Quality & Longevity - AMD 7970 vs. Nvidia 680

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

BoFox

Senior member
May 10, 2008
689
0
0
For viewing convenience, I'll just quote it from another thread so that you can see the details here:
the stats come from. They’re taken from a large French e-tailer, whose database we have had direct access to. We were, then, able to extract the stats we wanted directly. Of course they’re limited to the products sold by the retailer, but unfortunately there isn’t any other way of getting hold of reliable stats. Who would believe for example any returns rates given by the manufacturers themselves?


The returns rates given concern the products sold between April 1st 2010 and October 1st 2010 for returns made before April 2011, namely after between 6 months and a year of use. The statistics by manufacturer are based on a minimum sample of 500 sales, those by model on a minimum sample of 100 sales. Each time, we’ve compared the rates by manufacturer to those in our previous article on the subject published in December 2010.

- PNY 1.2% (against 1.1%)
- ASUS 1.3% (against 0.9%)
- Zotac 1.4%
- Sapphire 1.5% (against 1.8%)
- Club 3D 1.6% (against 2.2%)
- Gainward 1.6% (against 1.5%)
- Gigabyte 2.5% (against 1.7%)
- MSI 2.9% (against 1.7%)
- XFX 3.0% (against 2.4%)

ASUS and PNY swap places here as a result of the higher ASUS returns rate. Zotac has entered the classification in a respectable position, while Sapphire and Club 3D have both improved their scores. The rates for Gigabyte on the other hand, and above all MSI, are up a good deal and XFX is still in last place with an even worse average than before. Overall the numbers are acceptable, but three models have returns rates of at least 5%:

- 7.0%: Gigabyte GV-R587UD-1GD
- 7.0%: XFX ATI Radeon HD 5870
- 5.0%: ASUS EAH5870/2DIS/1GD5/V2

The three models are all based on the Radeon HD 5870! Here are the stats by GPU:

- Radeon HD 5770: 2.0%
- Radeon HD 5830: 2.5%
- Radeon HD 5850: 5.5%
- Radeon HD 5870: 5.0%
- Radeon HD 5970: 10.9%
- GeForce GTS 250: 1.6%
- GeForce GTX 460: 2.4%
- GeForce GTX 465: 3.4%
- GeForce GTX 470: 4.7%
- GeForce GTX 480: 3.0%

This confirms the Radeon HD 5870’s poor showing. The Radeon HD 5850 also does badly. Six months ago, the two solutions were at 3.2-3.4%! The Radeon HD 5970 is a very fragile card as is often the case with bi-GPU cards. None of the NVIDIA solutions are over 5% and the GeForce GTX 480 has a rate of just 3%, which is a very good score for a high end card.

I have no idea if the pattern holds to be true for newer cards, though. IMHO, the Geforce 5xx series seemed to be a bit better except for GTX 570 and the super-rare GTX 590 issue (only maybe 2 or 4 of GTX 590's blowing up were ever documented, out of maybe 50,000 cards). From reading the forums, it seems that HD 6990 also had some problems, perhaps a bit more commonly. The patterns are almost never a given.

It's still way too early to tell, though. Newegg reviews are a quick way to get a "feel" of it.
 
Last edited:

chimaxi83

Diamond Member
May 18, 2003
5,457
63
101
After reporting the rude, inflammatory, personally-directed remark saying: "nice try tho rolleyes", I would like to see your sources before I can agree with you.

Perhaps BallatheFeared missed the sources that you have in mind.


lol nice stealth edit bud :thumbup:

you should probably go back to your friends at ABT, the mob mentality there makes you feel right at home doesn't it?

Let's not tell people where to go
-ViRGE
 
Last edited by a moderator:

BoFox

Senior member
May 10, 2008
689
0
0
lol nice stealth edit bud :thumbup:

you should probably go back to your friends at ABT, the mob mentality there makes you feel right at home doesn't it?
I do remember one source that BallatheFeared was probably thinking of, so I think he was posting with honest conviction. Some sources might be saying different things, though.

BTW, the gun seems to be pointed the wrong way:

My post wasn't meant to be a tease, if I was "free" to name them then I would have done so at the outset. I am not free to name them, even if they gave me their approval of me releasing confidential info I would not do it because I no more want to see them publicly crucified and vilified than I wish to see Keysplayr vilified.

(its the same with employees of companies that intentionally seek to remain unaffiliated, I know some of them privately and I would not "out" them even if they gave me their OK to do so...but yes there are AMD employees in our midst here, and they hide for good reason because there are some truly ugly hateful members in our community, on both sides of the fence)

If you truly believe that AMD does not give anyone free gear then I don't know what to tell you.

Consider for the moment the fact that the ONLY reason you are aware of Keysplayr being a member of the NV Focus Group is because it is a requirement of the NV Focus Group that he "out" himself. It is not a requirement of the AnandTech Forums (ATF).

As such we (ATF) also do not have a policy regarding members of parallel organizations on the AMD side of the fence to "out" themselves, that is between them and the policies of their organization just as is the case with the NV focus group.
http://forums.anandtech.com/showpost.php?p=32301070&postcount=150
 
Last edited:

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
I think it's pretty obvious which card is more "heavily" built. That alone though doesn't guarantee longevity.

As far as card failures by %, that alone doesn't paint much of a picture. What failed and why? Was it VRM's when extra voltage was applied? Was it fans on a particular model? Was it a bad batch of capacitors or memory chips on a budget board? Was it video corruption because of bad drivers that made people think their cards were faulty? There are just so many variables. Polls and studies are only as reliable as the people who took them and the conditions they were taken under.

Having 2x 6pin connectors in no real world situation caps the power usage. Just because the spec is 75W each (plus 75W for the pcie connecteor) they aren't limited to that amount. 2x 8pin, or 6+8pin, by itself, won't make the card use more power. That, along with other changes they typically do to "premium cards" could afford more stability to the power delivery, though.
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
Yeah, didnt that come into effect with the 6xxx series, or was it 5xxx?

I think 6xxx since it was late and slow while offering tess performance on par with the 560ti.

Take it to PM. This has nothing to do with the subject at hand
-ViRGE
 
Last edited by a moderator:

Stayfr0sty

Senior member
Mar 5, 2012
465
0
0
This really depends more on the 3rd party manufacturer then whether its AMD or Nvidia.
They just make the GPU chips and send them out to the manfacturers, they are the ones who decide what PCB to put, capacitors, memory,etc.....
 

Elfear

Diamond Member
May 30, 2004
7,163
819
126
V

Watercooling might inadvertently cause vrms to heat up as well since there is no air blowing over the vrms. The tower type cpu coolers to a lesser extent have the same problem unless the fan is situated low enough to cool the vrm heatsinks on the mb.

Except for full-cover blocks. No need for airflow then.
 

ViRGE

Elite Member, Moderator Emeritus
Oct 9, 1999
31,516
167
106
I think it's pretty obvious which card is more "heavily" built. That alone though doesn't guarantee longevity.

As far as card failures by %, that alone doesn't paint much of a picture. What failed and why? Was it VRM's when extra voltage was applied? Was it fans on a particular model? Was it a bad batch of capacitors or memory chips on a budget board? Was it video corruption because of bad drivers that made people think their cards were faulty? There are just so many variables. Polls and studies are only as reliable as the people who took them and the conditions they were taken under.
The big issue, IMHO, is how honest the RMAs are. Overclockers have a poor reputation (and not unearned) for returning cards because they don't clock high enough or they break them. Which is why I'm not surprised the 5970 was at 10% in that chart.
 

Stayfr0sty

Senior member
Mar 5, 2012
465
0
0
The big issue, IMHO, is how honest the RMAs are. Overclockers have a poor reputation (and not unearned) for returning cards because they don't clock high enough or they break them. Which is why I'm not surprised the 5970 was at 10% in that chart.

Amen
5970 scored low probably cause the benchie wasnt using the 2nd gpu
So it was performing more like a 5870. After all a 5970 is basically two 5870s
 

BoFox

Senior member
May 10, 2008
689
0
0
The big issue, IMHO, is how honest the RMAs are. Overclockers have a poor reputation (and not unearned) for returning cards because they don't clock high enough or they break them. Which is why I'm not surprised the 5970 was at 10% in that chart.

That's a good point - to expand upon it, owners of GTX 465 and GTX 470 cards were also more likely to overclock their cards than GTX 480 owners (of which a higher percentage were just rich people who were not overclockers).
 

Jovec

Senior member
Feb 24, 2008
579
2
81
The big issue, IMHO, is how honest the RMAs are. Overclockers have a poor reputation (and not unearned) for returning cards because they don't clock high enough or they break them. Which is why I'm not surprised the 5970 was at 10% in that chart.

As consumers we in no way have access to any accurate stats on this issue. Actual failures? Poor reference design? Poor QC at manufacturer? Poor component choice? Poor engineering of a non-reference? End-user abuse? Shipping damage? End-user buyer's remorse? End-user's realistic/unrealistic expectations not met? End-user's lack of understanding (half-height OEM case, no PCIe power connectors on their 300w OEM PS, etc).
 
Last edited:

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
That's a good point - to expand upon it, owners of GTX 465 and GTX 470 cards were also more likely to overclock their cards than GTX 480 owners (of which a higher percentage were just rich people who were not overclockers).


I'd like to see the scientific conditions you took this study under. :D
 

Lonbjerg

Diamond Member
Dec 6, 2009
4,419
0
0
You should look at GK110, not GK14 for comparison.

Eg. GPGPU test shows that GK104 is not NVIDIA's "kepler" flagship...the missing DP and other GPGPU tidbits tells a interesting tale...of a midrage GPU that became (by name) a high end GPU...
 
Mar 10, 2006
11,715
2,012
126
That's a good point - to expand upon it, owners of GTX 465 and GTX 470 cards were also more likely to overclock their cards than GTX 480 owners (of which a higher percentage were just rich people who were not overclockers).

lol, you think that being able to buy a $500 GPU once per year makes you rich?
 

BoFox

Senior member
May 10, 2008
689
0
0
I'd like to see the scientific conditions you took this study under. :D

Hehe, probably just as scientific as Virge's claim that lots of people were upset with their 5970's. :p Just kidding, hehe..

lol, you think that being able to buy a $500 GPU once per year makes you rich?

*OOF*.. yep, you gotta be a billionaire to afford a $500 GPU. It's an "lol" way to interpret my wording!
 

vicbdn

Junior Member
Feb 25, 2009
11
0
0
I'm surprised nobody's asked which reference card has the least coil whine...



Both attempts at 7970s from Sapphire had very obvious coil whine. My EVGA GTX 680 is quiet.

I'd bet on the GTX 680 from that alone.
 

jimrawr

Senior member
Mar 4, 2003
888
1
81
I guess I am just unlucky with vid cards, but in the last 4 years I have had 3 RMAs and NONE of the cards were overclocked. First was an XFX 8800GTX, died in about a year and a half. Then I got a EVGA 295GTX, which died after a month! RMA and got another, which died about a year and a half later as well.

Thankfully all under warranty so no big issue, but ive since bought a XFX 7970 and hoping not to have problems with this one. I think part of my problem was my OLD atx case which had really bad airflow and no cable mamagement so the temps were pretty high.