Slashdot Mirror


Nvidia Faces Suit Over GTX970 Performance Claims

According to this story at PC World, Nvidia was hit with a class action lawsuit Thursday that claims it misled customers about the capabilities of the GTX 970, which was released in September. Nvidia markets the chip as having 4GB of performance-boosting video RAM, but some users have complained the chip falters after using 3.5GB of that allocation. The lawsuit says the remaining half gigabyte runs 80 percent slower than it's supposed to. That can cause images to stutter on a high resolution screen and some games to perform poorly, the suit says. It was filed in the U.S. District Court for Northern California and names as defendants Nvidia and Giga-Byte Technology, which sells the GTX 970 in graphics cards. Nvidia declined to comment on the lawsuit Friday and Giga-Byte couldn't immediately be reached.

161 comments

  1. Seagate by Anonymous Coward · · Score: 3, Insightful

    This doesn't look very different from when Seagate was taken to court over mislabeling hard-drives sizes, using 1000000 bytes for a MB instead of the commonly used 1048576 bytes for a MB.

    I fully expect they will lose this, lose some PR metric, and start to implement the age old skill of asterisks on packages and adverts.

    1. Re:Seagate by TapeCutter · · Score: 3, Insightful

      Looks exactly the same to me, people wanting compensation for their idiotic complaint. Who would go to the trouble of starting a class action suit over this sort of trivia? Is it really disgruntled customers or is it a competitor playing a dirty PR game?

      --
      And did you exchange a walk on part in the war for a lead role in a cage? - Pink Floyd.
    2. Re:Seagate by 0123456 · · Score: 5, Insightful

      Who would go to the trouble of starting a class action suit over this sort of trivia?

      Lawyers?

      They'll make a few million, and all the GTX970 owners will get a $5 discount coupon off their next Nvidia card.

    3. Re:Seagate by Dutch+Gun · · Score: 4, Interesting

      nvidia did much more egregious deceptive marketing well over a decade ago in the GeForce4 MX, which didn't have programmable shaders like the GeForce3 series. I actually learned about it because the company where I worked was fooled into buying one of these for my computer, when the program I was using required one of the higher-spec devices. That misleading advertising actually did real financial damage, at least in my company's case, both in lost productivity and in replacing the hardware.

      That really bothered me, and I started buying ATI cards for a time after that, since I didn't want to reward a company for underhanded behavior. I only started buying nvidia again quite a few years after that since the ATI drivers were giving me more trouble than they were worth.

      This is not quite as bad, but it definitely smells a bit the same. A "mistake", really? I'm betting they wouldn't have made a "mistake" in underestimating their card's performance. Marketing at it's finest.

      --
      Irony: Agile development has too much intertia to be abandoned now.
    4. Re:Seagate by wiredlogic · · Score: 1

      Why would they lose? The cards have 4GiB of RAM. They never guaranteed that performance would be sustained after the 3.5GiB mark.

      --
      I am becoming gerund, destroyer of verbs.
    5. Re:Seagate by LostMyBeaver · · Score: 2, Informative

      More accurately, nVidia could probably make it clear that 3.5 gigs is better than 3 gigs and 512megs is more expensive to add than the extra gig. So, a 4 gig card with 3.5 gigs active is the best you can expect right now. So the user with the 4 gig card can still expect better performance than a user with a 3 gig card.

      Pretty sad

    6. Re:Seagate by LostMyBeaver · · Score: 3, Insightful

      I was thinking that the $5 discount coupon would only be given to people willing to spend 15 minutes filling out the forms to get it. The lawyers will collect the difference after the expiration.

      I think it's more likely that there's lawyers sitting around somewhere who are reading news rags and looking for reviews which out this type of stuff. They then initiate the class action and make noise on sites like Slashdot to get people to sign up in order to establish the requirements for it to be considered a class action.

      There are far too many people who would do something stupid like say "I clicked the link out of principle!"

    7. Re:Seagate by hairyfeet · · Score: 3, Insightful

      Good lord, what is with all the corporate ass kissing? Nvidia is gonna need extra wipes tonight to clean all the lip prints off their behinds!

      Look, for those that seem to have trouble understanding the concept of truth in advertising I'll spell it out, okay? They ADVERTISED a 4GB card but if you try to load a HD texture pack that goes over 3.5GB does it work? NO IT DOES NOT because the last 512Mb might as well be turbocache system RAM for all the good it will do ya! So it is really REALLY simple, they gimped the chip, the way they gimped the chip made it for all intents and purposes a 3.5GB card (since you will NEVER EVER EVER BE ABLE TO USE THE LAST 512MB because of how big of a speed difference there is, it is IMPOSSIBLE for you to use a large HD texture with the last 512MB running at just 20% speed, so its worthless) and yet they sold it as a 4GB card.

      The solution is obvious, anybody who bought the 4GB should be refunded the difference between a 3GB version of the card and a 4GB and all future cards should be sold as 3.5GB since that last 512MB will never be used by a single game EVAR as it'll tie such a fucking boat anchor to performance you might as well be running it on a 9800GT. If they would have sold it as a 3.5GB card? No problem, case closed. But people shelled out for these specifically so they could load all those HD texture packs which just aren't gonna run on these, you will have to treat it as a 3GB card because if you go 1MB into that last 512MB? Say goodbye to your framerate. I'm sorry but I agree with the plaintiffs, they didn't get what they paid for.

      --
      ACs don't waste your time replying, your posts are never seen by me.
    8. Re:Seagate by 0123456 · · Score: 0

      It does work. There's 4GB of RAM. It's fully usable. It's just slower. Just like my laptop.

      So are you saying I should be suing Toshiba because the top 2GB of RAM in my laptop is half the speed of the bottom 4GB?

      And note that the actual frame-rate tests I've seen show only a few percent difference, because you'll rarely be accessing data in that upper 512MB. You need artificial tests to make it run drastically slower.

    9. Re:Seagate by Anonymous Coward · · Score: 1

      I'm pretty sure you'd be looking at suing if you were refused your money back or a replacement that functioned as you expected on a car sold to you with 4 wheels where after you received it, one turned out to be 80% smaller than all the others.

    10. Re:Seagate by Solandri · · Score: 1

      I thought it was Maxtor which kicked off the "1 MB = 1 million bytes" thing? I vaguely recall Seagate being one of the stubborn holdouts for 1 MB = 2^20 bytes. I do know IBM was the last one to switch. (Seagate bought Maxtor in 2006, so it's somewhat a moot point.)

    11. Re:Seagate by Anonymous Coward · · Score: 0

      Huh? What laptop has half its memory run at one-fifth of the rest's speed?

    12. Re:Seagate by Anonymous Coward · · Score: 0

      There's 4GB of RAM. It's fully usable. It's just slower.

      This is bullshit, performance is part of usability.

    13. Re:Seagate by Anonymous Coward · · Score: 0

      I thought it was Maxtor which kicked off the "1 MB = 1 million bytes" thing? I vaguely recall Seagate being one of the stubborn holdouts for 1 MB = 2^20 bytes. I do know IBM was the last one to switch. (Seagate bought Maxtor in 2006, so it's somewhat a moot point.)

      Not exactly, it was a bunch of Metric System purists who started the whole thing. (The relevant wiki page is: http://en.wikipedia.org/wiki/Mebibyte)

      The short story is that Maxtor was the first to have a Marketing department take advantage of that, knowing full well most people see MB and assume powers of 2.

    14. Re:Seagate by Anonymous Coward · · Score: 0

      Memory size doesn't work in decimal, it uses binary address decoding logic to select a memory location. These are the fscking tracks on the boards, perhaps you need to learn the absolute basics before talking crap? It would be preposterous for memory units of measurement to move to decimal. This is nothing like spinning platter hard drives.

    15. Re:Seagate by drinkypoo · · Score: 3, Insightful

      nvidia did much more egregious deceptive marketing well over a decade ago in the GeForce4 MX,

      Really? I had one of those cards, what was wrong with it?

      I actually learned about it because the company where I worked was fooled into buying one of these for my computer, when the program I was using required one of the higher-spec devices.

      Fooled? How?

      That misleading advertising actually did real financial damage,

      No. The idiot chump who made the purchasing decision and bought the card even though it did not have programmable shaders and didn't say it had programmable shaders is the one who did the company real financial damage. The GF4MX never claimed to have programmable shaders. Are you that idiot? Or did the GF4MX just touch you somewhere? Can you show us on this picture of the internet where nVidia touched you?

      I owned a GF4MX because it was a fantastic value. As a budget gamer, it was the best buy at the time if you weren't playing games which demanded shaders. As a corporate user who needed shader support, you'd have to be an asshole to buy the GF4MX anyway, because it was a budget card. You wouldn't buy a card like that to do work, unless you just needed a basic GPU.

      This may shock you, but you are going to have to do your homework no matter what you buy. The model numbers on cars are equally worthless. Let me guess, your business also bought a one ton truck and discovered after ignoring the weight allowances that it couldn't actually haul one ton in the bed legally, and therefore Ford stole your money and kicked you in the nuts because you were too dumb to read and understand the specs, and that the GVWR would only allow you 3/4 of a ton of actual hauling capacity in your one ton truck.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    16. Re:Seagate by drinkypoo · · Score: 1

      So are you saying I should be suing Toshiba because the top 2GB of RAM in my laptop is half the speed of the bottom 4GB?

      If it's not because the memory is of two different types, yes. That's exactly what is being said. If the system says it supports up to however much memory, and the memory specs don't make it clear that over a certain point performance will degrade, then yes you should sue, if they won't take a return.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    17. Re:Seagate by FyRE666 · · Score: 2

      Has Slashdot somehow attracted edgy Youtube commenters now? It's so dull to read useless diatribes full of personal attacks and passive-aggressive dick waving. I'd hope we're a little more intelligent here. Why do people even post this garbage? Do they think it makes them "cool", or maybe that it'll impress someone? Maybe it's just a cry for attention from a lonely 12 year old I guess...

    18. Re:Seagate by OverlordQ · · Score: 2

      > The short story is that Maxtor was the first to have a Marketing department take advantage of that, knowing full well most people see MB and assume powers of 2.

      Then most people are stupid. Stop trying to bastardize the SI prefixes for your hard drive edge case, in every other measure Mega is a base 10 power, not base 2.

      --
      Your hair look like poop, Bob! - Wanker.
    19. Re:Seagate by greg1104 · · Score: 2

      No, that dude has been on Slashdot for longer than any 12 year old has been alive. You can tell from both his uid and the terrible car analogy.

    20. Re: Seagate by Anonymous Coward · · Score: 0

      As if /. is attracting anything at all these days... Maybe dust notes.

    21. Re: Seagate by jones_supa · · Score: 2

      Here's an old Geforce 4 discussion from 2001 in Slashdot, for those who want to compare.

    22. Re:Seagate by jones_supa · · Score: 1

      It does work. There's 4GB of RAM. It's fully usable. It's just slower.

      Imagine if you bought an 8-pack of cola cans and 1 of those cans was always a bit shittier tasting than the others. Would you accept the explanation "there's 8 cans of cola, they are fully drinkable".

    23. Re: Seagate by CronoCloud · · Score: 1

      Ah, beowulf cluster and CowboyNeal jokes...those were the days.

    24. Re:Seagate by Wootery · · Score: 1

      the company where I worked was fooled into buying one of these for my computer, when the program I was using required one of the higher-spec devices. That misleading advertising actually did real financial damage

      What exactly did they lie about? It sounds like you needed a card with support for new graphics APIs, and it turned out it didn't have that support, correct? Surely that's the fault of the person who placed the order.

      Creating a 'GeForce 4' that's actually less capable than a 'GeForce 3' is at worst a bit scummy (at worse.... it's not totally unreasonable to have a low-end GF4 be outperformed by a high-end GF3, imo). I wouldn't go as far as to say it was misleading advertising, unless they really lied on the check-boxes about supported graphics APIs.

    25. Re:Seagate by Anonymous Coward · · Score: 0

      I was fooled with the Geforce 440MX (specifically) as well. I believe they said it would have shaders before it came out. Where I read about it it said it would have 24 of the 26 shaders implemented in the Geforce 4 (from memory). Totally wrong, there were no shaders, all this misinformation just before it was released.

    26. Re: Seagate by Anonymous Coward · · Score: 0

      Oh, it'd be priceless to find AMD were somehow, indirectly of course, behind this...

    27. Re:Seagate by Anonymous Coward · · Score: 0

      There are subtle, but significant, differences.

      The SI prefix mega has had a well defined meaning since 1960. It's potentially misleading to advertise using the SI value rather than the industry specific 2^20, but it isn't technically wrong. These Nvidia cards are advertised as having 4 GB of memory. This forms the perfectly reasonable expectation that all 4 GB behave the same way.

    28. Re:Seagate by Spamalope · · Score: 2

      nvidia did much more egregious deceptive marketing well over a decade ago in the GeForce4 MX,

      Really? I had one of those cards, what was wrong with it?

      No. The idiot chump who made the purchasing decision and bought the card even though it did not have programmable shaders and didn't say it had programmable shaders is the one who did the company real financial damage. The GF4MX never claimed to have programmable shaders. Are you that idiot? Or did the GF4MX just touch you somewhere? Can you show us on this picture of the internet where nVidia touched you?

      The GeForce 3 was a new processor core supporting DX8. The Geforce 4 line was marketed as an updated GF3 core supporting DX9, but Nvidia sold warmed over Geforce 2 cores supporting only DX7 labeled as Geforce 4 without making the switch clear - in fact burying any disclosure of what that meant in terms of performance and compatibility.

      At launch at retail from what I remember, the posters and pamphlets touted the GF4 cores supporting DX9. The box for the GF4 MX cards didn't contain any information outside the shrink wrap to let a customer know it was really a GF2 with a higher clock. If you advertise 'Buy GF4 for Direct X 9 capability' then your GF4 parts need to deliver that or you're being deceptive. It's not acceptable to require consumers to learn processor architecture to know the marketing material aren't true, rather the marketing mustn't contain out right lies.

    29. Re: Seagate by drinkypoo · · Score: 1

      Ah, beowulf cluster and CowboyNeal jokes...those were the days.

      Natalie Portman's Hot Grits poured on the GNAA. Those were truly classier times.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    30. Re:Seagate by Anonymous Coward · · Score: 0

      The only reason people buy a card like this is performance.

    31. Re: Seagate by Anonymous Coward · · Score: 0

      Hah! Hah! And yes today ATI's drivers still SUCK. My how little has changed.

      Beyond that all I see are more reasonable comments.

      As to this, yes the lawyers initiating this suit are the only ones who stand a chance of making anything.

      Personally I'd think that the suit is meritless as the 970 does indeed have 4GB of VRAM. The only "error" being the 1.75 L2 cache v. 2. Is that even on the boxes? Looking at an R9 280X and my 780 Ti box that I have lying around pretty much the only real technical "spec" that I see is how GDDR5 that the cards have. The balance being Sapphire and eVGA marketing fluff which should give you a clue that any advertising is done by the OEMs as nVidia doesn't sell any cards.

      But I guess that the 970s weren't cheap enough for the whiners to begin with. I'm still at a loss as to how they managed to push the cards so hard that they found that in the first place when every single review had nothing that lead them to look into it. I suspect that AMD was involved here somewhere since they're also phailing at GPU design now as well as CPU design. (And their drivers STILL suck ass.)

      As to the GF4MX. Wasn't it pretty obvious if you did ANY research that it wasn't the same as a GF4? Or am I being too logical and reasonable again?

    32. Re: Seagate by Anonymous Coward · · Score: 0

      Naked and petrified.

    33. Re: Seagate by jones_supa · · Score: 1

      As to the GF4MX. Wasn't it pretty obvious if you did ANY research that it wasn't the same as a GF4? Or am I being too logical and reasonable again?

      Yep, I know. My main point was just to show some Slashdot discussion from the old times. I tried to find something close to the GF4MX and the discussion was what I found with a quick search.

    34. Re:Seagate by electrosoccertux · · Score: 1

      his posts are the disgusting type that make me feel negative about life and oh by the way hate slashdot.

      it's likely he's created multiple accounts in multiple browsers and cycles them so that he has mod points ready to go any time he wants them

    35. Re:Seagate by MikeBabcock · · Score: 1

      Except that MB is a metric term that was co-opted in computing for no good reason (base 2 calculation rounding), and correcting it makes sense.

      1 million bits is a megabit. 1 million bytes is a megabyte.

      --
      - Michael T. Babcock (Yes, I blog)
    36. Re:Seagate by Anonymous Coward · · Score: 0

      They ADVERTISED a 4GB card but if you try to load a HD texture pack that goes over 3.5GB does it work?

      Of course it does.

      So it is really REALLY simple, they gimped the chip, the way they gimped the chip made it for all intents and purposes a 3.5GB card

      Except it has 4GB of memory, not 3.5 and as such that extra 512MB of RAM is usable and is much faster than paging out to system memory.

      it is IMPOSSIBLE for you to use a large HD texture with the last 512MB running at just 20% speed, so its worthless

      Dont be so melodramatic, it isn't impossible at all, it isnt even unplayable.

      and all future cards should be sold as 3.5GB since that last 512MB will never be used by a single game EVAR as it'll tie such a fucking boat anchor to performance you might as well be running it on a 9800GT.

      Wrong. That additional 512MB of RAM makes it a lot faster than just having 3.5GB of video memory and paging out to system memory when you go over.

      But people shelled out for these specifically so they could load all those HD texture packs which just aren't gonna run on these

      Which HD texture packs arent going to run on these?

      you will have to treat it as a 3GB card because if you go 1MB into that last 512MB? Say goodbye to your framerate.

      Umm...how do you think that works with every other card? Is every 3GB card treated as a 2.5GB card for the same reason? In fact using that extra 512MB is faster than having a 3.5GB card in the same situation because the latter would use system RAM for the extra instead.

    37. Re:Seagate by Anonymous Coward · · Score: 0

      Ok I think you are being a bit dramatic there, not much of what you have actually stated there is true (particularly use of the word 'impossible').

      Now how this architecture actually works is that the memory interface on the 980 has 8 ports and the memory accesses are distributed between them with 0.5GB each and you have 4GB of memory to access. Now on the 970 it only has 7 ports so immediately you can see a problem with distribution here. So what has been done is that the 7th port provides access to the last 1GB which of course creates a bottleneck there, if you hit it with a properly crafted synthetic benchmark you would see accesses to that last 0.5GB at 1/7 the speed of the rest of the memory (you can see this on Nai benchmark if you google for it).

      For a simplistic view normally the driver is given 2 pools of memory to manage with, system memory and graphics memory. In the case of the 970 the graphics memory is split into a 3.5GB and also a 0.5GB pool and the driver is smart enough to know (in this case just as in the case where there is only 2GB or 3GB or 6GB) that even if a game allocates a chunk of memory (be that more or less than what is physically available) it may not necessarily use that memory all at once which is why you can allocate more than is available and it won't reduce your performance until you actually use it.

      So now the question about running over that 3.5GB and into the 3.5-4.0GB range, well in this case you have a 512MB block that is not as fast as the first 3.5GB but absolutely much faster, about 4 times, than accessing the system memory over the PCIe bus. So no, you are absolutely not going to be better off with a 9800GT.

      So yes the marketing is causing some problems here and it should have been sold as a 3.5GB card with 0.5GB chunk of fast cache. But the issues arent as dramatic as you make out and in the real world you are highly unlikely to be balancing the load precariously between 3.5GB and 4GB for optimal performance on a 4GB card anyway. Saying it is "it is IMPOSSIBLE for you to use a large HD texture" and "you will NEVER EVER EVER BE ABLE TO USE THE LAST 512MB" is just rubbish, I suggest you read up on the architecture so you understand what you are talking about before making silly remarks like that.

      NB: I'm not defending it, it shouldn't have been sold flat out as a 4GB card. Just debunking some rubbish assertions.

    38. Re:Seagate by hairyfeet · · Score: 1

      Bullshit you CANNOT USE THE LAST 512MB, reviewers have to use CLI hacks just to run a bench with the last 512MB, want proof? Here ya go, he loads 4k textures and HD texture packs and NEVER is able to hit above 3.5GB, the games WILL NOT USE THE 512MB because of the performance hit, its seen as system RAM and therefore WORTHLESS.

      --
      ACs don't waste your time replying, your posts are never seen by me.
    39. Re:Seagate by Neil+Boekend · · Score: 1

      Why was Seagate sued for using the correct amount? The SI prefixes are older then the use of them in computing. "Mega" was standardized in 1960.
      Now as to why they were abused to mean powers of 2: I can understand that. However that abuse does not mean you should sue the guys that get it right.

      --
      Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
    40. Re:Seagate by Anonymous Coward · · Score: 0

      Not really. All the ram is there. The only difference is that architecture doesn't have the same number of interconnects between that last 500mb and the rest of the hardware. Synthetic benchmarks testing explicit memory ranges show horrible performance, real world applications will never be so specific.

    41. Re:Seagate by Agripa · · Score: 1

      Then most people are stupid. Stop trying to bastardize the SI prefixes for your hard drive edge case, in every other measure Mega is a base 10 power, not base 2.

      Bit and byte are not SI units. Let me know when I can buy computers with 17.179869184 GBytes of RAM.

      Take up your argument with JEDEC.

    42. Re:Seagate by Agripa · · Score: 1

      1 million bits is a megabit. 1 million bytes is a megabyte.

      So systems advertised with 17.179869184 GBytes of RAM should be appearing any time now. That will be a lot simpler than labeling them 16 GBytes and no doubt appeal to marketing.

    43. Re:Seagate by MikeBabcock · · Score: 1

      Actually they have 16 GiB of RAM; its a number very close to a billion but based on powers of 2 instead of 10.

      http://en.wikipedia.org/wiki/G...

      Words have meanings, and so do prefixes; the metric numbering system was usurped (stupidly) to mean something it didn't mean.

      --
      - Michael T. Babcock (Yes, I blog)
    44. Re:Seagate by OverlordQ · · Score: 1

      Bzzzzt, try again, IEC 80000-13.

      --
      Your hair look like poop, Bob! - Wanker.
    45. Re:Seagate by Agripa · · Score: 1

      IEC is not JEDEC and JEDEC does not explicitly support the IEC definitions.

      When they start using metric prefixes in IC part numbers and organization, then I will reconsider.

    46. Re:Seagate by Agripa · · Score: 1

      I am aware of the IEC's revision and believe it will lead to further confusion. Simply regarding bits, bytes, words, and whatnot as non-SI units with special meanings for k, M, and G would have been enough. It only got confusing with telecommunication standards requiring conversion.

  2. 80% Slower? by Anonymous Coward · · Score: 0

    What does that mean? It runs at 20% of the speed of the other 3.5GB?

    If it said, "the remaining half gigabyte runs at 80 percent of the speed it's supposed to" that would make perfect sense, but "80% slower" is very confusing and its meaning is open to interpretation.

    1. Re:80% Slower? by sjames · · Score: 2

      It's only confusing if you bend over backwards to make it confusing. Given speed X, the last 0.5 Gig runs at X - (X*.8) That is, speed minus 80% of speed or as one might say in English, 80% slower.

      So yes, in this case it runs at 20% of the speed of the other 3.5 GB.

      What other interpretation did you have in mind?

    2. Re:80% Slower? by Anonymous Coward · · Score: 1

      The last 0.5 GB is only connected by one memory lane, while the first 3.5 GB has many more of them.

    3. Re:80% Slower? by craigm4980 · · Score: 4, Informative

      80% slower = 20% of the speed = 5 times slower (takes 5 times as long to do something) = 500% slower.

      OR

      80% slower = takes 80% longer to do something = 1/1.8 times slower = 0.55555555555 times slower = 44% slower.

      See, ambiguous. There are 2 well defined definitions of 80% slower: gets 80% less done per unit time, OR takes 80% longer to get the same thing done. They have very different meanings. For one of them, 100% slower = gets nothing done ever, and the other means it takes takes twice as long (you have to wait an extra 100% of the original time). The same applies to 100% faster: one version means infinatly fast, and the other means twice as fast.

    4. Re:80% Slower? by sjames · · Score: 2

      You're playing fast and loose with the terminology and adding extra steps that were never suggested to confuse yourself. KISS and it will all make sense.

    5. Re:80% Slower? by Anonymous Coward · · Score: 0

      GP is trying to keep trolls like you from turning this the grammatical swamp like we have for the word "biweekly."

      Everyone knows what the word biweekly means, right? What happens if someone promises to pay you $X biweekly?

      Just ask Fred. He'll say it means every 2 weeks, and he can prove it by pointing to the dictionary.
      Just ask Barney. He'll say it means twice per week, and he can prove it by pointing to the same dictionary.

      Both are right, but both are confusing as hell. People like you are the reason we have useless words like "biweekly."

      Conclusion: Stop saying "x% slower." Fred will think you mean one thing, and Barney will think you mean another.

    6. Re:80% Slower? by sjames · · Score: 1

      I'm not the one who used the phrase in the first place. I'm just the one who somehow managed to understand a perfectly common phrase that really doesn't have a lot of ambiguity to it.

      Perhaps if more people would take a moment to explain 'bi-weekly' (fortnightly), there wouldn't be so much confusion there either.

    7. Re:80% Slower? by Anonymous Coward · · Score: 0

      80% slower = 20% of the speed = 5 times slower (takes 5 times as long to do something) = 500% slower.

      "5 times slower" does not mean something takes 5 times as long to do.
      Y is 5 times (X) slower than X. Y is -4 times as fast as X.

      Slower is a relative term. When you say Y is slower than X, "slower" is relative to X. "5 times slower" means "5 times X slower".

    8. Re:80% Slower? by greg1104 · · Score: 1

      There are plenty of examples of both usages of slower, showing that it is ambiguous across popular usage. For example, there's a whole 800% slower "genre" on Youtube where songs are played at 1/8 of their original speed. When popular usage is so sloppy, it's better to avoid the phrasing altogether in favor of something precise instead.

    9. Re:80% Slower? by Anonymous Coward · · Score: 0

      wait, what? since when did "biweekly" mean the same thing as "semiweekly"? is this a recent relaxation of the definition, or have i just completely in the dark on this confusing state for 30+ years...?

    10. Re:80% Slower? by danomac · · Score: 1

      Maybe he works for Nvidia and that's how this mess got started?

    11. Re:80% Slower? by sjames · · Score: 1

      So we take X - 800%X = -7X, hmmm I say to myself, spinning it backwards at 7x normal speed seems unlikely. They must not be using percentages right.

      Note how the absurdity of the statement when interpreted the standard way tips me off that something is wrong. Doesn't that make more sense than pretending to be baffled when someone uses it correctly?

    12. Re:80% Slower? by sjames · · Score: 1

      You may be on to something here!

    13. Re: 80% Slower? by Anonymous Coward · · Score: 0

      No all 4GB is attache to the chip. Only 7 of 8 Caches are enabled so the last 512MB is divided between all 7 caches so the backend (aka ROP) or the texture unit can't access it efficently as the first 3.5G

    14. Re:80% Slower? by MikeBabcock · · Score: 1

      No, its an ambiguous word and I've complained about it in life on numerous occasions.

      --
      - Michael T. Babcock (Yes, I blog)
  3. It's worse than just 0.5 GB of slow memory by Anonymous Coward · · Score: 5, Insightful

    Memory performance between the two segments (3.5 + 0.5 GB) of memory works in an XOR manner so that accessing the slow segment prevents access to the 3.5 GB segment. Also, the whole memory access issue is a distraction from the fact that Nvidia originally advertised that the 970 had 64 ROPs (when it really has 56) and that it has 2 MB of L2 cache (when it really has 1.75 MB).

    1. Re:It's worse than just 0.5 GB of slow memory by Anonymous Coward · · Score: 0

      muh 256kb of cache, I'm so outraged!

    2. Re:It's worse than just 0.5 GB of slow memory by Anonymous Coward · · Score: 0

      you're missing the point. It looks like it has only 7/8 of the promised processor, but the full memory, but the 7/8 processor can only use 7/8 of the memory

    3. Re:It's worse than just 0.5 GB of slow memory by aliquis · · Score: 1

      I'm not a GPU designer.

      But in the case of the ROP ..
      (should I Google it? ROP = Raster Operator and here comes the URL: http://techreport.com/review/2...) .. supposedly doesn't matter (at all? Or for the number of pixels only?) because as the URL above says it:
      "In an even crazier reality, that limit isn't even the primary fill rate constraint in this product, since the GTX 970's shader arrays can only send 52 pixels per clock onto the crossbar."

      If it's the case that 56 is more than what is needed due to limits in other places on the card then whatever. I guess it's even fair to round of 1.75 MB to 2 MB. Heck, even 3.5 to 4 GB =P

      As for how the slow access of the last 0.5 GB affect the other 3.5 GB I don't know and I won't look it up for this post. But sure, that's a real problem. And they should post the correct information and I guess being excessive on details is a good thing.

      In the end it's still the card I would had bought.

    4. Re:It's worse than just 0.5 GB of slow memory by ChumpusRex2003 · · Score: 1

      It has 7/8 of the processor. In traditional GPU designs, this would mean that only 7/8 of the RAM is addressable and that the RAM data bus is only 7/8 as wide - as parallel modules of both stream processor, L2 cache and memory controller get disabled.

      The 1st 3.5 GB operates on the 7/8 wide bus (224 bits); this is the same as on any other reduced core count card. However, nvidia added a new "bypass" bus inside the GPU permit 2 memory controllers to be driven by a single L2 cache module. This avoids the need to reduced addressable RAM in the event of a L2 cache module being disabled. This last memory controller without a dedicated L2 cache module then accesses the top 0.5 GB of RAM via a 1/8 wide (32 bit) bus.

    5. Re:It's worse than just 0.5 GB of slow memory by MikeBabcock · · Score: 1

      One wonders if disabling that last 500MB of RAM would in fact improve performance.

      --
      - Michael T. Babcock (Yes, I blog)
    6. Re:It's worse than just 0.5 GB of slow memory by Anonymous Coward · · Score: 0

      You're the idiot who makes ripoffs like this possible.

    7. Re:It's worse than just 0.5 GB of slow memory by Anonymous Coward · · Score: 0

      Short answer is YES. The card itself reports 4Gb of VRAM, so if a game tries to allocate as much as it can of it, it will result in unplayable throttlefest because
      a) The las .5 Gb of ram runs 8 times slower of the first 3.5 Gb
      b) You can access only one partition of memory at the time, meaning that stuff in the fastest portion has to wait until you finish accessing stuff in the last .5 Gb

      Thats why the driver tries as much as it can to avoid allocating memory in the last segment.

    8. Re:It's worse than just 0.5 GB of slow memory by g0bshiTe · · Score: 1

      I have the card and actually did run into issues from specifying my game to use the full 4gb for textures. Ended up having to back it down to 3gb. No significant loss in performance maybe -10 fps, since it's still pegging at +60 I'm happy with the purchase.

      I purchased the card based on performance reviews and the price point of the 980. The 970 seemed to offer performance my wallet could justify spending the money on vs not dropping the cash for the latest and greatest card.

      --
      I am Bennett Haselton! I am Bennett Haselton!
  4. So... by pruedz · · Score: 1, Troll

    First World Problems.

    1. Re:So... by Anonymous Coward · · Score: 3, Insightful

      What part of businesses cheating customers is a first world problem?

      Or is that shit okay elsewhere?

    2. Re:So... by Anonymous Coward · · Score: 0

      Cheating the customers is primarily a third world problem. In first world there are mechanisms like customer protection legislation and functional court systems that specifically protect customers from this practice.

      In third world, one or both are typically lacking and cheating customers is commonplace.

      So this is in fact a third world problem.

    3. Re:So... by Anonymous Coward · · Score: 0

      Cheating the customers is primarily a third world problem. In first world there are mechanisms like customer protection legislation and functional court systems that specifically protect customers from this practice.

      At least in theory. In practice, "cheating" on the customers is a fairly common practice in consumer electronics (such as audio, digital cameras, etc.). Optimistic or misleading specs can be published without much risk of legal consequences, since the customers do not have the expertise or equipment needed to find out about it. How many people complain because their cheap 12 MP camera actually only uses 10 MP of image information from the sensor to create a 4000x3000 JPEG file with interpolation ? The wording of the specs can also be made vague enough not to actually guarantee anything (Nvidia could simply have advertised 4 GB of VRAM at up to 224 GB/s), or important details can be hidden in fine print 99% of the customers never read (an ISP may advertise "unlimited" traffic and hide the actual quota of 100 GB/month or whatever somewhere in a 500 pages long document). Finally, an increasingly popular approach is not to publish much usable technical specification at all, but mostly just useless marketing fluff.

  5. probably won't go anywhere by Anonymous Coward · · Score: 3, Interesting

    I think they will find it real difficult to prove that Nvidia was intentionally (Ntentionally?) misleading people with the advertisements. Nvidia's response and explanation for what happened seemed pretty detailed and made sense to me. These kind of lawsuits actually piss me off a little anyway, because the lawyers are the only ones who really benefit. Even if Nvidia is made to compensate people who purchased the card, it's unlikely going to amount to any more than a few dollars for each person. Except the lawyers, who will get their huge fees regardless.

    Do the people who file these things actually think they are somehow taking the company to task or making the world a better place? They certainly can't be doing it for actual money or greed.

    1. Re:probably won't go anywhere by sjames · · Score: 2

      I don't get your problem with the people suing. They were promised X for Y dollars. They only got a fraction of X. That Nvidia didn't do that intentionally means it wasn't fraud (and that's good). That still leaves them owing their customers 100% of X or compensation for only giving them a fraction of X.

      Specs like that are generally not interpreted as being on a best effort basis.

      As for the rest, yes it is an injustice that the lawyers will get millions and the actual plaintiffs will get nearly worthless coupons.

    2. Re:probably won't go anywhere by 0123456 · · Score: 3, Insightful

      I don't get your problem with the people suing. They were promised X for Y dollars. They only got a fraction of X.

      They were promised 4GB. They got 4GB. Complaining that 0.5GB of that runs slow is like complaining that 2GB of the 6GB in my laptop runs at half the speed of the other 4GB.

      The only valid complaint they have is that Nvidia said it had 64 ROPs when it only has 56. That's not really something worth a lawsuit, when, as I understand it, the chip isn't capable of generating pixels fast enough to make use of 64.

      As I said above, the only people who will really benefit from this are lawyers.

    3. Re:probably won't go anywhere by sjames · · Score: 4, Informative

      They were also promised a specific performance at the same time as the 4GB of memory. They aren't getting it. At the least, Nvidia should have offered a driver that avoids the last half Gig of ram and a partial refund.

      This isn't a car where it is well understood that top speed and maximum fuel efficiency don't happen at the same time. This is an unusual situation for a graphics card that substantially degrades it's performance, and so, it's value.

    4. Re:probably won't go anywhere by Kjella · · Score: 0

      This isn't a car where it is well understood that top speed and maximum fuel efficiency don't (sic) happen at the same time. This is an unusual situation for a graphics card that substantially degrades it's (sic) performance, and so, it's (sic) value.

      Legally, that might make all the difference in the world as a false claim by nVidia would be entirely different from a false assumption by the consumer. For example, for the CPU it's common that I have more RAM than I can access at any one time at top speed. Just because that's been commonly the case with GPUs - though in the aftermath of this several instances of non-uniform memory has been proven - it doesn't make it a defect that it's not. Unless you have some kind of generic fitness for purpose clause, but I don't think it fails that.

      --
      Live today, because you never know what tomorrow brings
    5. Re:probably won't go anywhere by ericloewe · · Score: 1

      They were not promised any performance numbers. They were promised something below the 980... which it is.

      Anything else comes from reviews - including thorough performance testing. The card didn't magically become slower than it was for those reviews.

    6. Re:probably won't go anywhere by sjames · · Score: 2

      Nvidia has already admitted that it's own people were confused by the claims. That says a lot about who the courts might blame.

      As for computers in general, I find that when I do access all of the memory (HPC), those accesses happen at the rated speed.

      As for the rest, an anti-emetic might keep you from being shoved in a locker.

    7. Re:probably won't go anywhere by sjames · · Score: 1

      You might not realize this but most of those review numbers came right from Nvidia's marketing department.

    8. Re:probably won't go anywhere by Anonymous Coward · · Score: 1

      Correct they were promised 4GB. They were also promised 224 GB/sec memory bandiwdth (not only on the first 3.5GB). Do you really want to allow companies to misrepresent their products?
      http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-970/specifications

    9. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      You keep talking about partial ram being less fast in your machine. What the fuck are you talking about?

    10. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      So if you bought a 4-core CPU and one of the cores was crippled to the point where you would be better off not using it, would you be okay with that since you technically got "4 cores"?

      The majority of the world doesn't appreciate it when someone uses weasel word tactics like that to worm their way out of doing the right thing.

    11. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      Correct they were promised 4GB. They were also promised 224 GB/sec memory bandiwdth (not only on the first 3.5GB).
      http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-970/specifications

      Technically the memory isn't the problem, it's all the same RAM running at the same clock speed. It's how it's integrated onto the bus that causes the bottleneck, so if you want to get super nitpicky about the details.... they didn't technically lie in their specs and nowhere did they state that you can achieve 224 GB/sec while using the entire 4Gigs of RAM. This is not at all the first product I've seen with stated specs that do not hold up well at all once you actually examine the hardware running in a live environment.
      Maybe the plaintiffs have a case, maybe they don't, it will be interesting to see what comes out of this in either event. Personally I think it's highly unlikely this was accidental, I'm pretty sure the marketing department was told to put some lipstick on the Pig.

      eally want to allow companies to misrepresent their products?

      Of course not, quit being a dick. But I also think it's easy to take things too far, to the point where companies start to simply NOT provide the technical details at all, and just rely on vague marketing promises. Most hardware actually doesn't get that detailed about specs anyhow. Think about your mouse, there's almost no technical information in the marketing material for it. Same with your consumer grade router, your keyboard, hell even your TV and monitor. Unless you're building your own computers, chances are there's more useful technical specs on the GPU than the rest of the system combined. So be careful what you wish for.

    12. Re: probably won't go anywhere by Anonymous Coward · · Score: 0

      What you are describing is the Arm big little setup. I don't think in that case that people care.

    13. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      So i guess you wont complain the next time they do this and its 2.0gb that is slow instead of 512mb?
      You still got 4gb, so what are you complaining about? Yeah, i thought so.

    14. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      This is an unusual situation for a graphics card that substantially degrades it's performance, and so, it's value.

      It actually only substantially degrades the performance of the card under certain circumstances (such as running specific games at extreme settings). That is why it took months to be discovered. Performance guarantees for worst case conditions are not exactaly a standard practice for consumer products that are mainly intended to provide good value for the money.

    15. Re:probably won't go anywhere by drinkypoo · · Score: 3, Insightful

      For example, for the CPU it's common that I have more RAM than I can access at any one time at top speed.

      No, no it isn't, and it hasn't been since the Amiga.

      There is no PC where it is common to have different speeds of memory.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    16. Re:probably won't go anywhere by sjames · · Score: 1

      However, it happens under a condition that is within the spec of the card. Had the card been specced as 3.5 GB and it just happened to be able to go to 4GB (outside of spec), all bets about performance would reasonably be off.

      If specifications are to mean something they must be met. They COULD have qualified the specs by saying 4 GB max ram and then explicitly state that the performance figures happen at 3.5 GB and they would be fine.

      Allowing the fast and loose approach distorts the market and shafts the best in the market. Consider, what if another card was out there that cost $5 more but didn't have the massive fall-off when accessing the last 600M of it's RAM.

      Functional healthy markets cannot coexist with deceptive specifications (even unintentionally deceptive).

    17. Re:probably won't go anywhere by ericloewe · · Score: 1

      Are you claiming not a single site ran their typical suite of benchmarks, instead getting their data from Nvidia?

      Because that is the most asinine conspiracy theory of the month. Sure, there are lazy sites that don't actually review stuff, but that doesn't apply to all.

    18. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      The burden of proof is pretty low in a lawsuit. Why would Nivida have designed a card like this when a card with 3.5GB would have featured a more elegant design, if not to trick people into believing it was closer in specification to their 980 cards?

    19. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      Seems to me that, technically, this is very similar to having a 256MB DIMM and a 128MB DIMM put into a system with 2 memory channels, but configured to operate as a single wider channel. Not sure if systems still do that (because multiple independent channels is going to be faster for most systems), but it was certainly done in the past. Why? Because it provided double the bandwidth for certain usage applications. And when so configured, accessing the "top" 128MB was going to be with half the bandwidth, since it's only attached to one channel.

      Change the numbers to 7 channels, logical and 8 physical, and the memory per physical channel to 512MB, configured to operate the 7 logical channels as a single wider channel (GPU is all about the bandwidth), and that's what you have here.

      Systems came like this all the time, not so long ago...

    20. Re:probably won't go anywhere by sjames · · Score: 1

      They certainly didn't do testing before they got a test unit. Then, once they did, I'm guessing they ran a very standard test suite that probably doesn't try to stress memory capacity.

      I'm quite sure the numbers in Nvidia ads and on the box didn't come from review site tests.

      And yes, I'm sure some sites actually tested, but I'll bet more than we would like just grabbed the numbers from Nvidia.

    21. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      Maybe someday you'll learn about registers, L1, L2, L3 caches, bus expansion memory, virtual memory, etc.

    22. Re:probably won't go anywhere by ericloewe · · Score: 1

      No graphics card has "XX FPS minimum in GameY" on the box or in any ad. Only vague promises and concrete data like clocks and RAM.

    23. Re:probably won't go anywhere by sjames · · Score: 1

      But in this case, the clocks and RAM data is wrong.

    24. Re:probably won't go anywhere by meerling · · Score: 1

      Telling marketing to "put some lipstick on the Pig" is kind of pointless as they've already done that, and body shimmer, and stuffed it into a slinky red mini-dress.
      (Don't ever ask anyone in tech support what they honestly think about the marketing people.)

    25. Re:probably won't go anywhere by meerling · · Score: 1

      From a review I saw a while ago, it's still faster than if you didn't have the last .5mb since otherwise it would have to pull it off the system memory and that's even slower.

    26. Re:probably won't go anywhere by meerling · · Score: 1

      sorry, .5gb, got distracted and typed the wrong thing.

    27. Re:probably won't go anywhere by drinkypoo · · Score: 1

      Seems to me that, technically, this is very similar to having a 256MB DIMM and a 128MB DIMM put into a system with 2 memory channels

      Is this a troll? A 128MB DIMM, all else equal, has the same number of chips as a 256MB DIMM. Each chip has half the capacity, so for the same bus width, you get half the memory. Of course, there are DIMMs which only have a couple of chips on them, but those chips have wider buses than the normal ones. You can install just one DIMM at a time into modern PCs, because a DIMM is two SIMMs on one board. And also a stupid name, because a SIMM was a single module, not a single memory chip, and a DIMM is also a single module, but the name is dual ... module. How idiotic.

      Dual-channel memory controllers let you install four DIMMs. Or, you could install three. But usually, memory controllers won't do that. If you install two dimms in the proper slots you get double bandwidth, more or less. If you install three, you get single bandwidth. Four gets you double again. Maybe that's changed but last I checked it was true.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    28. Re:probably won't go anywhere by drinkypoo · · Score: 2

      Maybe someday you'll learn about registers, L1, L2, L3 caches, bus expansion memory, virtual memory, etc.

      I owned an IBM PC-1 with 384kB of RAM on an 8-bit ISA card along with a RTC and maybe a serial port or something, so I'm quite familiar with the idea. But that was a long, long time ago. Maybe someday you'll learn that when people say "memory", by default they mean main memory.

      I also owned several Amigas, and still have an A1200 in fact. Now THAT is an architecture where it was normal to have memory running at different speeds. Hell, you could actually put SRAM on Amigas.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    29. Re:probably won't go anywhere by ericloewe · · Score: 1

      Clocks are not wrong and the whole 4GB of RAM are accessible.

      If you want to complain about something, at least find something that was really misrepresented.

    30. Re:probably won't go anywhere by sjames · · Score: 1

      You mean like that the last 0.5 GB runs slower (you know, the clocks is wrong) and creates bus contention? Or how it can only do 7 accesses in parallel, not 8? Even Nvidia admits they gave out incorrect information, how is it you are so sure they didn't?

    31. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      Car analogy time!
       
      This is like your car having a 2l V6 engine and six speed gearbox, but when in sixth gear doing over 3500rpm (The very top-end speed of your car), the engine management system cuts fuel input to five of the cylinders. In every day use you wouldn't notice; Your car will hit and maintain highway speeds without any issue. You take your car to a track day, and on the straight you flick the car into 6th, and you car immediately slumps with the engine displacement of a soda can and the torque output of an arthritic hedgehog.
       
      Tell me that you wouldn't flip your shit at the manufacturer.

    32. Re:probably won't go anywhere by Anonymous Coward · · Score: 0

      pedantic> So, you're saying that all the addressable memory space in my PC does, in fact, have the same clock speed? So the "RAM" of my sound card, video card, network card, and system RAM all have the same clock speed? /pedantic>

      Yes, I know originally it was just RAM... however modern PC architectures lump all addressable hardware memory into one big pool (disregarding paging and other OS optimizations per piece of hardware...).

    33. Re:probably won't go anywhere by rdnetto · · Score: 1

      For example, for the CPU it's common that I have more RAM than I can access at any one time at top speed.

      No, no it isn't, and it hasn't been since the Amiga.

      There is no PC where it is common to have different speeds of memory.

      Maybe not on desktops, but it definitely exists in servers, where it's referred to as NUMA.

      --
      Most human behaviour can be explained in terms of identity.
    34. Re:probably won't go anywhere by MachineShedFred · · Score: 1

      The other effect is that Nvidia gets a nice fat punitive settlement levied against them, which acts as a deterrent from doing this crap again. I think that's the point.

      --
      Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
    35. Re:probably won't go anywhere by g0bshiTe · · Score: 1

      I bought a 970 and it was based off site benches, it seemed for the price it would suit my needs. It fully does, I did run into issue with one game where on the command line I was forcing launch using all 4gb of ram on the card, it didn't work out as well. I ended up backing off until I found the magic number and all works well.

      In the end I have a card I thought I paid a decent amount for, for the performance I receive. Even after all this has come to light I'm still happy with my purchase.

      --
      I am Bennett Haselton! I am Bennett Haselton!
    36. Re:probably won't go anywhere by MachineShedFred · · Score: 1

      No, just the throughput which is a function of clock * bus width.

      You can have all the clock rate in the world, but still get shit performance on a 32-bit bus, which is exactly the problem.

      --
      Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
    37. Re:probably won't go anywhere by MachineShedFred · · Score: 1

      it's the same clock. The problem is bus width. 7/8ths of the memory is on an 7x wider bus than that last 1/8th. Thus, the performance loss.

      --
      Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
    38. Re:probably won't go anywhere by MachineShedFred · · Score: 1

      What's really funny, is that if they would have marketed it as a 3.5GB card, and then people started to figure out that there was really 4GB of RAM on there, they probably would have gotten better sales of the part through the hardware hack community trying to figure out how to "unlock" it.

      --
      Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
  6. This will be settled by fred911 · · Score: 3, Insightful

    When the firm representing the class bills up a few million, the defendant agrees to paying the fees and to mail all class members a $5 discount coupon or some useless download.

     

    --
    09 F9 11 02 9D 74 E3 5B - D8 41 56 C5 63 56 88 C0 45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
    1. Re:This will be settled by seepho · · Score: 1

      Bullseye. I bought a 970 and I'm apathetic about this for that exact reason.

    2. Re:This will be settled by Anonymous Coward · · Score: 0

      The point of class action is not to make profit(except for lawyers), it's there to make sure that businesses don't have the option to do small damages to a lot of people and get away with it.

  7. Good luck with that... by Kjella · · Score: 5, Interesting

    ...and I say that as an owner of 2x970s. Every benchmark you see is as advertised. It actually has 4GB of RAM that can be reached at the advertised speed, just not all 4GB of it simultaneously. Nobody's come up with a "smoking gun" benchmark where framerates tank for gaming. The 780 Ti with 3GB RAM beats it in pretty much benchmarks - even at 3840x2160 in SLI - so it seems that the last 512MB don't make much of a difference at all, at least not in today's games. They'd better find some compelling examples of actual harm, because I still haven't seen it. I might be biased though, since I'm kinda hoping there won't be.

    What is certain though is that nVidia screwed up big, because this really would have been a footnote if they'd just informed about it. It would have been known as a 4GB card that's really 3.5GB-ish. When I bought it I thought it had the same memory subsystem as the GTX980, like two GTX970 in SLI with 2x13 = 26/16ths the shaders will always perform better. Now that might not be true in a 3.5-4GB scenario but it's a maybe kinda thing. I've long since learned that you buy computers for what you want today, tomorrow.... maybe something entirely new comes around and you want to replace it anyway. Not that I see these two being out of date for a while, seriously kicking ass at 2x145W GPU + 88W CPU it's ~500W ass-kicking system.

    --
    Live today, because you never know what tomorrow brings
    1. Re:Good luck with that... by Anonymous Coward · · Score: 1

      The 970 is a fantastic card, I also have one and have never seen issues. HOWEVER, it definitely can't access all 4GB at the same speed. it can access the first 3.5GB at full speed, and the second 500mb at only a fraction of that full speed and it can't access both at the same time, hence when the second 500mb is in use it actually impacts the performance of the entire 4GB. For me there is no affect, but there are those with 4k displays and multi card configs that this definitely impacts, the effect is very real and most definitely can be huge. but regardless of whether you are affected or not it was a shitty scummy act on the part of Nvidia and their marketing of the product.

    2. Re:Good luck with that... by Anonymous Coward · · Score: 0

      ...and I say that as an owner of 2x970s. Every benchmark you see is as advertised. It actually has 4GB of RAM that can be reached at the advertised speed,

      Wrong.
      512 MB of the 4 GB RAM runs at a small fraction of the speed of the rest of the memory.
      It has 56 ROPs instead of 64.
      It has 1.75 MB cache instead of 2 MB.

      You're a fucking twit and the rest of your post is just corporate ballwashing. The fact that you admit to owning 2 970s in SLI just highlights how fucking retarded you are - for that money a single 980 would have been a far better choice.

    3. Re:Good luck with that... by Anonymous Coward · · Score: 0

      ...and I say that as an owner of 2x970s. Every benchmark you see is as advertised. It actually has 4GB of RAM that can be reached at the advertised speed,

      Wrong.
      512 MB of the 4 GB RAM runs at a small fraction of the speed of the rest of the memory.
      It has 56 ROPs instead of 64.
      It has 1.75 MB cache instead of 2 MB.

      You're a fucking twit and the rest of your post is just corporate ballwashing. The fact that you admit to owning 2 970s in SLI just highlights how fucking retarded you are - for that money a single 980 would have been a far better choice.

      Well if you're going to be that much of a dick, then I'll play the Devil's Advocate here. The RAM DOES run at the advertised speed and bandwidth. You just can't access it all at the same time, and there are issues with accessing the last 512 at full speed... but it's not technically the RAM which is gimped and causing the problem. So technically speaking their specs are not incorrect.

      But more to the point, I don't think they really understood the problem they had with this card. If they had, they'd have released it with drivers which would have been able to compensate for the quirks of the design, thus preventing the performance issues seen. And I think overall they're a bunch of assholes... but then again so are you.

    4. Re:Good luck with that... by aliquis · · Score: 1

      Just to make sure on the later part.

      Does any graphics card enable you to access memory from various parts at the same time?

      If not how is that part relevant?

      Maybe one can read from multiple of the other segments / chips at the same time what do I know.

    5. Re:Good luck with that... by Jesus_666 · · Score: 1

      (Note: This is all due to my understanding of the situation. I did not extensively research the GTX 970 and might be entirely wrong.)

      The problem is that GPUs usually have a uniform memory layout. If your GPU advertises 4 GiB of RAM then all 4 GiB of it behave in pretty much the same way. Accessing one part of the memory does not significatly affect accesses to other parts. Thus it's unnecessary to take special care in how to structure your memory handling; you just use whatever's there.

      The 970, as I understand it, has a non-uniform memory layout where the segment between 0x00000000 and 0xDFFFFFFF cannot be accessed at the same time as the segment between 0xE0000000 and 0xFFFFFFFF. Try to access one segment and all accesses to the other segment will stall until this one access has been handled.

      This could be used without appreciable performance impacts if the software accessing the memory is aware of it and specially structures its memory management so that accesses to the upper segment are sparse and happen in bulk (ie. it switches between blocks of lower segment accesses and blocks of higher segment accesses). That's the kind of optimization you see in game console programming and actually smells kind of like how PS3 games had to structure their memory handling around the Cell's peculiarities. If I remember correctly, this made the PS3 somewhat unpopular to develop for.

      Of course, no one in their right mind is going to add special Geforce GTX 970-specific logic to their game (potentially having to restructure half the engine for it) just to make best use of the hardware. Even making a codepath that detects the 970 and avoids the upper 0.5 GiB of VRAM entirely is unlikely. Thus, in situations where more than 3.5 GiB of VRAM are needed, the 970 will exhibit stuttering because of stalled memory accesses and there's not much anyone can do about it - except Nvidia, who could release a driver that reports the 970 as having 3.5 GiB of RAM.


      (I find it interesting how a Google search for "VRAM" ended up having several articles about the 970's slowness on the first page. I have never searched for the 970 before; my 660 from 2012 still has more power than I need.)

      --
      USE HOT GRITS WITH STATUE OF NATALIE PORTMAN (NAKED AND PETRIFIED)
    6. Re:Good luck with that... by Anonymous Coward · · Score: 0

      > Does any graphics card enable you to access memory from various parts at the same time?

      No, they cannot. In fact, separating tasks into separate 'threads' is the purpose of a stream multi-processor--

      You can read more about that here: Confidentiality Issues on a GPU in a Virtualized Environment, http://www.eurecom.fr/en/publication/4205/detail/confidentiality-issues-on-a-gpu-in-a-virtualized-environment, or here: www.cs.rice.edu/CS/Architecture/docs/spa.pdf

    7. Re:Good luck with that... by Anonymous Coward · · Score: 0

      to your point about nvidia handling this in the driver.... well, they have done exactly that.

    8. Re:Good luck with that... by g0bshiTe · · Score: 1

      Benchmarks show that SLI 970's outperformed a single 980 in every aspect.

      --
      I am Bennett Haselton! I am Bennett Haselton!
  8. No kidding by Sycraft-fu · · Score: 5, Interesting

    It is just a bunch of whiny asshats who care about specs on paper rather than real world performance. The 970 is damn amazing. It makes the 980 nothing more than a overpriced luxury toy, and I say that as a 980 owner. Its performance is within 10-15% of a 980s and it is like half the price, what's not to like?

    Also as for the memory thing this is actually a BONUS from nVidia, not a cripple. What I mean is in the past, they'd have just stuck 3.5GB on it and called it good. Then, if something needed more than 3.5GB, you go to system memory which is very slow 16GB/sec if you are running 16x PCIe 3 and much slower if you run less (like if you are doing SLI on a consumer board with PCIe 2 it would be 4GB/sec). However with this, you get another 512MB of RAM that is faster. Not as fast as the primary RAM, but much faster than hitting the system RAM over the PCIe bus. It won't perform as well as a 980 in those high memory situations, but it would perform better than if it just didn't have it at all.

    I agree they should have noted it better, but really who gives a shit in reality? The 970 is the best "step down" card they've ever made compared to the highest end. Amazing value for the money and real world benchmarks from somewhere like HardOCP show it kills at modern games.

    It's also funny how they act like nVidia did this to "harm" people for some business reason. If anything, they'd want to make the 970 look worse so people would be more likely to spend the near double to get a 980. However instead they made the 970 as close to the 980 as they could and I'm sure that ate in to the 980 market.

    1. Re:No kidding by Anonymous Coward · · Score: 0

      It's also funny how they act like nVidia did this to "harm" people for some business reason. If anything, they'd want to make the 970 look worse so people would be more likely to spend the near double to get a 980. However instead they made the 970 as close to the 980 as they could and I'm sure that ate in to the 980 market.

      If they had the 970 advertised so that it looks worse, as it is in fact, then a lot of buyers would have went for AMD cards.

      Also, just because it doesn't impact on your system right now, it doesn't mean that it won't for anyone else and for all times.

      I think the lawsuit will do good to show gfx manufacturers where to draw the line between marketing blabber and outright false advertising.

    2. Re:No kidding by Anonymous Coward · · Score: 0

      Wow. Its not the performace of the card that is in question here, it's blatant lying about the card's actual specs.

    3. Re:No kidding by bloodhawk · · Score: 3, Informative

      It's also funny how they act like nVidia did this to "harm" people for some business reason. If anything, they'd want to make the 970 look worse so people would be more likely to spend the near double to get a 980. However instead they made the 970 as close to the 980 as they could and I'm sure that ate in to the 980 market.

      That is just silly. Nvidia benefit by making the product look better to highend gamers so they don't choose the competitions card. You act like Nvidia is the only maker of highend gaming cards. It is actually worse than a 3.5GB card as at least with 3.5GB gamers and drivers by default understand they are going to have a performance impact if beyond that is accessed and hence will only use it when necessary, with 500MB of slow memory by default the games had no understanding they would be crippling game performance by accessing that part of the memory. The 970 is a nice card, but Nvidia did blatantly mislead purchasers and while the issue won't even impact more than 90% of owners it is still an issue and one that could have been avoided by Nvidia being honest.

    4. Re:No kidding by Anonymous Coward · · Score: 1

      That extra 0.5Gb is not only worthless, it actually IS detrimental! The card still advertises a unified 4GB of memory space to the application. There is no mechanism to advertise that this remaining 0.5Gb is 5x slower with XOR contention against the lower 3.5Gb. As a result, the underlying application treats the entire address space uniformly (i.e. if you were playing this game with a 3.5Gb card, it would EITHER always be consistently bad (since stuff would be moving across PCI-E), or the game would auto-lower its quality) Playing the same game on a GTX 970 "4GB" card just results in inconsistent performance.

      This explains a ton of micro-stuttering / stuttering problems and general performance inconsistency people have been seeing with the 970's at higher memory usage levels. (e.g. https://www.youtube.com/watch?v=RsQsP0dyQNQ#t=58 That video doesn't just reflect a 10-15% difference in FPS performance between the 970/980. It shows extreme stuttering in the 970 when things exceed 3.5Gb... THIS is the type of thing every 970 owner will have to deal with as newer games use more VRAM)

      This is also unfortunate, since most benchmarks just report average FPS! As a result, the 970 benchmarks at higher resolutions artificially make the it look better than it actually is (i.e. the average numbers might look impressive, but the standard deviation when memory gets juggled above the 3.5Gb boundary may make it completely unplayable at that setting in real life).

    5. Re:No kidding by PitneFor · · Score: 0

      Wrong. The issue is, people bought the 970 expecting all 4GB of ram to be full speed. The 'real world' performance you speak of is relative. A person may have a certain computer configuration to where 'real world' performance is impacted. Also, people purchase graphics cards for long time usage. As time passes, the amount of graphics memory that software uses increases. Software two years from now are more likely to use full 4GB of ram than software today. Had the card been advertised correctly in the first place, many people would have not purchased the 970 at all, because of the above reasons I mentioned. For this, NVIDIA should be sued; it is not acceptable. To the people who purchased the 970, they have no recourse. They spent a good deal of money on a card that they would not have otherwise purchased if they were given the correct facts when the card was offered for sale.

    6. Re:No kidding by Anonymous Coward · · Score: 0

      It is just a bunch of whiny asshats who care about specs on paper rather than real world performance.

      Really? Are you arguing that it's okay for products not to be up to spec?

      Specs let you know what you are getting and if you look at the nvidia spec for 970: http://imgur.com/CFUOUzs
      Where does it say that ONLY 3.5GB operates at the specified bandwidth?

    7. Re:No kidding by Anonymous Coward · · Score: 0

      Wow. Its not the performace of the card that is in question here, it's blatant lying about the card's actual specs.

      I read the actual specs. I don't see this lie you people keep claiming. Post the lie, verbatim, and point out exactly why it's a lie. I'm not saying they're in the Right, I'm saying that you and most other people are making a LOT of assumptions about what they promised vs. what the actual problem is with the product.

    8. Re:No kidding by bspus · · Score: 1

      I wonder if it would be possible to completely disable the "problematic" 0.5GB of RAM by a patch or driver update.

      Then it would really just be a 3.5GB card and avoid those issues.

      I'm sure that just the idea that whatever stuttering is experienced at any given game *might* be due to the RAM issue is enough to drive some of those gamers mad.

      So if it could be disabled (optionally of course) it would be good I think.

    9. Re:No kidding by g0bshiTe · · Score: 1

      The game I bought the card specifically to play allows me to specify how much VRAM the game will use.

      Today I'll never see that problematic .5GB of memory.

      --
      I am Bennett Haselton! I am Bennett Haselton!
    10. Re:No kidding by g0bshiTe · · Score: 1

      As the purchaser of a 970 I could agree with that had my only reason for buying the card been the advertised 4GB VRAM, which it wasn't.

      --
      I am Bennett Haselton! I am Bennett Haselton!
    11. Re:No kidding by RyoShin · · Score: 1

      It is just a bunch of whiny asshats who care about specs on paper rather than real world performance.

      If "real world performance" cannot at least meet "specs on paper", then it is false advertising.

      Yes, in this case it's an extremely little thing compared to the overall card, but corporations are trying their damnedest to slip in/out whatever they can at the expense of the consumer. So long as they can get away with it (which includes paying a fine/lawsuit that costs less than the profit from their misleading statements) they will continue to do it. If we can use lawsuits or laws to stop such practices now, while the claims are relatively small, it's for the best.

      If we instead say "eh, that's no big deal", then it will continue and latter suits will be harder to win. At some point it might become easy to win, say when the box reads "Will not set your house on fire!", no less than five house fires are caused directly by normal operation of the card, AND the card spontaneously bursts into flames just sitting on the defendant's desk, but it's in our best interest to fight these practices now rather than waiting for that time.

  9. Deep pockets by Anonymous Coward · · Score: 1

    there's a whole industry devoted to suing "deep pockets". I once attended a presentation on this that explained to lawyers which companies they should target, how long they have to wait until they should gather plaintiffs for a class action law suit and how lucrative this is. There are even people financing these suits as an investment. I'm still not sure what is more unethical, lawyers leeching off of companies or companies lying to customers.

  10. Just bought a mITX version of the 970 by Anonymous Coward · · Score: 0

    I don't see the problem if there is one (yet?)?

    This card is the smaller form factor one meant for mITX platforms (stuffed in hadron hydro case), supposedly it is a tad bit slower and not the greatest for over clocking (which i have done none of). I could water cool it and change that, but I haven't.

    I haven't used all of the memory yet. When I open 3x Eve game clients (the norm if ever) I'm only tapping a little under 2gb. I haven't checked all games but I have checked a few and most of the games I do play even in the highest detail don't use but half of the video memory (gpuz).

    Now that said if Giga-byte wants to offer a juicy deal to GTX 970 owners to shut this guy up, I can guarantee my next purchase will be another giga-byte product (and I'm a regular!)!

  11. frivilous by Anonymous Coward · · Score: 0

    'nuff said.

    Where's my noprize?

  12. Re:Just a thought by 0123456 · · Score: 2

    No. The reason it affects the 970 is because it's missing some of the hardware that's in the 980.

  13. Still waiting for the Purevideo functionality by Anonymous Coward · · Score: 0

    Also, 3d Tech News and Pixel Hacking reported that the 970 doesn't have this issue when running games in OpenGL mode.

    http://www.geeks3d.com/20150127/opengl-apps-not-impacted-by-gtx-970-vram-limitations/

  14. Re:Just a thought by arbiter1 · · Score: 1

    970 is only effected cause a shader core and ROP are disabled cause they are bad. the last 512mb ram shares a cluster so its slower cause that. Reviewers have test this and only time it became a problem is running super high rez with MSAA and scaling, pretty much to point probably running sub 40fps anyway.

  15. Oh boy! I can't wait! by EmagGeek · · Score: 1

    I can't wait for my postcard-check in the mail for $0.21 explaining how the lawfirm got $34,500,000 of the $35,000,000 settlement and the remaining $500,000 gets divided among all GTX970 owners.

    1. Re:Oh boy! I can't wait! by Anonymous Coward · · Score: 0

      Even better if Nvidia were caught paying the law firm to start the case, to avoid paying out an even bigger settlement because those numbers look a bit low if this involves defrauding customers / investors.

  16. I got a problem with my GTX 970 by Anonymous Coward · · Score: 0

    Just as the article describes, while playing Battlefield 4 I'll have these sudden moments of graphics lag. It happens quite often.

  17. in other news... by Anonymous Coward · · Score: 0

    I'm suing Atari, because when I needed help I pressed the HELP key on my 800XL, but help never came.

  18. CUDA and OpenCL by Anonymous Coward · · Score: 1

    It does matter for CUDA and Opencl performance. So even though games might work around it (except on the mentioned high resolutions), it simply runs into problems if you are going to use it for computing on >3.5GB.

  19. SHOCKING interview with Nvidia engineer by Roman+Mamedov · · Score: 1, Insightful

    SHOCKING interview with Nvidia engineer about the 970 fiasco https://www.youtube.com/watch?...

    1. Re:SHOCKING interview with Nvidia engineer by Anonymous Coward · · Score: 0

      SHOCKING interview with Nvidia engineer about the 970 fiasco https://www.youtube.com/watch?...

      Loved the brilliant clarity that this engineer used to explain the harshness of the GPU world today. Its dog eat dog... all about the $$$.

      Fantastic look inside NVIDIA ... must see.

  20. Re: Hurr durr by Anonymous Coward · · Score: 0
  21. tldr; no you're an idiot by Anonymous Coward · · Score: 0

    Where to start?

    Given your love for calling people idiots, I'll start here:
    "I owned a GF4MX because it was a fantastic value. As a budget gamer, it was the best buy at the time if you weren't playing games which demanded shaders."

    Nope... the GF4MX was a terrible value, considerably worse than the Geforce 3, and worse than a comparable card from ati. The fact that you thought it was a good value is just a greater indication of nvidia's spin campaign at the time. The real value at the time was the GF3 Ti200 which could typically be overclocked to Ti500 levels, or just short thereof.

    Let's also not forget the considerable lack of honesty in marketing a product as the Geforce4 when it was in fact slower, and older than the Geforce 3. This was the first point that a computer industry company employed this practice to my knowledge (sadly it's rather commonplace now). To compare with your truck analogy it would be like Ford bringing an "F450" to market which actually has a lower towing capacity than the F350 (note, I realize that Ford already has an F450, but the analogy still stands).

    1. Re:tldr; no you're an idiot by drinkypoo · · Score: 1

      Nope... the GF4MX was a terrible value, considerably worse than the Geforce 3,

      it was a lot cheaper than the GF3, especially if you bought an OEM card.

      Let's also not forget the considerable lack of honesty in marketing a product as the Geforce4

      They didn't. It was the 4MX. It did in fact have some of the bits of the GF4 slapped onto its GF2 roots.

      To compare with your truck analogy it would be like Ford bringing an "F450" to market which actually has a lower towing capacity than the F350 (note, I realize that Ford already has an F450, but the analogy still stands).

      People get confused that diesels have less hauling capacity than gassers because of the heavy motor all the time, but nobody sues over it.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"