to be not very reliable. Whats up with that?

 

[Abwx]

Hardware.fr published these data about a month ago. The data is from 2012-2013.

*google translated* It is easy to conclude that the cards based on AMD GPUs are less reliable than maps based on NVIDIA GPU

Source: Hardware.fr
Hardware.fr translated

Let s look at the complete text that you pretend is google
translated ;

It is easy to conclude that the cards based on AMD GPUs are less reliable than maps based on NVIDIA GPU
Without denying thoses numbers we must add that Sapphire
weight a lot on thoses 7850 and 7870 numbers since without
Sapphire we get 1.5% and 1.64% for thoses GPUs.

I m fluent in french , of course, but anyway how convenient
to translate only half of a tiny paragraph such that the whole
meaning is completely stretched and changed.

 

[Imouto]

As chimaxi83 said it's been posted already and yet again debunked by the bloody high return rate of Sapphire cards, specially the 7870 with VRAM problems.

So no, AMD GPUs are as reliable as NV. The problem is with assemblers like Sapphire throwing all its brand prestige out the board. If Sapphire wasn't so popular the failure rates would be the same.

 

[Abwx]

Sapphire cards seems to be not very reliable. Whats up with that?

What is up is that you removed the part of the quote that
say that failure rates are deeply influenced by Sapphire ,
implying that it was the AMD GPUs that were less reliable ,
read my post above...

 

[blackened23]

The Radeon 7850 is an overclocking beast.

It also seems to not enjoy being overclocked so much, and it has low resistance to soup.

No real shock value here, despite my objections kepler voltage locking made sense on every level to Nvidia except selling me one of their cards.

Voltage was locked on the 680 IIRC, but slight voltage adjustments are allowed on the 700 series, depending on the versatility of the cooler and what nvidia deems acceptable with their green light program (i'm assuming). Most aftermarket 780 cards have at least +37mV, while the lightning has +100mV, and the classified can do more than that, but it isn't officially sanctioned by nvidia. The EVGA classified controller isn't technically allowed, but it's EVGA's dirty little secret I guess. (it can do up to 1.4V IIRC)

Or you can use a custom BIOS and use whatever voltage you want up to 1.35V (I think? That's what the skynet bios allows..) So it's not full voltage control but most 780 cards overclock extremely well despite this. It isn't uncommon for many 780s to hit ~1100 boosts out of box when the advertised boost is 941mhz.

In any case, i'm curious as to what constitutes a failure per this article. Does the mere act of RMA'ing a card count as a "Failure"? I know a lot of guys like to return cards just "because", not necessarily that they failed...Personally I haven't had any failure issues with AMD cards, who knows though. 3% sounds excessively high, i'm thinking the mere RMA of a card is counted in those stats. I could be wrong though.

 

[Abwx]

So no, AMD GPUs are as reliable as NV. The problem is with assemblers like Sapphire throwing all its brand prestige out the board. If Sapphire wasn't so popular the failure rates would be the same.

On another note AMD should be cautious because they
gave Sapphire the exclusivity for their professional series
Fire Pro GFX and that such failure rate would be a disaster
in such high priced products , hope that some AMD guy
read by there and do his work since we have early infos
that Sapphire are not up to some other manufacturers..

 

[runzwithsizorz]

Have used AMD cards for years and never had one fail and I never bought Sapphire brand. MSI or Gigabite. Currently using 5870 (laptop four yrs) and 7870 Hawk (fairly new, 1 year) and 4250 (laptop 8 years)

 

[Cloudfire777]

Let s look at the complete text that you pretend is google
translated ;

I m fluent in french , of course, but anyway how convenient
to translate only half of a tiny paragraph such that the whole
meaning is completely stretched and changed.

The Sapphire cards still have AMD silicon.
We don`t know if its the other components or the silicon that failed.

And since you bash me about not quoting the whole text, did you notice that no Sapphire 7970 cards was listed? And they say:

In contrast to the 7950 and 7970 rates are higher if the Sapphire cards are excluded

Too bad they didn`t have the sample required to post data

 

[Abwx]

i'm thinking the mere RMA of a card is counted in those stats. I could be wrong though.

In the begining of the article they stipulate that
thoses rate include any card that is returned even
if it is later found that the card has no issue.

 

[Abwx]

The Sapphire cards still have AMD silicon.
We don`t know if its the other components or the silicon that failed.

And since you bash me about not quoting the whole text, did you notice that no Sapphire 7970 cards was listed? And they say:

Too bad they didn`t have the sample required to post data

There s all the necessary datas , including previous articles
that show clearly that Sapphire is the one increasing thoses
return rates so contrary to your assumptions this has nothing
to do with the GPU brand , what could be less reliable is obviously Sapphire s products but not AMD s GPUs , that s
why i think that the title of this thread is more than
misleading and just totaly irrelevantb since nowhere
HFR did say that AMD GPUs could be less reliable.

If it was the silicon then all cards brands would had
exactly the same failure rate as Sapphire , isnt it..??.

 

[Cloudfire777]

Look at Newegg ratings:

XFX 290X
24% giving it 1 out of 5 eggs.

Sapphire 290X
19% giving it 1 out of 5 eggs


EVGA GTX 780
2% rating it 1 out of 5 eggs

EVGA GTX 780 Ti
0% rating it 1 out of 5 eggs


The 680/7970 with most ratings

XFX 7970 GHz
33% rating it 1/2 out of 5 eggs

EVGA GTX 680
12% rating it 1/2 out of 5 eggs

Asus GTX 680
13% rating it 1/2 out of 5 eggs


I`m beginning to see a pattern here..

 

[Imouto]

Dear TFSM... Now that your OP and sensationalism are exposed you're moving goalposts to the HD 7970 only. If you remove the Sapphires from the HD 7850 and HD 7870 samples the failure rate is being beaten only by the GTX 660 and both are better than the GTX 660Ti, GTX 670 and GTX 680.

So it depends on assemblers and some GPUs, not brands.

EDIT: Not happy with that you had to go and make a case of the XFX 290X with 17 samples. Bravo, you're such an statistician.

 

[horza]

The Sapphire cards still have AMD silicon.
We don`t know if its the other components or the silicon that failed.

And since you bash me about not quoting the whole text, did you notice that no Sapphire 7970 cards was listed? And they say:

Too bad they didn`t have the sample required to post data

He didn't "bash" you it's a very fair question to ask when you chop a quote as you did and removed a very important aspect of the conclusion. Never-mind lets act like we are getting attacked unfairly and change the focus fast! This thread sucks.

 

[Cloudfire777]

Dear TFSM... Now that your OP and sensationalism are exposed you're moving goalposts to the HD 7970 only. If you remove the Sapphires from the HD 7850 and HD 7870 samples the failure rate is being beaten only by the GTX 660 and both are better than the GTX 660Ti, GTX 670 and GTX 680.

So it depends on assemblers and some GPUs, not brands.

EDIT: Not happy with that you had to go and make a case of the XFX 290X with 17 samples. Bravo, you're such an statistician.

It was actually the one with the biggest sample of reviewers I could find.

Here is
HIS 290X
29% giving it 1/5 eggs

Asus 290X
0% giving it 1/5 eggs

Combine all of them together, the dissatisfaction about 290X is pretty high compared to the GTX 780 and 780 Ti`s.
So it doesn`t look like Sapphire cards are the only one with high failure rate.

I`m not cherry picking here. I took the GPUs with the most reviews.

 

[Abwx]

It was actually the one with the biggest sample of reviewers I could find.

Here is
HIS 290X
29% giving it 1/5 eggs

Asus 290X
0% giving it 1/5 eggs

Combine all of them together, the dissatisfaction about 290X is pretty high compared to the GTX 780 and 780 Ti`s.
So it doesn`t look like Sapphire cards are the only one with high failure rate.

I`m not cherry picking here. I took the GPUs with the most reviews.

You will notice that there s return rates but in no case
we can conclude that it was due to GPU failure since
we have not the numbers for the exact kind of failure,
as indeed aknowledged by HFR , so their numbers , as well
as newegg notations, are for cards and nowhere we have
info about the GPUs failure rates but still you used a survey
about graphic cards to say that all thoses failures are due
to GPUs even no hard data has been available about this,
so the title of your thread is voluntarly misleading since
hardware/fr aknowledge that they dont know what were
the failures causes.

 

[Imouto]

WTF is wrong with you? If I go out my house right now and count the people walking on the street can I assume that I'm alone in my country with that guy getting the trash out? That there are no women, only freezing cats. Can I extend the study to the whole world?

 

[DaveSimmons]

Thread title is not supported by thread contents. Also: repost.

we must add that Sapphire
weight a lot on thoses 7850 and 7870 numbers since without
Sapphire we get 1.5% and 1.64% for thoses GPUs.

- GeForce GTX 660 Ti: 2.81%
- GeForce GTX 670: 2.87%
- GeForce GTX 680: 1.99%

Hmm, I see a lower % for AMD once you take out Sapphire's boards. Does that mean nvidia is less reliable?

Luckily my GTX 680 is still working perfectly.

 

[Cloudfire777]

Let me quote again

In contrast to the 7950 and 7970 rates are higher if the Sapphire cards are excluded

GTX 680 had a failure rate of 1.99%.
7850/7870 1.5% and 1.64%. That means the 7970 is higher than 680.
But the sample was too low so I don`t have any numbers. Which makes this discussion useless.

Also, there was a similar thread posted earlier where someone commented on the bad Newegg ratings.
I was active there too I see. We had this discussion before. haha
:$


I didn`t read the whole hardware.fr article and should have noticed that the Sapphire increased the bad rates for the AMD GPUs. I really didnt see it.
My bad. Sorry. I have reported this thread to be locked.

 

[rgallant]

It was actually the one with the biggest sample of reviewers I could find.

Here is
HIS 290X
29% giving it 1/5 eggs

Asus 290X
0% giving it 1/5 eggs

Combine all of them together, the dissatisfaction about 290X is pretty high compared to the GTX 780 and 780 Ti`s.
So it doesn`t look like Sapphire cards are the only one with high failure rate.


I`m not cherry picking here. I took the GPUs with the most reviews.

so they only sold 17 cards all this time ?

 

[3DVagabond]

The Sapphire 7870 had issues for a while because they used caps (IIRC it was caps) that were not on AMD's approved parts list. They fixed that and allowed anyone who had cards with the improper caps in them to RMA for new cards.

 

[blastingcap]

The Sapphire 7870 had issues for a while because they used caps (IIRC it was caps) that were not on AMD's approved parts list. They fixed that and allowed anyone who had cards with the improper caps in them to RMA for new cards.

This.

Furthermore, I will never buy a locked-voltage card if I can help it. There are many others out there who feel the same way. Are you listening NV?