Thursday, January 29th 2015

Perfectly Functional GTX 970 Cards Being Returned Over Memory Controversy

In what is a major fallout of the GeForce GTX 970 memory allocation controversy, leading retailers in the EU are reporting returns of perfectly functional GTX 970 cards citing "false advertising." Heise.de reports that NVIDIA is facing a fierce blowback from retailers and customers over incorrect specs. Heise comments that the specifications "cheating could mean the greatest damage to the reputation of the company's history."

Major German PC hardware retailer Caseking.de says that retailers don't have any explanation from NVIDIA to give to their customers. A similar sentiment is being expressed by the NVIDIA add-in card partners (AICs) we spoke to. Retailers and AIC partners are on their own, for now. One AIC partner rep told us that NVIDIA has no worldwide action plan, as of now, to deal with a potential flood of returns.
In absence of every other recourse, laws in most EU member states dictate that the retailers accept returns for a full refund, if they are not able to "repair" the defect, or exchange with another unit that works as advertised (which a retailer obviously can't, in this case). Retailers' options in the matter boil down to: 1. Taking back cards from whoever isn't happy with their GTX 970 and giving them a refund; 2. compensating with something of value (eg: game-coupons, in-game currency, etc.,) and 3. Springing up a surprise, such as exchanging GTX 970 cards purchased before a set date, with a GTX 980 (if that's your idea of a "repair."). This will come at the expense of a cascading lawsuit-chain (customers suing retailers, who in-turn sue AICs, and who in-turn sue NVIDIA).

NVIDIA, on the other hand, plans to issue a driver update that will "improve" the way the chip allocates resources, but there's no word on whether it re-enables disabled components that NVIDIA wasn't honest about, the first time around. They're counting on the issue to simply blow over, because at $329, there really isn't much you can complain about the GTX 970, given how it's positioned in comparison to the GTX 980.
Add your own comment

172 Comments on Perfectly Functional GTX 970 Cards Being Returned Over Memory Controversy

#76
R-T-B
btarunrHeise comments that the specifications "cheating could mean the greatest damage to the reputation of the company's history.".
Seriously? No one remembers the FX series benchmark cheating they got caught doing?

This is like nothing by comparison. At least in my books.
Posted on Reply
#77
xfia
El_MayoWOW I don't see how that post sounded troll-like, what do you mean?



Oh yeah it's a good card, but I wonder if I'll see any noticeable difference in performance if I switch to a 960 at 1920x1080. But I do wanna get a 4K monitor eventually so I'll probably keep it
even if you had the full 4gb it would still not be ideal for 4k gaming with 2.
Posted on Reply
#78
REAYTH
R-T-BSeriously? No one remembers the FX series benchmark cheating they got caught doing?

This is like nothing by comparison. At least in my books.
That and outside like 5 tech forums the vast public will have no idea. This is just another "Nerdageddon" that a few people will necro in a AMD driver release thread.
Posted on Reply
#79
xfia
dont hate on the AMD drivers :peace: they have gotten better every few months since WQLH and I find the Omega drivers to be pretty nice.
Posted on Reply
#80
Xzibit
I see this scenario also making the issue stick around

A.) 970 Happy user

B.) 970 Unhappy user

C.) 970 User experiencing problems

B user get a refund for his card. Retail sells open/refurbished 970 at lower price. User A buys at discount, now SLI since he was playing at 1440p with no issue decides to turn up detail now or play at 4k. Starts experiencing problems C user is having.
Posted on Reply
#81
Fluffmeister
XzibitI see this scenario also making the issue stick around

A.) 970 Happy user

B.) 970 Unhappy user

C.) 970 User experiencing problems

B user get a refund for his card. Retail sells open/refurbished 970 at lower price. User A buys at discount, now SLI since he was playing at 1440p with no issue decides to turn up detail now or play at 4k. Starts experiencing problems C user is having.
D.) Buys 8GB GTX 980 Ti
Posted on Reply
#82
Uplink10
Serves them right. Quotes from The Wolf of Wall Street:

That is what I hoped it would happen because they have such high prices:

00:49:57,828 --> 00:50:01,649
One of these days, the chickens are gonna come home to roost.

And now:

02:49:08,430 --> 02:49:11,158
<i>"The chickens had come home to roost."</i>
Posted on Reply
#83
BorisDG
FluffmeisterD.) Buys 8GB GTX 980 Ti
980 Ti will be 100% 6GB of memory to deal with 384bit bus.
Posted on Reply
#84
AsRock
TPU addict
qubitTeacups and storms come to mind.
Agree and disagree, if your one of those who replace your GFX card yearly or a little over for the most part i agree but for those like my self who keep a card 3-5 years the 3.5GB is going to get a load more often there fore would see the issue more often by time they replaced it.

If some ones returning a 970 due to just so they can get a 980 i am truly lmfao @ you. You really think nVidia are going to learn if they end up getting more money of you in the end.

If i had one i would return it due to the fact i said above but i do wounder how bad it really is. But still i would as telling people they did not know about this is just total BS.

All so i would feel like i been conned too but not having one to try the current games i play that take more than 3.5GB would lead me to sending it back.

Next time nVidia will thin twice and make sure what they are selling what they say they are.

What will not help is it's near tax return too and new toys are always fun.
Posted on Reply
#85
Lionheart
btarunrWe're all nerds on this forum, for knowing what a graphics card is. Else we'd be gaming on an Xbox.
Oi that's cruel, I love my updated VCR :cool:
Posted on Reply
#86
L337One91
"Perfectly functional GTX 970..."

This sounds like shilling.
Posted on Reply
#87
bwat47
GC_PaNzerFINWhat I see going on around is that ANY problem with ANYTHING related to GTX 970 is getting blamed for memory allocation now. Even game engine related bugs, other driver issues, SLI microstuttering and not being capable of handing +30 FPS at ultra high settings at 4K resolution.

No doubt there are some with valid driver issues related to memory allocation. But come on, this is a huge storm in a glass. Mass panic. Few people were ready to return GTX 970 and buy MUCH WORSE card from AMD. Does not make sense at all anymore.

NVIDIA made a huge PR mistake handling this. No doubt they should make official company statement and give some kind of compensation for the pain like game bundle.

Also I would love to see NVIDIA give users option to disable this last 512 MB partition from drivers if they experience problems. !
This. During this entire controversy people have been running around claiming that any and all stuttering issues must be the fault of the last 0.5gb vram, with little to no compelling proof. Most videos/benchmarks attempting to 'prove' this issue involve loading up games with ridiculous settings like ridiculous amounts of downsampling which will almost certainly result in hitting other bottlenecks on the card making it impossible to determine if its the fault of the vram or not.

I wouldn't really call an r9 290/290x a 'much worse' card though, they are pretty close in performance.
Posted on Reply
#88
R-T-B
REAYTHThat and outside like 5 tech forums the vast public will have no idea. This is just another "Nerdageddon" that a few people will necro in a AMD driver release thread.
As much as I'd like to believe people are reasonable and would be like that, the returns suggest the general public does know and cares... Heck I'll take a second hand one at this rate!

PS: I'm not a fanboy. I'm running an R9 290X right now and love it for it's own qualities. But the 970's energy consumption would be excellent for another PC without tripping my UPS!
I wouldn't really call an r9 290/290x a 'much worse' card though, they are pretty close in performance.
I agree there. If it weren't for energy consumption I'd call them equal. As it is, 970 is a little better IMO but I gotta hell of a deal on my R9 290X
Posted on Reply
#89
xorbe
What IS the desired outcome anyways? I say, fix the specs (3.5GB, 224-bit, 56 rop), and add a driver option to disable the unwanted 512MB side-port vram. That's it.
Posted on Reply
#90
GhostRyder
GC_PaNzerFINNo doubt there are some with valid driver issues related to memory allocation. But come on, this is a huge storm in a glass. Mass panic. Few people were ready to return GTX 970 and buy MUCH WORSE card from AMD. Does not make sense at all anymore.
Uhh I think you need to look again because the 970 is normally below or on par with the R9 290X and the 290 is not much below it...

Either way this was expected, people are going to be outraged one way or another and you cannot expect anything less. People do not like being lied to on products especially when it comes to alternatives to other products on the market because things like that can influence decisions on purchasing products. When people drop a lot of money on a products they have certain levels of expectation no matter what it is, and when its something that actually does cause a difference they are likely to make a stand. All in all its healthy for the consumer because it means that things like this will not happen (Or will slow down) in the future.
Posted on Reply
#91
Lorenzo
Ok, first, those are not perfectly functional 970s, every single 970 has the same uncorrectable memory problem, on the other hand, nVidia broke the law when they didn't disclose the real specs of the product.

The shouldn't get a free pass just because.

The card still performs in a stellar fashion, yes, that does not change the fact that nVidia knowingly deceived its customers, unless you buy the miscommunication error explanation, and even if that was true, their customers shouldn't have to pay for the company's mistakes.
Posted on Reply
#92
RichF
SeverusEven if there's a slight drop in performance, at the price point you still don't have any better option. Why bother returning it and adding another couple hundred dollars for a 980? It won't be worth the trouble and it won't be a good time / price / performance deal.
I'll just wait for whatever Nvidia will offer us for the trpuble and be happy with it :)
It's not a slight drop. It's a massive drop in terms of bandwidth for that second VRAM partition.

People who bought two or more of these cards for 4K gaming should be pissed. VRAM totals are very important for making SLI decisions in particular.
Posted on Reply
#93
RichF
"Perfectly function" is misleading wording because it depends upon whose point of view you're talking about.

From Nvidia's point of view they're perfectly functional because that's how they were designed.

From a buyer's point of view they're faulty because they do not contain the design they were sold.

When OCZ switched the NAND in its Vertex 2 ssds from 32-bit to 64-bit without telling anyone and without changing the specs printed on the box, people were upset. Not only did they lose capacity, they lost performance as well.

According to the point of view of this topic, though, those 64-bit NAND Vertex 2 drives were "perfectly functional".
Posted on Reply
#94
mxp02
xfiaahh a 970 has no big advantage over your 780.. give it a little overclock and it will be right there most of the time.

the 970 is certainly not mid-range though.. the 780(ti)-970-980-290(x) are all in the same high end performance class that do perfect 1080p gaming. people with 4k rigs hardly understand where graphics technology is at and pay a shit load for something they will soon need to upgrade.

in reality just going by performance and nothing else it would be crazy to buy a 980 and a 290 gives you the best bang for your buck so it was kinda garbage people saying the 970 has the best price to performance ratio.
This is interesting.According to this theory, 6600 8600 250 460 are certainly not mid-range.Or should that be,just because they're close to previous gen high end,so they are certainly mid-range.
Posted on Reply
#95
xfia
mxp02This is interesting.According to this theory, 6600 8600 250 460 are certainly not mid-range.Or should that be,just because they're close to previous gen high end,so they are certainly mid-range.
what exactly do you mean? I'm a gamer so I base my experience off of performance.. new architectures are great but how much is it really worth when the most they can boast is efficiency. spending millions on research and not increasing ipc is almost sad to me and borders pathetic if a new architecture sits on a smaller lithography.
this can give you a bit of a idea how much performance has really increased in recent generations
www.tomshardware.com/reviews/gaming-graphics-card-review,3107-7.html
www.tomshardware.com/reviews/gaming-cpu-review-overclock,3106-5.html
a lot of gamers have zero reason to make upgrades with years old hardware
Posted on Reply
#96
inferKNOX
༼ つ ◕_◕ ༽つ Gif Games, nVOLVO!!
Posted on Reply
#97
zAAm
Just semantics, but wouldn't a "perfectly functional GTX970" actually have 4 GB of full speed memory according to the definition given to consumers? Which means the returns are actually not perfectly functional?
Posted on Reply
#98
unspoken
Games are already hitting the 3.35 GB limit these cards provide. I would not have purchased two of these considering the low ram available specifically in SLI. In a single card setup, sure. But knowing this information would have saved me money and the trouble of building a high resolution setup.

Perfectly functional? Pfft. No. These cards are mislabeled. Why did tech sites not test the new Maxwell architecture? It seems they can be as unreliable as nVidia.
Posted on Reply
#99
inferKNOX
btarunrWe're all nerds on this forum, for knowing what a graphics card is. Else we'd be gaming on an Xbox.
These are my sentiments too. I take quite a bit of offense to his nerd-bashing!

Like Tomgang, I feel like this:


What I came to understand is that because it can only access one memory partition at a time, it starts alternating between the two partitions, slowing the whole card down. Basically both partitions become slow when the slower 2nd one starts being used, due to the alternating nature of the config!
That means the whole card suffers a performance drop when performance is needed most!:banghead:
Posted on Reply
#100
xfia
I think the HSA foundation needs to give NV a lesson on the parallel nature of gpu's.. The whole design is flawed by basic principle. gaming? yeah it can hurt there but what about editing, rendering and NV own Cuda apps... ouch...
Posted on Reply
Add your own comment
Nov 11th, 2026 17:23 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts