AHO: Memory Inductors Are Frequently Burning Out on Certain NVIDIA GPUs

im_shocked.gif

Nvidia is sure on the shady train lately.
 
I'll have you know the last 5 AMD cards I bought have died on me, but my Nvidia GTX Awesome is still running just fine. This is clearly AMD making false claims on Nvidia.

[sarcasm]Yes, I'm joking. If I don't do this, someone here will think I'm an Nvidia fanboy. [/sarcasm]
 
Both of my Kepler Titans (that I fortunately paid not much for off FS/T but were just out of warranty) blew an inductor after a couple of months doing what the card was built to do: DP compute. At least now I feel like it wasn't anything I did wrong, but still, it hurts.
 
Actually Hardcore Overclocking has published a video alleging that NVIDIA has been skimping on memory inductors and other components for their reference GTX 780/780 Ti/Titan(XM)/980Ti cards. Numerous photos of PCBs with burned out inductors are shown, and some commentators are calling this planned obsolescence.

The planned obsolescence comment from about 8 hours ago was me.
 
Would be interesting if they could compile a list of which brands do it the most. I've had a 780 6gb and a 980 that I bought when they were first released die this way within 2 years and they were both EVGA SC models. My other cards are doing just fine at the moment: Titan (Asus), Maxwell Titan X (Nvidia), Titan XP (Nvidia).
 
I wonder if you can replace those inductors as preventative maintenance? You would need to be good at surface mount soldering.
 
After watching the video I have a question. Was it Nvidia who designed the board, or was it the board manufacturers? Cause unless these are based on the reference design, this maybe the fault of the board makers. Like XFX, Asus, and etc.
 
It says reference cards, so it is nvidia.
Like 100% reference? Cause that's one hell of a cheapening from Nvidia. From what the guy says, Nvidia had really went cheap with the inductors. And he was being very generous with his estimates.
 
The only good thing for Gamers is for nVidia to get dethroned.
And where is the equal competition beyond the $400 MSRP pricepoint? Yeah... it's non-existent.

I'd love to see AMD be a serious competitor beyond the mid-range, hell I'd like to see 2 or 3 competitors. The reality is that they don't exist, and it's not as if the market for high end GPUs is going away any time soon. Considering Nvidia has basically had that market segment not only cornered but managed to find new markets to re-sell that technology to over the past 5 years, they aren't going to get dethroned any time soon. Even Intel didn't take much of a hit in the long run when they were playing stupid games, it's not going to turn out any different for nvidia.
 
Remember the huge stink about the PCIe-o-caust the Polaris cards would cause, hyped by the usual suspects here, which became a big FUD of nothing?
Well it's funny how that shoe fell on the other foot.
 
Not surprised, I’ve had a few nvidia cards go bad from memory alone.

My AMD/ATI cards had their issues too but they were all over the place.
 
Hanlons Razor said:
"Never attribute to malice that which is adequately explained by stupidity."

after watching that YT (And damn that person is annoying... wtf did he have to list ever-single-gpu-using-switcher... ) and a few comments here. People are postulating that this was a design intent. As someone that designs powersupplies ... A more plausible explanation is a design oversight... IF the ATX psu (full range) spec wasn't designed for and the 12V coming in was dirtier than the ideal design case then the inductor could be hit with more losses.
 
One is a fluke, two a coincidence, three is a trend. Even if everything is on the level with this particular incident, the signs that Nvidia has lost its way are strong.
 
One is a fluke, two a coincidence, three is a trend. Even if everything is on the level with this particular incident, the signs that Nvidia has lost its way are strong.
I agree, I question the premise that this is malice and I would lean towards design oversight (more than likely due to drop in design assurance within NV). I work in the electronics industry in OE, MRO and it is all to easy to have an oversight in a design.
 
You have to be fucking kidding me? With the price we've been paying they've been shorting us on quality. I'm fucking pissed tbh

you almost make it sound like this is the first time Nv have done this (with full knowledge of what they were doing) from Geforce 200 series all the way up to "modern" they find the best way to build as "cheaply" as possible to maximize revenue over maximizing quality even if it means a bit less $ (some rare exceptions, but they are rare with Nv)

vreg etc built for 105/85c instead of 125/115c (rely on fancy digital circuits to prevent potential overheat which is not a sure fire will never happen)
there are many other reasons ofc, but, just not worth wasting time pointing them out.
Nv is a shaddy company, "quick" is one thing but overall quality is another, AMD has one thing going for them outright and that is quality of the finalized product. the "core" design of Radeons (that is the pcb, memory, vregs, solder etc) is bar none superior to what Nv offers, no they are not always faster but they are competitive overall (with more robust features that are not proprietary BS just for the sake of being so)

If anything can be said, if AMD was not always looking to be as good as they can be, we would not have the modern things we have or are used to having on various computer products
such as eutectic solder being used (or should be) as solder instead of lead (for the high temperaure high speed circuitry i.e vregs) GDDR (mainly used by ATi/AMD then a while later sometimes used by Nv once the kinks were worked out) HBM, tessellation etc etc....

anyways, simply put would not surprise me, am sure Nv will post a "it is users fault" reply to such hate speech on their product, they always find a way to downplay where they fucked up yet people still buy their products like they are the best thing since sliced bread no matter how badly they screw the dog, speed is not everything, it is but a factor, people need to take it as only a factor IMO nothing more.

Analogy
A "car" that does 80mph will get you to the store just as easily as the "car" doing 90mph, I rather the "car" that does 80mph and can do it for years (even if uses a bit more fuel and gets a little warmer) than the "car" that does 90mph and you are lucky that it lasts a year or 2 before you have to go and buy another expensive one that uses shiny paint with the same shitty door hinges ^.^
 
Unless something changed AMD's failure rates are much higher:

https://hardforum.com/threads/amds-failure-rates-part-two-2015-edition.1899524/
The source which is French was updated and here's the highest offenders.

https://www.hardware.fr/articles/947-5/cartes-graphiques.html
  • 10,34% Sapphire R9 290 4G GDDR5 Tri-X OC (New Edition)
  • 9,80% Inno3D iChill GTX 980 Ti X3 Air Boss Ultra
  • 8,85% Inno3D iChill GTX 980 Ti Black Series
  • 7,59% Sapphire R9 290 4G Vapor-X
  • 6,47% Gigabyte GV-N970IXOC-4GD
  • 5,29% MSI GTX 960 2GD5T OC
So yea, Sapphire is like really bad, but so is Inno3D, which I've never heard of. But also their sample size is small. The only R9 290 they show is from Sapphire, which the GTX 980Ti has Inno3D, EVGA, MSI, and Asus. Asus has like 0% failure rates, which either means that Asus has some amazing quality or that the sample size was too small.

The only real conclusion you can get from this is stay away from Sapphire and Inno3D when it comes to buying a R9 290 and GTX 980 Ti. There's not enough data here to show that AMD or Nvidia has a higher failure rate. Cause again, I'm sure other's have made a R9 290 besides Sapphire. If I limited the sample size to just Inno3D for the 980Ti, you would say that Nvidia has a higher failure rate.
 
I had an Inno3D iChill 980 Air Boss something something and it became unstable at idle despite being ok under load.
It was a handy failure because the ti has just come out so got a refund to buy one - not an Inno3D.
 
The source which is French was updated and here's the highest offenders.

https://www.hardware.fr/articles/947-5/cartes-graphiques.html
  • 10,34% Sapphire R9 290 4G GDDR5 Tri-X OC (New Edition)
  • 9,80% Inno3D iChill GTX 980 Ti X3 Air Boss Ultra
  • 8,85% Inno3D iChill GTX 980 Ti Black Series
  • 7,59% Sapphire R9 290 4G Vapor-X
  • 6,47% Gigabyte GV-N970IXOC-4GD
  • 5,29% MSI GTX 960 2GD5T OC
So yea, Sapphire is like really bad, but so is Inno3D, which I've never heard of. But also their sample size is small. The only R9 290 they show is from Sapphire, which the GTX 980Ti has Inno3D, EVGA, MSI, and Asus. Asus has like 0% failure rates, which either means that Asus has some amazing quality or that the sample size was too small.

The only real conclusion you can get from this is stay away from Sapphire and Inno3D when it comes to buying a R9 290 and GTX 980 Ti. There's not enough data here to show that AMD or Nvidia has a higher failure rate. Cause again, I'm sure other's have made a R9 290 besides Sapphire. If I limited the sample size to just Inno3D for the 980Ti, you would say that Nvidia has a higher failure rate.

Ouch, holy crap 10% is high and I agree with you.
 
Couldn't it just be that the video cards that are burning out are just old and naturally failing over time, or is there something else at work here?
 
If not mistaken, Nvidia locks voltage adjustment in the Bios. AMD you still have overvolting without modification. AMD tends to brute force core voltage as well too.
 
Couldn't it just be that the video cards that are burning out are just old and naturally failing over time, or is there something else at work here?
Everything eventually fails of course, but in this case is just a stupid oversight that makes little sense. Im calling it an oversight, high failure rates can lead to the complete ruination of your brand, and Nvidia wants to make sure you know its their brand you are buying, it would be ruin for them.
 
Like 100% reference? Cause that's one hell of a cheapening from Nvidia. From what the guy says, Nvidia had really went cheap with the inductors. And he was being very generous with his estimates.


Wait up, any manufacturer that makes cards could "cheap out" on their reference boards, NVidia doesn't make those for the manufacturer, the manufacturer makes the boards off the reference spec, but they don't have to follow it line by line. The reference design provides manufacturers a ready to go solution, use these parts this way, add our software and change the logo, and get to market while the market is hot.
 
you almost make it sound like this is the first time Nv have done this (with full knowledge of what they were doing) from Geforce 200 series all the way up to "modern" they find the best way to build as "cheaply" as possible to maximize revenue over maximizing quality ................................

How is it you can say this? Until recently, NVidia didn't even build video cards. They designed the chipsets and the architecture, hey laid out a "reference card" with reference drivers/software. Then NVidia licenses the solution to the builders, ASUS, EVGA, MSI, Gigabyte, etc who actually built the cards. The first NVidia actually started building cards for retail was the Founders Edition 10 series cards. All those years prior, the reference cards, were not built by NVidia.

I think you need to verify what I am saying, make sure I am right or find out if I am wrong. Because I think it will have a huge impact on how you perceive the devil ... huunnnmmm, NVidia.

EDITED:
https://hardforum.com/threads/what-are-reference-cards-and-non-reference-cards.1730682/
 
Last edited:
Couldn't it just be that the video cards that are burning out are just old and naturally failing over time, or is there something else at work here?
The inductors are just a coil of wire. They shouldn't go bad due to age. Not unless they're given more then they're meant to handle, which they were.

Wait up, any manufacturer that makes cards could "cheap out" on their reference boards, NVidia doesn't make those for the manufacturer, the manufacturer makes the boards off the reference spec, but they don't have to follow it line by line. The reference design provides manufacturers a ready to go solution, use these parts this way, add our software and change the logo, and get to market while the market is hot.

One way to see if this is truely Nvidia's fault is to look at their reference boards made by Nvidia themselves. If we see the same shit inductors then it's Nvidia's fault. Otherwise it's the board makers for being cheap.
 
One way to see if this is truely Nvidia's fault is to look at their reference boards made by Nvidia themselves. If we see the same shit inductors then it's Nvidia's fault. Otherwise it's the board makers for being cheap.

I'll say it again, NVidia does not make reference cards. NVidia designs reference cards. NVidia licenses reference cards, but NVidia doesn't make reference cards, not until the 10 series. And even then, technically, the Founders Edition cards were not reference cards.

Therefore, you could say that NVidia "fucked up" the engineering on the card. But they wouldn't gain by doing this unless you see it as a planned obsolescence kind of thing and you know what ..... manufacturers usually don't maintain reference cards through the life of the product cycle. After awhile they all turn into non-reference cards either beefed up for the enthusiast, or cheapened with cut corners for the bargain hunters. Reference designs are good for one thing, getting your product on the shelves fast, with all the others.
 
Is that why my 980ti stopped working after 2 years? It was only factory OCed and I don't tend to push it and my system if very well cooled. So wonder how long it will last after I get it back? Maybe look at AMD for my next card? I usually try get the best price/performance/longevity I can afford but if that last thing doesn't apply AMD might look better esp. @ 4k.
 
Anyone
The source which is French was updated and here's the highest offenders.

https://www.hardware.fr/articles/947-5/cartes-graphiques.html
  • 10,34% Sapphire R9 290 4G GDDR5 Tri-X OC (New Edition)
  • 9,80% Inno3D iChill GTX 980 Ti X3 Air Boss Ultra
  • 8,85% Inno3D iChill GTX 980 Ti Black Series
  • 7,59% Sapphire R9 290 4G Vapor-X
  • 6,47% Gigabyte GV-N970IXOC-4GD
  • 5,29% MSI GTX 960 2GD5T OC
So yea, Sapphire is like really bad, but so is Inno3D, which I've never heard of. But also their sample size is small. The only R9 290 they show is from Sapphire, which the GTX 980Ti has Inno3D, EVGA, MSI, and Asus. Asus has like 0% failure rates, which either means that Asus has some amazing quality or that the sample size was too small.

The only real conclusion you can get from this is stay away from Sapphire and Inno3D when it comes to buying a R9 290 and GTX 980 Ti. There's not enough data here to show that AMD or Nvidia has a higher failure rate. Cause again, I'm sure other's have made a R9 290 besides Sapphire. If I limited the sample size to just Inno3D for the 980Ti, you would say that Nvidia has a higher failure rate.


I have the 980 ti EVGA SC 2.0, so I started to look up 980 ti pcb's I couldn't really find my specific one but the Asus ROG memory inductors are massive and look pretty custom on those cards you can't even tell what they are. The EVGA cards I could find in the 980ti like the FTW looked bigger then reference and the classifieds have what looks like a fairly bigger custom inductor with an er33 indication on them (not sure what the e means) and it looks like their using a r44 inductor at the 2x4 pin power to boot.

What does it all mean, I Have no clue but I'm definitely spending a bit more for an ROG or a higher tier EVGA card next upgrade..as someone else said too, I'm wondering what if anything this has to do with PSU quality as well.
 
Back
Top