GPU RMA rates

[thilanliyan]

http://www.behardware.com/articles/881-5/components-returns-rates-7.html

"The returns rates given concern the products sold between October 1st 2011 and April 1st 2012 for returns made before October 2012, namely after between 6 months and a year of use. Over the lifetime of a product the returns generally form a spread out U on the graph, with the end virtually flat. Our figures therefore cover the early part of the lifetime of products, where returns rates are high.

The statistics by brand are based on a minimum sample of 500 sales and those by model on a minimum sample of 100 sales, with the biggest volumes reaching tens of thousands of parts by brand and thousands of parts by model. Each time, we’ve compared the rates by manufacturer to those in our previous article on the subject published in May 2012."

I'm guessing the nV 6-series is not there at all because it was later to market so they didn't have enough data?

DISCLAIMER: These results are from ONE French retailer, so may not be indicative of any overall trend.

 

[OVerLoRDI]

Ouch some of those top end Fermi cards have high return rates.

 

[Jimzz]

A lot seem to be high end cards that may have power issues or systems with other specs holding the GPU back so they return it.

Return =/= bad GPU.

 

[96Firebird]

I'm guessing the nV 6-series is not there at all because it was later to market so they didn't have enough data?

Yeah, it was only about a week from release date of the GTX 680 until the end of the "sales" deadline for the study, so I don't think they got enough data in.

 

[thilanliyan]

A lot seem to be high end cards that may have power issues or systems with other specs holding the GPU back so they return it.

Return =/= bad GPU.

True, but it doesn't matter though...if pretty much any part of the card fails, you have a brick with a useless GPU on it. The GPU being fine doesn't mean you can still use the card.

 

[Meghan54]

True, but it doesn't matter though...if pretty much any part of the card fails, you have a brick with a useless GPU on it. The GPU being fine doesn't mean you can still use the card.

Again, the returns to that retailer do NOT equate to failure rates, simply the card was returned for whatever reason the buyer had. European consumer law is vastly different than U.S. consumer law in that the consumer has much more right of return to a Euro retailer than any U.S. company allows.

The only way one could actually see failure RMA rates is to access the manufacturer data, which is never public.

 

[thilanliyan]

Again, the returns to that retailer do NOT equate to failure rates, simply the card was returned for whatever reason the buyer had.

I understand but the retailer does do some testing to see whether they ARE faulty:

"Under what conditions is a part declared as defective by this etailer? There are two possible cases: either the technician considers the exchange of information with the client (type of problem, cross testing) sufficient to declare that the product isn’t working, or there’s a question mark over the component and the etailer tests it to check if it’s working or not.

Among the returns that aren’t tested, some of the components announced as having an issue by customers probably aren't actually defective, in spite of the precautions taken by the technician. This is something inherent in the etailing sector and in practice, it’s unlikely that any model or product is more affected by this phenomenon than any other (at least we’re aware of no objective argument that shows this)."

Of course some items deemed defective may not actually be so, but that would be the minority I'd wager, considering they do perform some sort of testing prior to accepting a return. My original response was more to do with the fact that IF the card no longer works, it doesn't really matter which part of the card failed.

 

[ShintaiDK]

When you look at top5. It got nothing to do with Fermi. But alot to do with Grainwards Phantom cards.

If we exclude the two Gainward Phantom’s from the figures for the GTX 580, its figures drop back to 5.27%.

Then HD7970 cards are the worst suddenly with 5.56%. But thats the point, One manufactor can easily ruin a trend. Specially in highend cards. I would only worry if mainstream and value cards had a high RMA rate. But a custom AIB, certainly not.

Now OCZ under SSD. Thats where the fail is.

 

[magomago]

I wonder if, for higher end cards, people RMAing in hopes of obtaining a higher clocking card plays a role in this at all.

 

[ShintaiDK]

I wonder if, for higher end cards, people RMAing in hopes of obtaining a higher clocking card plays a role in this at all.

That wouldnt be impossible. We seen that happen with CPUs. Some people dont have issues committing fraud.

 

[Plimogz]

Now OCZ under SSD. Thats where the fail is.

Sure, say that to the people who were told to avoid OCZ and spend more on Samsung 840s. Whereas from my experience (only 3 drives so far, but still) the OCZ's are both reliable and well supported by their manufacturer.

Frankly, this attitude has been around since DDR1 days, and has never quite squared with my personal experience: OCZ RAM and now OCZ SSD's have been nothing if not satisfactory (and fast!) throughout. And the one time I've had one of their products fail (Gold VX 4000 modules) their customer support was exemplary. I still can't believe that they warrantied it @3.5V for life. lol.