From: First of One on
Interesting failure statistics from a "large French e-retailer" covering
parts sold between September 2008 and March 2009.

Failure rate by brand:
- PNY 1.69%
- ASUSTeK 1.71%
- MSI 2.05%
- Gigabyte 2.63%
- Sapphire 2.7%
- Gainward 3.15%
- Point Of View 4.67%

PNY seems to be doing quite well.

Failure rate by chipset:
- NVIDIA GeForce GTX 260 : 2.4%
- ATI Radeon HD 4850 : 3.3%
- NVIDIA GeForce GTX 295 : 6.5%
- ATI Radeon HD 4870 : 6.1%
- NVIDIA GeForce GTX 280 : 8.3%
- ATI Radeon HD 4870 X2 : 10.5%

The GTX 295 is more reliable than the GTX 280, despite its greater
complexity.

Article is in French, so use Google Translate if required:
http://www.hardware.fr/articles/773-5/taux-pannes-composants.html

--
"War is the continuation of politics by other means.
It can therefore be said that politics is war without
bloodshed while war is politics with bloodshed."


From: KCB on

"First of One" <root(a)127.0.0.1> wrote in message
news:D4mdnaoL_ZTbXUrXnZ2dnUVZ_rKdnZ2d(a)giganews.com...
> Interesting failure statistics from a "large French e-retailer" covering
> parts sold between September 2008 and March 2009.
>
> Failure rate by brand:
> - PNY 1.69%
> - ASUSTeK 1.71%
> - MSI 2.05%
> - Gigabyte 2.63%
> - Sapphire 2.7%
> - Gainward 3.15%
> - Point Of View 4.67%
>
> PNY seems to be doing quite well.
>
> Failure rate by chipset:
> - NVIDIA GeForce GTX 260 : 2.4%
> - ATI Radeon HD 4850 : 3.3%
> - NVIDIA GeForce GTX 295 : 6.5%
> - ATI Radeon HD 4870 : 6.1%
> - NVIDIA GeForce GTX 280 : 8.3%
> - ATI Radeon HD 4870 X2 : 10.5%
>
> The GTX 295 is more reliable than the GTX 280, despite its greater
> complexity.
>
> Article is in French, so use Google Translate if required:
> http://www.hardware.fr/articles/773-5/taux-pannes-composants.html

I think the article is only reporting the "rate of return". That doesn't
necessarily mean failure. The author even states at the end that it is hard
to draw conclusions from the statistics. If the reason for return had been
included, then we would have some useful information.

How many of those 10.5% returns on the 4870 X2 were because of inadequate
power supplies in the buyer's computer, or even difficulty in driver
installation? I also think we need to include the "user stupidity factor",
because some people just shouldn't have a computer.


From: Steve on
In article <hb8ml6$psm$1(a)news.eternal-september.org>,
bcgc_qc(a)hootmail.com says...
<snip>

> I think the article is only reporting the "rate of return". That
> doesn't
> necessarily mean failure. The author even states at the end that it is hard
> to draw conclusions from the statistics. If the reason for return had been
> included, then we would have some useful information.
>
> How many of those 10.5% returns on the 4870 X2 were because of inadequate
> power supplies in the buyer's computer, or even difficulty in driver
> installation? I also think we need to include the "user stupidity factor",
> because some people just shouldn't have a computer.

How many of those returns were because the card wouldn't even fit in
the case?






s

From: First of One on
"KCB" <bcgc_qc(a)hootmail.com> wrote in message
news:hb8ml6$psm$1(a)news.eternal-september.org...
> I think the article is only reporting the "rate of return". That doesn't
> necessarily mean failure. The author even states at the end that it is
> hard to draw conclusions from the statistics. If the reason for return
> had been included, then we would have some useful information.
>
> How many of those 10.5% returns on the 4870 X2 were because of inadequate
> power supplies in the buyer's computer, or even difficulty in driver
> installation? I also think we need to include the "user stupidity
> factor", because some people just shouldn't have a computer.

Alas, the same "user stupidity factors" should apply to the GTX295, yet it
manages a 6.5% return rate. And why did Sapphire's 4870 suffer 10.1% return
rate while the average for all 4870s was only 6.1%? When the sample size is
reasonably large (each brand and model needed minimum sales of 100 units to
be included), the small details do not skew the results as much.

And these are cumulative return rates for cards sold between September 2008
and March 2009, so they include both infant mortality items (DOA, user
stupidity) and defects that may have occurred as much as six months to a
year in service.

Note we are not passing judgement on anybody. The dual-GPU cards can be
expected to be returned more often because they are twice as complex and
demand more from the host system. Gainward cards probably fail more
frequently than PNY cards because PNY doesn't offer overclocked "Golden
Samples".

--
"War is the continuation of politics by other means.
It can therefore be said that politics is war without
bloodshed while war is politics with bloodshed."



From: Benjamin Gawert on
* First of One:
> Interesting failure statistics from a "large French e-retailer" covering
> parts sold between September 2008 and March 2009.

[snip]

> PNY seems to be doing quite well.

I don't think this statistics is really related to hardware reliability.
While PNY indeed does produce good cards (they are also the only
boardpartner for Quadro professional gfx cards), it is probably more
related to what the buyers do. For example, PNY hardly is the brand
choosen by gamers that want to overclock. Those that show the high
return rates on the other sides are brands that aim on the "enthusiast"
market (gamers) with factory-overclocked products. Such products are
probably overclocked even more, and often also modified (i.e. by
replacing the cooler).

The chipset statistics is equally useless. There very likely is no
relaibility difference between a GTX 260 or GTX 280 GPU itself, however
differences in reliability often are due to different operating
conditions (i.e. clock rates, temperaures etc. Therefore it is silly to
compare chipsets without telling under what conditions (i.e. what gfx
card model, what clock rate, what temperatures etc) they ran.

At the end of the day it is just what it is, a return statistics of a
certain french retailer. It however says nothing about the reliability
of a specific gfx card brand or chipset.

Benjamin