Nvidia explains the GTX 970's memory 'problems'

[crossy]

Put yourself in those peoples positions.....you just upgraded your card to a 970. You have either had an issue or your performing a knee-jerk reaction to the media news about the "slow" 0.5GB........you just sold your old card.....what are you going to replace the 970 with?

The old cliche about a fool and his money springs to mind - just drop graphics quality settings a notch if you get stutter. BTW, isn't that pretty much "Fix 101" for ALL graphics stuttering issues, irrespective of NVidia or AMD?

Also, as I have said before AMD drivers have been plagued with issues that they have taken too long to fix, this turns people off, especially the non-tech-savvy who just switch a nVidia card back in and the problem goes away. It's a problem AMD have created for themselves through heel dragging.

Actually I've been pleasantly surprised how trouble-free the AMD Windows drivers have been. They cause me less hassle than an iTunes update does these days.

The problems nVidia have had wouldn't effect them so much as their biggest function-effecting mistake only really effected laptops. How many people check the video card in a laptop before purchase? Or buy an AIB VGA card for a laptop?

Aren't most laptops using the embedded graphics? (In which case you're definitely correct about folks not bothering with spec checking) Certainly the kind of pile 'em high and sell 'em cheap that you see in PC World, Lewis's, etc. In fact I'd go further and say it's only really the high end lappies and gaming ones that use discrete graphics cards/modules and someone buying a gaming PC will definitely be interested in the graphics engine. A high-end buyer's probably more interested in outside appearance etc.

[tribaljet]

...

Aren't most laptops using the embedded graphics? Certainly the kind of pile 'em high and sell 'em cheap that you see in PC World, Lewis's, etc. In fact I'd go further and say it's only really the high end lappies and gaming ones that use discrete graphics cards/modules and someone buying a gaming PC will definitely be interested in the graphics engine.

Actually there is a surprising number of cheap laptops that come with Optimus/Switchable Graphics. Now, whether the discrete graphics are high performing, that's a whole different matter, but they pretty much always outperform the CPU's IGPs.

[shaithis]

Just because some people find the AMD drivers fine, doesn't detract for the slew of problems they have had. The damage was done a while ago.

If their hardware and drivers were so damn good, why have nVidia being constantly eating into their market share? Why do people seem more likely to go AMD > nVidia rather then the other way round?

The one time recently we saw a glut on AMD cards was when mining peaked.....did you happen to see how many cards fried or were unstable while mining? And as soon as that craze ended the AMD cards were dumped like hot potatoes.

But hey, I am sure we will go round in circles again with people saying AMD drivers are no worse/better.....and that it's all some conspiracy between nVidia and the press to try and slay AMD for some random reason.

[CAT-THE-FIFTH]

Just because some people find the AMD drivers fine, doesn't detract for the slew of problems they have had. The damage was done a while ago.

If their hardware and drivers were so damn good, why have nVidia being constantly eating into their market share? Why do people seem more likely to go AMD > nVidia rather then the other way round?

The one time recently we saw a glut on AMD cards was when mining peaked.....did you happen to see how many cards fried or were unstable while mining? And as soon as that craze ended the AMD cards were dumped like hot potatoes.

But hey, I am sure we will go round in circles again with people saying AMD drivers are no worse/better.....and that it's all some conspiracy between nVidia and the press to try and slay AMD for some random reason.

So how come Nvidia marketshare was not smashed during the rubbish FX era then?? The one that Valve had to degrade image quality and framerates in HL2,which was on of the biggest titles launched until Crysis. Or how come the iPhones outsell all their competitors all the time despite massive hardware problems??

Its funny how you rubbish Apple fans,yet they use all the same arguments as you.

Why did VHS,do better than superior formats?

I assume you never buy milk then? Most milk is sold at a lost and the UK dairy industry is collapsing.

In fact why are you so obssesed about marketshare - in fact everytime you go into a shop do you bother about things like marketshare or the next thing a person might buy??

So I assume you only buy stuff at Tesco or off Amazon. Only drive a Ford or a Toyota and so on.

I assume you must love Eastenders too.

Its funny how I know nearly 20 gamers myself with everything from an HD7850 2GB to a R9 290X 4GB,and cards from a GTX750TI to a GTX980. Yet they appear not to have all this problems you talk about,and you keep spreading more and more fud all the time,and that includes one chap I know who has dual R9 290X cards and a GTX980 rig.

Hilariously when I listed one or two minor issues with my GTX660 you instantly went bezerk the last time and try to rubbish all of it,and that was issues with the 330 branch(known issues) that even some of the Nvidia fans I know over on other forums said was meh.

Also,how come none of the R9 290X/R9 290 owners in real-life and the multiple people on OcUK I know over there still have functioning cards??

You seem to conflate reliability and drivers as some indication of why people buy Nvidia - its actually people like you who peddle the same all crap all the time,being the reasons people still chose Nvidia even with the FX series,or when the had mass failures of GPUs costing them $200 million.

Sounds massively like the reality distortion field round Apple.

Edit!!

Heck,lets look at something like laptop marketshare where Nvidia has done well down to technical reasons. The GK104 had similar performance/watt to lower end AMD GPUs but was another performance bracket above - so no wonder they took marketshare.

It was the mobile market where AMD score high during the HD5000 series,and any gains in the market were removed by Kepler,which is where Nvidia score highly,and even GPUs like the GM107 and GM206 are further improvements on this strategy that AMD cannot seem to compete with.

But the drivers rubbish you spout is hilarous especially when PCPER who is close to Nvidia,has shown on multiple occasions that the AMD desktop cards have done very well in terms of playability.

However,its even funnier that in the professional markets where Nvidia had 90% of the market 3 years ago,AMD has reach nearly 30% - must be those **** drivers and exploding hardware,no??



PS:

I already know how you are going to answer,so don't bother!

[watercooled]

did you happen to see how many cards fried or were unstable while mining?

Frankly, no. And it's something I followed quite closely at the time. One or two people complaining they'd destroyed their card doesn't equal a defect.

One thing I did see was certain models of cooler having fan bearings leak, but that was due to fans being manually set at 100% and run 24/7 which they were never intended for, and absolutely nothing to do with AMD.

Edit: Oh and one other thing was one specific model of IIRC 7970 from one manufacturer had a defect in the VRM circuitry, causing some to fail. Again, nothing to do with AMD, and something the MFR fixed in future versions.

[CAT-THE-FIFTH]

Its no point watercooled,shaithis will keep spouting his AMD cards have unreliable hardware and software crap til the cows come home which is his right OFC. But I feel sorry for all the people asking him for tech advice who he will mislead with his nonsense. None of the dozens of people I have suggested Kepler,Maxwell or GCN cards in the last two years have seemed to have any massive issues.

I have had close to two dozen cards from each company since 2003. Magically I never had these MASSIVE software problems.

Edit!!

Plus he will ignore any issue Nvidia has had with their hardware and software too.

Pfft.

Second Edit!!

The only card which ever died on me was an 8800GTS 512MB,and my 6800LE had dodgy capacitors which affected multiple cards from several companies,but my card worked fine when I had it. I had to help close to 10 people who had problems with their G8* GPUs in their laptops.

Using his logic, NVIDIA SUCKORS! MAKES HARDWARE WHICH EXPLODES AND DIES. NEVER BUY N-V-I-D-I-A! INTEL RULES! THEY HAVE MORE MARKETSHARE IN GRAPHICS.

Third Edit!!

It kind of reminds last week of a conversation I had with a random person in a pub. He looked at my phone and what all my mates had(which were Android phones from different companies) and instantly said they were all crap,since the iPhone 5 he had before and his current iPhone 6 were better than all Android phones and all of his mates were changing over,and it had better sales figures than the top end Samsungs meaning it was better.

I had similar interactions with Mac fans who said Mac desktops were better than any PCs,and that I should switchover to OS X and after I used it I would not go back to crappy Windows and PCs which were unreliable,hard to use and prone to viruses.

What they had not realised is that I had used Macs with OS 8,OS 9 and OS X beforehand.

[CAT-THE-FIFTH]

Supposedly the GTX980M might have the same issue:

http://tech4gamers.com/nvidia-geforc...emory-problem/

However,one of the comments is saying the opposite. Might need more investigation although I expect its clickbait.

[spl]

To actually put some numbers on it so we have quantifiable evidence instead of general fanboy crap ("I have 800 friends who have no problems bla bla..."), this is quite interesting:

http://www.hardware.fr/articles/927-...raphiques.html

Both sides have some very poor failure rate models, but generally, the rate for Radeons is considerably higher. Maybe it's not a case of one being great and the other being crap (or the common "both are great") but rather, both suck, but Radeons suck more. I'd say returns rates are far too high on both sides. I'm sure our 'militant battle moose' and watercooled will figure out some way to dispute that though :-)

By the way, Chrome's built-in Google translate for this (at least for me) wrecks the content layout, so if you want a translation I'd suggest copying and pasting the bit you want translated straight to another tab.

P.S. Cat - "I already know how you are going to answer,so don't bother!" reads as "Listen to what I have to say and then shut up and don't prove me wrong." - you're honestly better off leaving that bit out.

[jigger]

How many of those got mined to death in lofts and cellars around the world?

- 15,23% Sapphire HD 7950 With Boost (11196-19)
- 13,85% HIS R9 280X iPower IceQ X² Boost Clock
- 11,97% ASUS Radeon R9 280X R9280X-DC2T-3GD5
- 11,31% Sapphire Radeon R9 280X Dual-X OC UEFI 11221-07

[watercooled]

To actually put some numbers on it so we have quantifiable evidence instead of general fanboy crap ("I have 800 friends who have no problems bla bla..."), this is quite interesting:

http://www.hardware.fr/articles/927-...raphiques.html

Both sides have some very poor failure rate models, but generally, the rate for Radeons is considerably higher. Maybe it's not a case of one being great and the other being crap (or the common "both are great") but rather, both suck, but Radeons suck more. I'd say returns rates are far too high on both sides. I'm sure our 'militant battle moose' and watercooled will figure out some way to dispute that though :-)

By the way, Chrome's built-in Google translate for this (at least for me) wrecks the content layout, so if you want a translation I'd suggest copying and pasting the bit you want translated straight to another tab.

P.S. Cat - "I already know how you are going to answer,so don't bother!" reads as "Listen to what I have to say and then shut up and don't prove me wrong." - you're honestly better off leaving that bit out.

And... AMD and Nvidia actually make the cards right?

The significant differences in return rates between different brands/models of the same exact GPU show this. Faults of actual processors are exceedingly rare.

Not to mention it's far from an exhaustive list so hard to draw any realistic conclusions - look at how the media are making a huge deal about these Backblaze HDD failure rates; quite a few places, IIRC even Backblaze themselves, have explained why it's dangerous to over-extrapolate like that, if it wasn't already obvious.

Also, take some time to understand what you're reading with those numbers as they're potentially quite misleading. The per-card averages only seem to take into account cards above a certain threshold, so particular models of card with exceptionally high return rates skew the numbers.

For instance, look at Hawaii; 4.38%. But according to their 'aggregated' numbers, 6.1%. See how that works? Strange for a shop to only sell two of each 290/X, so probably not a big shop which comes back to what I said about limited sample size, but hey lets not worry ourselves over fanboy crap like that eh?

GK110 according to their figures = 3%. Averaging all the listed cards, 5.23%. Funny that isn't it? And incidentally higher than Hawaii either way you look at it. Not that I'd cite that as evidence, because it's not.

And I'm not overly criticising the article here, it doesn't look like the ever set out to compare AMD to Nvidia failure rates, hence the comparison was not set up fairly for it. This is well known in the scientific field, you spell out your intentions, methods and expected findings beforehand. You don't pull a load of data at random from a small sample size, filter out stuff you don't like on a whim and call it conclusive. I could make a new water cure for colds looks exceptionally good if I filtered out those pesky inconclusive results.

Edit: They even mention cryptocurrency mining in the article, as jigger said above. In many countries, Tahiti stock was constantly emptied by miners as it was far and away the best value card to do it with, even at the hugely inflated prices it was selling for at its peak.

[kalniel]

To be fair, if I remember correctly from the last time this came up, they're filtering out low sample size results, not inconclusive ones. That's why averaging cards also isn't satisfactory because you (presumably) didn't account for the numbers of each model.

But you're right - it's nothing like any approximation of failure per usage hours or anything.

[spl]

If anyone has better data about failure rates then I'd love to see it. This is the best data I know of for comparison of brand reliability.

[watercooled]

To be fair, if I remember correctly from the last time this came up, they're filtering out low sample size results, not inconclusive ones. That's why averaging cards also isn't satisfactory because you (presumably) didn't account for the numbers of each model.

Filtering out low sample size results also skews the AMD/Nvidia comparison then, as models should be irrelevant if you're doing a dumb GPU to GPU comparison.

Fact is, the numbers they give work out differently depending on which part of the article you use them from, I've no idea what's going on with them. Some raw data might be useful, a load of random percentages without anything resembling a decent explanation of where they came from is not. They're probably useful for the intent of the article, which is AFAICT rates for AIB's, but as I said you have to be extremely careful of original methodology if you're pulling data to make other comparisons.

E.g. If I had a shop which sold cables, some USB with a low failure rate, some 1394 with high failure rate. Out of that, and providing I'd sold enough to draw some reasonable conclusions, I could state that cable A had more failures than B, possibly due to their brand. However, someone else could not sensibly take that data and say that USB cables must be more reliable than 1394 cables, even though at a high level, that's what the data implies.

Edit: Just to clarify, I was speaking hypothetically about the 'inconclusive results' bit, not suggesting that's what had been done in the article. However there is obviously some results filtering going on, which greatly distorts numbers for AMD-Nvidia comparisons drawn from the data displayed.

[watercooled]

If anyone has better data about failure rates then I'd love to see it. This is the best data I know of for comparison of brand reliability.

As far as I can tell it's from a small retailer selling an (understandably) limited selection of cards. It's not a comparison of brand reliability, that much should be obvious.

Edit: There was a similar myth about Intel being more reliable than AMD CPUs. That got destroyed by a large randomised sample with no cherry-picking: http://www.extremetech.com/gaming/13...thusiast-myths

[crossy]

If anyone has better data about failure rates then I'd love to see it. This is the best data I know of for comparison of brand reliability.

Problem is that this kind of data is quite hard to get - manufacturers are notoriously (and understandably) not keen to hand it out. I also hear that they "lean" on the major sellers to not release this kind of info either. I'm sure the figures from one of the large US operations - like BestBuy - might be interesting.

Shame that there's not more in the way of raw data available - I'd like to dig deeper and see if perhaps some of the results were due to a particular card from a specific manufacturer, although I see that HIS get singled out for criticism. Then again, no XFX?

[HalloweenJack]

since were talking failed hardware - shall we discuss the NVidia drivers which turned off the fan and thus directly caused GTX 480 and 580 cards to burn up?