Reference vs. non-reference lifespan

[Pwndenburg]

I realize any input I get on this will be anecdotal at best; but, I'd really like to get some opinions on any potential benefits of aftermarket cooling for cards run at stock speeds. Assuming both cards are run at factory settings, would a non-reference cooled card be likely to outlast a reference one due to lower temps?

 

[railven]

My experience, , I've always bought reference since I usually like to buy day 1.

I've never had a card die on me without some kind of outside influence (well one, and it was watered accidentally.)

I even have a reference 9800 GTX+ and have owned two 8800 GTS without ever having to bake either (knock on wood since I still use the 9800 GTX+).

If you plan on doing some OCing, maybe you should look more at non-reference, since I normally stay within the reference spec, might be why my cards have lasted me so long (except a 9500 Pro I softmod to 9700 pro, still, worked fine for years.)

 

[SlowSpyder]

For whatever it is worth to you, I had two Sapphire Vapor-X cards. They used non-reference cooling and a non-reference board. I have nothing but good things to say about those cards, had they been equiped with 2GB of memory I wouldn't have gotten rid of them. I gamed on them and mined quite a bit. Never a problem... so that's my experience to report.

But then again I really have never had a problem (well, a BFG 6800nu OC, but I am not sure if that was a reference board with a difference cooler or a non-reference board). So all in all, I've had good luck with both. I would not be afraid to buy non-reference card from a good manufacturer. Just be aware that some non-reference cards do not allow voltage control that many reference cards do allow.

 

[bdunosk]

Theoretically, cooler = longer lasting. Realistically, most GPU's will outlive their usefulness if you're gaming.

It also depends on the card and what you want out of it. I like aftermarket cooling that's quiet, which hasn't always been achieved with reference cards.

 

[Pwndenburg]

You've got that noise thing right, I had an EVGA 260 core 216 with a reference cooler. That thing has made me wary of reference coolers since. The thing would whine louder than an AMD Phenom II stock fan and died an early death to what I can only presume was excessive heat. Thanks to these forums and ones like that like them; however, I feel like I do a better job of having semi-proper airflow within my case.

To cut to the chase, the question, specifically, is assuming the card being run at stock for its lifespan would it be worth it to wait for non-reference cooling for the 680 gtx specifically. Another variable I've mulled over is the possibility that the auto overclock might be better with more robust cooling without me doing a thing.

Decisions... decisions

 

[railven]

You've got that noise thing right, I had an EVGA 260 core 216 with a reference cooler. That thing has made me wary of reference coolers since. The thing would whine louder than an AMD Phenom II stock fan and died an early death to what I can only presume was excessive heat. Thanks to these forums and ones like that like them; however, I feel like I do a better job of having semi-proper airflow within my case.

To cut to the chase, the question, specifically, is assuming the card being run at stock for its lifespan would it be worth it to wait for non-reference cooling for the 680 gtx specifically. Another variable I've mulled over is the possibility that the auto overclock might be better with more robust cooling without me doing a thing.

Decisions... decisions

The GTX 680 "stock" is relatively quiet. I'm actually impressed with it's acoustics. Now, of course if you start giving it more juice, you'll create more heat, and of course the fan will have to ramp up. But this is true about any card.

So now you're changing your question from "running stock" to "what if I want to OC." You can OC the GTX 680 and keep it relatively quiet from what I've seen/read. But of course at a certain point it will get loud (but that's after some high clocks like 1.3ghz+).

Stock, my HD 7970 and GTX 680 are very tolerable. With a mild OC my HD 7970 gets a little noiser but acceptable, my GTX 680 I've yet to increase the offset. WIth a heavier OC my HD 7970 is loud.

 

[Pwndenburg]

Thanks for the input. Nah, It might be beneficial but I'll never OC my vid card. I like the peace of mind that keeping it at stock provides. I guess that's the main question then, acoustics and you say they will be fine. Thanks for the thoughts I guess I'll give that reference a whirl since the noise factor is lower and the reviews say the top temp isn't that bad. I'll admit, it's going to be hard looking at numbers higher than the current arctic accelero min 29c, max 50c

 

[BrightCandle]

Google has been showing over a quite a long period of time that computer components are incredibly tolerant to heat as long as its within the tolerance. You can put them in a surprisingly hot ambient environment and still have similar failure rates and consumption figures.

The data suggests there is no appreciable difference in life length of reference verses customised cooling. It is about acoustics, overclocking and looks mostly.

 

[BD231]

I realize any input I get on this will be anecdotal at best; but, I'd really like to get some opinions on any potential benefits of aftermarket cooling for cards run at stock speeds. Assuming both cards are run at factory settings, would a non-reference cooled card be likely to outlast a reference one due to lower temps?

Its a total crock actually, they just end up blowing all kinds of hot air into your case and don't typically keep the cards any cooler than a blower would because the sink density you'd need to keep the GPU super cool just isn't there. You'd wipe the floor with even the cheapest aftermarket cooler you can buy yourself on any reference/non reference card out there.

The only issue is that for some god awful reason aftermarket cooler makers don't take the time out to make a dual slot coolers for cards and instead chuck compatibility out the window with their 3 slot behemoths. SUCKS. You're either stuck with the potentially annoying blower, a non reference cooler which isn't much better and in some cases worse, or a gigantic bulky thing that takes up 3 pci slots.

 

[Pwndenburg]

Obviously in my case the 3 slot behemoth seems to be the most appealing lol. Then again, I'm certainly not an "enthusiast" in that I only run one High-end GFX at stock. So, I can actually fit the big sucker in there. There is the rumor out there about the vram cooling being insufficient; but, I chose to believe that it is incorrect installation. Btw, do we have the means to measure the vrm temp with the 680 gtx?

 

[Subyman]

Google has been showing over a quite a long period of time that computer components are incredibly tolerant to heat as long as its within the tolerance.

Very insightful

I've never had a card totally die on my, as in it won't post. I've had some slowly degrade where they aren't useful at doing 3D anymore, but those were old cards from the early 2000's. I've had no problem with reference or nonreference cards in recent years.

 

[Pwndenburg]

Well, again, thanks to all for your input. I've decided to simply buy the reference (when they are in stock, aside from the gougers on Amazon) and if I simply can't live with reference then the new Accelero will be available on May 23. The page for it claims that the thermal paste for the VRMS is improved.

 

[taltamir]

I realize any input I get on this will be anecdotal at best; but, I'd really like to get some opinions on any potential benefits of aftermarket cooling for cards run at stock speeds. Assuming both cards are run at factory settings, would a non-reference cooled card be likely to outlast a reference one due to lower temps?

Since you are only compared stock speed cards, and only stock PCB, with the only difference being the cooling.
Then cooler = better.

Any professional review of a non reference card is going to compare its thermals to the reference card, and if they are better then it will last longer.
Whether you NEED that extra longevity or the exact increase is another question altogether.

 

[taltamir]

Google has been showing over a quite a long period of time that computer components are incredibly tolerant to heat as long as its within the tolerance.

I thought that was HDDs.
And it wasn't "amazingly tolerant" it was "the old adage that HDD fail due to heat and must be kept extremely cool is false"

 

[SlowSpyder]

Since you are only compared stock speed cards, and only stock PCB, with the only difference being the cooling.
Then cooler = better.

Any professional review of a non reference card is going to compare its thermals to the reference card, and if they are better then it will last longer.
Whether you NEED that extra longevity or the exact increase is another question altogether.

I'm not disagreeing with you at all, I agree 100%. But, my guess is that something like a GTX480 (very high heat output GPU) will not die due to components failing before the card is so old that you wouldn't want to use it anyway. I wouldn't be surprised if there are GTX480's running five years from now, wait, actually I would be because why would you want to? But my point is that I believe most cards will not have components fail typically before the usefulness of the card is long past.

 

[taltamir]

I'm not disagreeing with you at all, I agree 100%. But, my guess is that something like a GTX480 (very high heat output GPU) will not die due to components failing before the card is so old that you wouldn't want to use it anyway

I tend to think you are right.

This is what I alluded to but did not specifically say in

Whether you NEED that extra longevity or the exact increase is another question altogether.

I didn't want to commit since there has been the occasional dud that broke faster (eg: nvidia's bumpgate)

 

[clockerspiel]

I have an old BFG 8800 GT OCX that I purchased back in 2008 - still running quiet and cool:

http://www.hardwarecanucks.com/foru...-8800gt-ocx-9600gt-ocx-video-card-review.html

 

[Pwndenburg]

Now, that I think about it.. my 8800 GTS is still in service in a friends computer and my even older 6800 GT is in service in another friend who is extremely down on his luck. Guess that answers that, especially coupled with the fact that I upgrade every other year lol. Yep, I'm down to acustics. Given the relative acustic performance of the 680 GTX, the case for waiting for nonreference is getting weaker.

 

[bononos]

Obviously in my case the 3 slot behemoth seems to be the most appealing lol. Then again, I'm certainly not an "enthusiast" in that I only run one High-end GFX at stock. So, I can actually fit the big sucker in there. There is the rumor out there about the vram cooling being insufficient; but, I chose to believe that it is incorrect installation. Btw, do we have the means to measure the vrm temp with the 680 gtx?

There were some issues with non-reference cooling neglecting the vrms for the higher end around 4xxx/5xxx era cards I think which got to the point of being problematic. The gpu temps would be improved over stock but the vrms got hotter vs stock.

I can't remember which gpu series and cooler model since I wasn't gaming much at the time but this link might give some idea of what I'm talking about-
http://www.xbitlabs.com/articles/coolers/display/gelid-icy-vision_5.html


There was a thread where someone mentioned that Nvidia reference models were stingy with dimensioning the vrm capacity and it is just my own musing that it could be the reason for Asus/Gigabyte/others cards with non-reference vrm designs.

More recently there was talk about the xfx 7970 double fan models which had high vrm temps and the aftermarket acceleros for 7970 which gave vrm temps in the 100Cs.

 

[bononos]

Found this old AT thread which mentioned issues with 5850/5870 aftermarket cooling and vrms- http://forums.anandtech.com/showthread.php?t=2048818

 

[BrightCandle]

I thought that was HDDs.
And it wasn't "amazingly tolerant" it was "the old adage that HDD fail due to heat and must be kept extremely cool is false"

Since the initial HD study they have published studies about the failures of other components as well, whole computers being the one I am referring to. More so they have been showing the results of their so called "hot" data centres which run at 50C ambient temperature during the summer. All of the data points to hardware not failing appreciably faster at increased temperatures. So its more than just HDDs, in all Google has published about 10 studies on heat in the data centre.

Its worth mentioning of course that two things limit overclocks:
1) You hit the maximum voltage that the chip will reasonably take before you start destroying its structure rapidly.
2) You hit a thermal limit because you can't cool the increased clock speed and voltage.

All overclocks should only ever be limited by (1). If you are limited by (2) you should get better cooling because its a solveable problem, unlike (1) which is a problem that you can do nothing about.

 

[taltamir]

Since the initial HD study they have published studies about the failures of other components as well, whole computers being the one I am referring to. More so they have been showing the results of their so called "hot" data centres which run at 50C ambient temperature during the summer. All of the data points to hardware not failing appreciably faster at increased temperatures. So its more than just HDDs, in all Google has published about 10 studies on heat in the data centre.

Fascinating. I need to find though studies. lemme google them