Study: Modern Graphics Card Failure Rates

[malware]

Here's an interesting story with emphasis on graphics cards failure rates. It was interesting for me to see the results, and I hope it will be interesting for you too. The original story comes from Hardware France and is based upon results from a major un-named French retailer. The data of faulty graphics cards was collected in the period of seven months from August 2008 to March of this year. To build the statistics, an ideal sample amount of at least 500 units was used. The statistics provide reliability information on a 'per brand' and 'per GPU model' basis. You're welcome to comment on the results. Read the English version at GPU Cafe here, or the original French version here at Hardware.fr.



View at TechPowerUp Main Site

 

[FordGT90Concept]

I've never heard of the manufacturer "Point of View." Europe only? I can't find any information on them.

 

[Supreme0verlord]

What, no EVGA? That's very interesting to see how high the failure rate's for the GTX280 are, and its odd to see how big a difference there is between the 280 and the 260...

 

[Error 404]

Yay, go ASUS! I love my Asus Black Pearl 9600 GT, good to see Asus products have the lowest failure rates in GPUs.

 

[HammerON]

Interesting - I would like to see how they arrived at these scores/stats.

 

[tcorbyn]

Cheking how many RMAs are passed through each company I would gues. POV is shiiit! Good to know when decideing on a manufacturer tho. I had two sappire 4870x2s die on me, I gues I was just unlucky!

 

[HammerON]

I am glad to see that from their results that ASUS is not at the top of their list

 

[Hayder_Master]

mmmmm, seems ATI GPU's better durable here

 

[Gzero]

Remember this is from a European site, only cards sold over here will be recorded. We have only just been getting Evga now. Point of View used to offer the cheaper alternative to the bigger companies like Gigabyte, Gainward and Asus.

Whoa check the over 10% failure for GTX280's.

 

[Gzero]

Google translated hardware.fr to english

Has more detail about certain models and makes.

 

[GFC]

I've never heard of the manufacturer "Point of View." Europe only? I can't find any information on them.

When i saw the chart i just went 'gick', my video card is from POV xD.And um.. i think it's just euro retailer, and even here (atleast in this part of europe [north]) it's not even that popular.

 

[DarkMatter]

The high GTX280 failure rate probably is directly related to Point of View failure rate. For some reason, mainly price, POV is a very successful manufacturer in Europe, and specially in France, Portugal and Spain, but it has always been a bad manufacturer. It's typical for me to recommend "If you want a Nvidia card take just any manufacturer except Point of View". It's just cheap.

 

[mtosev]

I've never heard of the manufacturer "Point of View." Europe only? I can't find any information on them.

POV is from the netherland.


http://www.pointofview-online.com/default2.php?content_id=5

 

[theorw]

I think i have the less failing card! SAPPHIRE and 4850!
Both are the most reliable!
ATI rules btw...Most reliable and cheaper cards exept the 3870x2...

 

[hrvoje]

Got to sell my PowerColor 3870 X2 while is still working

 

[DarkMatter]

Got to sell my PowerColor 3870 X2 while is still working

I wouldn't, unless you have other reasons. Just one retailer over a few months is insufficient to show the real picture statistically speaking. Much less when they are lacking some of the best vendors, like EVGA, BFG, HIS and XFX...

In fact, only Saphire is known to be good with it's cards from that list. Good manufacturers for Ati cards have been Saphire and HIS lately and in Nvidia's camp it's been EVGA, Palit and Zotac IMHO. Things just change and the manufacturers that are on top varies with the time and now it's their time IMHO. That doesn't mean others are bad, but certainly aren't the best.

 

[lemonadesoda]

CAREFUL with these statistics.

It seems that the data might not be statistically significant based on the sample size. Two cards sold more than 100 units in the dataset being analysed and had a return rate of 10%+. That's bad for sure.

However, what can we say about the other cards? The number of sales for any one card could be between 20 and 200 (let's say - the data was not given, but taking the example of "100" cards being the average, and they said 500 was the minimum dataset for a brand. But for a specific card, the figure will be a lot lower). Just ONE return (note this is RETURN for complaint of failure... but could be nothing more than unhappy or incompetent user, and not necessarily failure/faulty GPU) is enough to add between 0.5% and 5% to the statistic!!!

The data is unreliable due to the sample size. YES, it is interesting and we would like to know more... but currently, dont make too many decisions/conclusions based on that table/stats.

 

[DarkMatter]

CAREFUL with these statistics.

It seems that the data might not be statistically significant based on the sample size. Two cards sold more than 100 units in the dataset being analysed and had a return rate of 10%+. That's bad for sure.

However, what can we say about the other cards? The number of sales for any one card could be between 20 and 200 (let's say - the data was not given, but taking the example of "100" cards being the average, and they said 500 was the minimum dataset for a brand. But for a specific card, the figure will be a lot lower). Just ONE return (note this is RETURN for complaint of failure... but could be nothing more than unhappy or incompetent user, and not necessarily failure/faulty GPU) is enough to add between 0.5% and 5% to the statistic!!!

The data is unreliable due to the sample size. YES, it is interesting and we would like to know more... but currently, dont make too many decisions/conclusions based on that table/stats.

I agree, I completely agree, but how many times are you going to edit the post until you are satisfied with it?

 

[ZoneDymo]

POV is from the netherland.


http://www.pointofview-online.com/default2.php?content_id=5

Headquaters is in the Netherlands but Point of Few cards are made and by far mainly distributed in Asia.

 

[theorw]

Headquaters is in the Netherlands but Point of Few cards are made and by far mainly distributed in Asia.

So thats why its the most crappy...The millions of chinesse buy them and therefor u have millions of crappy cards...

 

[csendesmark]

I've never heard of the manufacturer "Point of View." Europe only? I can't find any information on them.

Point of View is a value brand (I have a 8800GTS-640, works fine since 2006)
btw

I wanna see the failure rate of Inno3D cards

 

[hrvoje]

I wouldn't, unless you have other reasons. Just one retailer over a few months is insufficient to show the real picture statistically speaking. Much less when they are lacking some of the best vendors, like EVGA, BFG, HIS and XFX...

In fact, only Saphire is known to be good with it's cards from that list. Good manufacturers for Ati cards have been Saphire and HIS lately and in Nvidia's camp it's been EVGA, Palit and Zotac IMHO. Things just change and the manufacturers that are on top varies with the time and now it's their time IMHO. That doesn't mean others are bad, but certainly aren't the best.

Little off topic but...
It doesn't worth selling, because I paid 350E for if and now I can get around 100E (if lucky) for it. Ill rather put it in my another system because its not too bad card for gaming and I never got any problems with it.
Was just kidding about Fig3

Now in topic...
I have had 2 ATi cards from Sapphire and didnt have any problems, exept that i got one with wrong bios flashed and almost fixed it.
Nvidia cards from Gigabyte, Leadtek also no problems.

 

[gumpty]

Cheking how many RMAs are passed through each company I would gues. POV is shiiit! Good to know when decideing on a manufacturer tho. I had two sappire 4870x2s die on me, I gues I was just unlucky!

My second Sapphire 4870x2 is currently being RMA'd also. Between the two of us we can reek havoc on these statistics.

 

[newtekie1]

CAREFUL with these statistics.

It seems that the data might not be statistically significant based on the sample size. Two cards sold more than 100 units in the dataset being analysed and had a return rate of 10%+. That's bad for sure.

However, what can we say about the other cards? The number of sales for any one card could be between 20 and 200 (let's say - the data was not given, but taking the example of "100" cards being the average, and they said 500 was the minimum dataset for a brand. But for a specific card, the figure will be a lot lower). Just ONE return (note this is RETURN for complaint of failure... but could be nothing more than unhappy or incompetent user, and not necessarily failure/faulty GPU) is enough to add between 0.5% and 5% to the statistic!!!

The data is unreliable due to the sample size. YES, it is interesting and we would like to know more... but currently, dont make too many decisions/conclusions based on that table/stats.

Agreed, not to mention this is just from one retailer. A simple bad batch of cards, it happens to everyone, could throw the numbers way off.

 

[DrPepper]

These results don't say how many cards were tested only the percentage. We assume it counts for all but it could only be 50 cards tested and 5 are bad which means 10% is faulty. 10% looks bad because it is bigger than 5 but they are both the same number of cards.