Nvidia’s GTX970 has a rather serious memory allocation bug

Cyberghost

Federal Agent Area 51
Staff member
NVidia’s GTX 970 is the current price-to-performance darling, offering incredible visual for incredible value. It seems, however, that it’s harbouring a dark secret. It’s a 4GB card, but it looks like a significant chunk of that VRAM doesn’t work.


According to a number of rather angry people on Reddit, Overlock and the Guru3D forums who’ve noticed their shiny, powerful GTX 970’s come to a screeching halt when maxing out their cards. Clever people, using VRAM benchmarking software, have discovered that when the last 500`700MB of VRAM gets accessed, memory performance drops significantly. Some users have even found that their cards go belly up when hitting 3GB.


Source : Lazygamer.net
 

sam_738844

Wise Old Owl
NVidia’s GTX 970 is the current price-to-performance darling, offering incredible visual for incredible value. It seems, however, that it’s harbouring a dark secret. It’s a 4GB card, but it looks like a significant chunk of that VRAM doesn’t work.


According to a number of rather angry people on Reddit, Overlock and the Guru3D forums who’ve noticed their shiny, powerful GTX 970’s come to a screeching halt when maxing out their cards. Clever people, using VRAM benchmarking software, have discovered that when the last 500`700MB of VRAM gets accessed, memory performance drops significantly. Some users have even found that their cards go belly up when hitting 3GB.


Source : Lazygamer.net

Get your facts straight.


I’m not dismissing the idea that there may be a small group of users out there for whom this RAM allocation is an actual problem, but in most cases, the gap betweeen 3.5GB and 4GB is nonexistent. The impact on the overwhelming majority of GPU users is going to be similarly small, and the GTX 970 remains an excellent value for almost everyone. Those few contemplating 4K SLI builds or something similar may want to more carefully examine the question and consider either a GTX 980 solution or a pair of AMD GPUs in Crossfire

--ET

For users that are attempting to measure the impact of this issue you should be aware that in some cases the software you are using report the in-use graphics memory could be wrong. Some applications are only aware of the first "pool" of memory and may only ever show up to 3.5GB in use for a game. Other applications, including MSI Afterburner as an example, do properly report total memory usage of up to 4GB. Because of the unique allocation of memory in the system, the OS and driver and monitoring application may not always be on the page. Many users, like bootski over at NeoGAF have done a job of compiling examples where the memory issue occurs, so look around for the right tools to use to test your own GTX 970. (Side note: we are going to try to do some of our own testing this afternoon.)

NVIDIA has come clean; all that remains is the response from consumers to take hold. For those of you that read this and remain affronted by NVIDIA calling the GeForce GTX 970 a 4GB card without equivocation: I get it. But I also respectfully disagree. Should NVIDIA have been more upfront about the changes this GPU brought compared to the GTX 980? Absolutely and emphatically. But does this change the stance or position of the GTX 970 in the world of discrete PC graphics? I don’t think it does.

--pcper

- - - Updated - - -

Oh well...AMD just needs to hammer it down now. :D

ya and they seem to be trying to do it with a cheap marketing hammer, disgraceful.

*wccftech.com/amd-offering-discounts-gtx-970-owners-memory-controversy/
 

Nerevarine

Incarnate
AMD Offering Discounts To Former GTX 970 Owners Following Memory Controversy

lol :lol:



*cdn4.wccftech.com/wp-content/uploads/2015/01/B8dV86eCYAEMLwx.png


BTW, AMD 3xx launches with HBM, AMD was the first to use GDDR5 in 2009 and its been 5 years .. Im sure HBM will be a significant bump to GPU technology.. nvidia on the otherhand did not vouch for HBM Tech early on, there fore will get to use it in GPUs starting next year
 
AMD Offering Discounts To Former GTX 970 Owners Following Memory Controversy

lol :lol:



*cdn4.wccftech.com/wp-content/uploads/2015/01/B8dV86eCYAEMLwx.png


BTW, AMD 3xx launches with HBM, AMD was the first to use GDDR5 in 2009 and its been 5 years .. Im sure HBM will be a significant bump to GPU technology.. nvidia on the otherhand did not vouch for HBM Tech early on, there fore will get to use it in GPUs starting next year

The very reason AMD lost GPU race was the fact that their drivers were sub-par(keyword:were, Omega release contains a (faint)fine line of hope). AMD always had upper hand in hardware...just the drivers....no comments.

What nvidia did is not justifiable, I like nvidia hardware(I use Quadro for both mobile and desktop), "die harvesting" concept is nothing new, but the problem remains with the fact that they deliberately disabled a slice of L2 cache. Now, this might not effect gaming grade consumer GPUs much but for a professional GPU, something like that will be a complete blunder. Memory usage is around 90% most of the time and the transfer is in between frame buffer and gpu core, not system memory; I wonder if something like that happened(disparity in specifications), I might as well buy a Firepro(even better, Xeon phi, atleast Intel is honest about specifications).

Edit: Looks like they are having PR problems with Gsync too, whereby GSync controller is not actually needed to implement the technology.
 
Last edited:

topgear

Super Moderator
Staff member
I think the bugged GTX 970s are going to cretae some massive issue later on for those who are going to maxing it out with later game titles. So though many did purchase GTX 970 I think it's better to stay away until nvdia releases a "fixed" GTX 970 or better 970 Ti, anyone ? ;)
 

sam_738844

Wise Old Owl
The very reason AMD lost GPU race was the fact that their drivers were sub-par(keyword:were, Omega release contains a (faint)fine line of hope). AMD always had upper hand in hardware...just the drivers....no comments.

What nvidia did is not justifiable, I like nvidia hardware(I use Quadro for both mobile and desktop), "die harvesting" concept is nothing new, but the problem remains with the fact that they deliberately disabled a slice of L2 cache. Now, this might not effect gaming grade consumer GPUs much but for a professional GPU, something like that will be a complete blunder. Memory usage is around 90% most of the time and the transfer is in between frame buffer and gpu core, not system memory; I wonder if something like that happened(disparity in specifications), I might as well buy a Firepro(even better, Xeon phi, atleast Intel is honest about specifications).

Edit: Looks like they are having PR problems with Gsync too, whereby GSync controller is not actually needed to implement the technology.

First off, drivers and hardware? You are going the opposite direction my friend, for the last one and half years there has been no driver issues from AMD side, contrary to common misconception they have improved remarkably in fixing micro-stutter, lag both in single gpu and Crossfire setups which sometimes have proved better in some cases than even nvidia. Now coming to hardware, I will simply skip the Maxwell innovation part because it got monster praise even from the hardcore AMD fanboys to say the very lest, let alone critics and reviewers, am talking about mad performance figures in way too less power, noise and heat profile! if that does not account for technological marvel, i dont know what is?

Now coming back to 970, what nvidia or their so called marketing team has put up is a stupid act, that's for sure, and its gonna cost them, but in real world performance figures it changes nothing, GTX 970 is still treading heavy blows with R9-290X in recent games and is mind-blowing VFM at its price tag.

synthetics benches which pushes this card above 3.5GB are lame in real-time FCAT and FPS runs "More" matters .Games now or in future that will require 4GB of "high-speed" VRAM or a 4K setup will dry out the GPU processing power before RAM. It makes no sense just becaue the access to 0.5GB of VRAM is slower than the other 3.5GB makes a card slower or "NOT have" 4GB of on board memory, that's a lie.

And buddy...read this up properly

There is one caveat as is always with these things. The module-less implementation was leaked by GameNab and consists of Drivers that appear to be buggy. The laptop monitor is being detected as G-Sync enabled and is clearly superior to a non-G-Sync alternative. It raises some very interesting questions about the G-Sync technology and this is where it gets interesting. Analysis done by PCPer reveals that you arent paying such a huge premium for nothing, there is a pretty noticeable difference between a dedicated module and this one.

isnt this the same thing that was debated a year ago when Gsync was moving up? there is always a software approach but it was not working as good as the HW approach, that was the very motion of this tech, just now undergoing another spin by some just-another-tweak-head who realized it could be half-done with drivers too! isnt that what Free-sync is? Which i am afraid have not seen the light of success!?
 
Last edited:
OP
Cyberghost

Cyberghost

Federal Agent Area 51
Staff member
I've seen Amazon US giving partial refunds to who buy GTX 970. Anyone who bought this card from Flipkart or other online shops in India got a partial refund??

Amazon: Partial Refund for GTX 970, kept the card
 

sam_738844

Wise Old Owl
gtx 970 buy or not?

buying a GTX 970 is a no brainer. If you are falling for the "memory not enough" FUD over the internet, then this card is not for you. But if your up for a nutella trip in 2560*1440 in recent titles , then shut up and give your money :D
 
A

amit.tiger12

Guest
buying a GTX 970 is a no brainer. If you are falling for the "memory not enough" FUD over the internet, then this card is not for you. But if your up for a nutella trip in 2560*1440 in recent titles , then shut up and give your money :D

in simple language please...
and I already decided to buy this.. but after I saw this post I ask question...
something comes in between your decisions..
 

sam_738844

Wise Old Owl
^^ Few weeks back, a rumour started to brew up that GTX 970 has a serious memory allocation problem which is causing high-res gaming come to a halt. The amount of VRAM it was supposed to have, which is 4GB, is found to be suspiciously segmented to two partitions, one 3.5GB high speed/priority access for draw commands, and a rather low speed 0.5GB segment.

The air lit up with slew of allegations from sites and reviewers that these caused games to stutter like hell when those tried to utilise all 4GB of VRAM which nvidia promised the card would have at the time people bought it. If you look it up in web, you would find that GTX 970 is only able to fully utilize 3.5GB out of the 4GB of available VRAM optimally. While the card can still access all 4GB of VRAM, the last 0.5GB of memory is not accessed or managed as efficiently as the rest of the available memory. Which leads to the reported performance degradation when an application needs to access more than 3.5GB. But the reports pour in from sites soon and nvidia came up with a elaborated explnation of why and how it could affect games.

In real-world benchmarks from games, most users are not affected by the segmentation, and the performance hit that could occur if the same card would be pitted against a full 4GB GTX 980, is in 1-3% of drop margin. Meaning, in 1920*1080 or 1440P, you would barely notice anything, except you want to crank up details to insane levels in 4K or surround gaming. GTX 970 is not a card which would prove very useful in 4K gaming anyway, because cards with more VRAM is suggested to even get a playable FPS at those resolution.

Funny thing is , nobody was noticing any hiccups before the whole drama went online, there are notoriously rigged examples of the card stuttering while requiring 4GB of VRAM, at resolutions with games that SHOULD require at least 6GB of VRAM to start with to get at least 40+FPS, and SLI does not even change that. The whole fiasco about less performance is more like joke now.

Nvidia has addressed this, came clean and revised the specs of GTX 970 now as having 56 ROPs, 1.75MB of L2 Cache, and a memory bandwidth divided into 196 GB/s (3.5 GB)||28 GB/s (512MB).

Now the point, its totally justified that nvidia took a lot of heat because of their purportedly false advertising about the specs of GTX 970, had they not cleared the air out of what the offerings are from GTX 970 right now to owners and future buyers, they were bound to look at long list of lawsuits ready to burn their a$$ down. But the performance, yes the performance figures are saying otherwise, despite the spec revision, true configuration, the numbers did not change with games. GTX 970 is still performing brilliantly in all titles in full fidelity. Thats what still makes the card best VFM.

John Peddy research reported that the return rates of GTX 970 has been less than 5% all over the globe to big players in AIB. There are polls going on everywhere in which it is shown that 64% majority of people are directly saying that "specs dont matter but perf does" , 15% want to sit and let the dust settle down, 5-10 % accused nvidia of lying still keeping the card or was not owner in the first place.... and the rest returned thier cards.

So now, its about time YOU ask yourself a question. BTW , i am an owner of GTX 970, I dint play any games in last 5.2 months which exceeded 3.5 GB of VRAM, so it didnt matter to me, you can call that bias, but i seriously doubt that would make 970 any less of a champion card.
 
A

amit.tiger12

Guest
^^ Few weeks back, a rumour started to brew up that GTX 970 has a serious memory allocation problem which is causing high-res gaming come to a halt. The amount of VRAM it was supposed to have, which is 4GB, is found to be suspiciously segmented to two partitions, one 3.5GB high speed/priority access for draw commands, and a rather low speed 0.5GB segment.

The air lit up with slew of allegations from sites and reviewers that these caused games to stutter like hell when those tried to utilise all 4GB of VRAM which nvidia promised the card would have at the time people bought it. If you look it up in web, you would find that GTX 970 is only able to fully utilize 3.5GB out of the 4GB of available VRAM optimally. While the card can still access all 4GB of VRAM, the last 0.5GB of memory is not accessed or managed as efficiently as the rest of the available memory. Which leads to the reported performance degradation when an application needs to access more than 3.5GB. But the reports pour in from sites soon and nvidia came up with a elaborated explnation of why and how it could affect games.

In real-world benchmarks from games, most users are not affected by the segmentation, and the performance hit that could occur if the same card would be pitted against a full 4GB GTX 980, is in 1-3% of drop margin. Meaning, in 1920*1080 or 1440P, you would barely notice anything, except you want to crank up details to insane levels in 4K or surround gaming. GTX 970 is not a card which would prove very useful in 4K gaming anyway, because cards with more VRAM is suggested to even get a playable FPS at those resolution.

Funny thing is , nobody was noticing any hiccups before the whole drama went online, there are notoriously rigged examples of the card stuttering while requiring 4GB of VRAM, at resolutions with games that SHOULD require at least 6GB of VRAM to start with to get at least 40+FPS, and SLI does not even change that. The whole fiasco about less performance is more like joke now.

Nvidia has addressed this, came clean and revised the specs of GTX 970 now as having 56 ROPs, 1.75MB of L2 Cache, and a memory bandwidth divided into 196 GB/s (3.5 GB)||28 GB/s (512MB).

Now the point, its totally justified that nvidia took a lot of heat because of their purportedly false advertising about the specs of GTX 970, had they not cleared the air out of what the offerings are from GTX 970 right now to owners and future buyers, they were bound to look at long list of lawsuits ready to burn their a$$ down. But the performance, yes the performance figures are saying otherwise, despite the spec revision, true configuration, the numbers did not change with games. GTX 970 is still performing brilliantly in all titles in full fidelity. Thats what still makes the card best VFM.

John Peddy research reported that the return rates of GTX 970 has been less than 5% all over the globe to big players in AIB. There are polls going on everywhere in which it is shown that 64% majority of people are directly saying that "specs dont matter but perf does" , 15% want to sit and let the dust settle down, 5-10 % accused nvidia of lying still keeping the card or was not owner in the first place.... and the rest returned thier cards.

So now, its about time YOU ask yourself a question. BTW , i am an owner of GTX 970, I dint play any games in last 5.2 months which exceeded 3.5 GB of VRAM, so it didnt matter to me, you can call that bias, but i seriously doubt that would make 970 any less of a champion card.

very big but nice explanation.. thanks a lot.. ;) :)
btw I will use 970 for FHD screen, so I don't need to worry about 4GB or 3.5GB, whatever it is.. right?
 

warfreak

Talk to the hand!!!
At 1080p, this memory allocation thing is hardly a problem.

Only if you want to game at 4k or if you have a highrez texture mod installed for your game, then maybe the card will be pushed to limit.

I hear Shadow of Mordor requires and Withcer 3 is gonna require 6 GB of VRAM (to play at ULTRA) but these are rare exceptions.

We are at a phase when we feel 1080P is more than enough and 4K is overkill but a few years down the line, I think this is going to change. This kind of thing has happened before, it will happen again. What is privilege today will become necessity tomorrow.
 

sam_738844

Wise Old Owl
very big but nice explanation.. thanks a lot.. ;) :)
btw I will use 970 for FHD screen, so I don't need to worry about 4GB or 3.5GB, whatever it is.. right?

no worries. You can even use DSR to play games at 2560*1440 and use MFAA with this card with no problem.
 
A

amit.tiger12

Guest
no worries. You can even use DSR to play games at 2560*1440 and use MFAA with this card with no problem.

ohhk thanks...

- - - Updated - - -

At 1080p, this memory allocation thing is hardly a problem.

Only if you want to game at 4k or if you have a highrez texture mod installed for your game, then maybe the card will be pushed to limit.

I hear Shadow of Mordor requires and Withcer 3 is gonna require 6 GB of VRAM (to play at ULTRA) but these are rare exceptions.

We are at a phase when we feel 1080P is more than enough and 4K is overkill but a few years down the line, I think this is going to change. This kind of thing has happened before, it will happen again. What is privilege today will become necessity tomorrow.

4k resolutions means 4 times HD...
Even FHD is not good at 32" size, for FHD 40" is good size.. As per size of monitor required for 4K, for so big size our eyes are not that "big" to cover that whole size from small distance..
I think main purpose of 4K is to provide better quality for long distance viewing..
 
Top Bottom