Law Firm investing Nvidia for potential Class Action.

[Headfoot]

Class action suits don't really benefit the victims, but paying out enough in damages might keep nvidia from hiding problems a third time (bumpgate + VRAMgate).

And threat of suits has forced companies like Apple to "make things right" in the past.

Very true. The class action mechanism is better understood as a society-wide enforcement mechanism, like a back-up FTC/SEC etc. The victims might see a couple of bucks. But the goal is more so that the companies fear the cost of the class action mechanism and that acts as a deterrent in future cases.

 

[Enigmoid]

If you can do 2+2 too, you will understand that if that 0.5GB partition only can access 32b of bus width, you can NEVER claim the card has full 256b. It is indeed as the editor claimed a XOR situation. A trully 256b part would have the same access speed across every chunk of memory, while the 970 has only 224b bus width for the 3.5Gb chunk, and the other chunk has only 32 bit. IT IS NOT ADDITIVE as you paint it. So please cut the nonsense.

Usability is certainty XOR but you physically do have a 256 bit bus with 224 GB/sec. Whether you can use that is a different thing.

My car's speedometer goes to 260 km/h. My engine and tires will not survive 220 km/h.

IMO, the vram is the concern. The ROP count and bandwidth is easily offset by the loss in shaders. As far as marketing, Nvidia never marketed the L2 and I can't recall marketing on the ROP count of the 970. The bandwidth, bus, and vram were marketed but as they technically are there, I suspect it would be hard to sue Nvidia.

I don't like it but that's the way it is.

 

[frozentundra123456]

Very true. The class action mechanism is better understood as a society-wide enforcement mechanism, like a back-up FTC/SEC etc. The victims might see a couple of bucks. But the goal is more so that the companies fear the cost of the class action mechanism and that acts as a deterrent in future cases.

Actually it is a society wide mechanism for lawyers to make a lot of money.

 

[AnandThenMan]

As far as marketing, Nvidia never marketed the L2 and I can't recall marketing on the ROP count of the 970. The bandwidth, bus, and vram were marketed but as they technically are there, I suspect it would be hard to sue Nvidia.

I don't like it but that's the way it is.

It never appeared on Nvidia's site but the info was sent out as part of the review guide. Nvidia is responsible for this information being correct, which it still is not on many sites, one example. Nvidia can claim they are not responsible for what 3rd party sites publish, but Nvidia links to these reviews from their own site so that doesn't fly.

 

[Silverforce11]

Usability is certainty XOR but you physically do have a 256 bit bus with 224 GB/sec. Whether you can use that is a different thing.

You don't, because they fused 1 off. The 0.5gb segment shares the lane with the neighbor. It's not 256 bit bus.

If its 256 bit bus, then the first 3.5gb can be fully accessed at 224 gb/s, but it can't because its fused off so its only 196 gb/s.

Because of the XOR nature, it never achieves the advertised 224 gb/s throughput. Hardware.fr and TR actually noted that in the initial 970 launch, its bandwidth was lower than advertised but they didn't follow up, due to taking NV's word on it.

It would have been better off without that split segment or sharing of bandwidth, they would have sold just fine as 3.5GB, at its launch $330 was a bargain compared to what's at the market. This whole situation is stupid because NV didn't NEED TO LIE. They would have been just fine and gamers would have went with the 970 happily if its 3.5gb. I mean the 780/ti sold fine and it was only 3gb compared to the R290/X!

Edit: I think if the 970 was 3.5gb they would have likely made more 980 sales, getting even more profit. Not sure why they went stupid mode then lied about it.

 

[SteveGrabowski]

Very true. The class action mechanism is better understood as a society-wide enforcement mechanism, like a back-up FTC/SEC etc. The victims might see a couple of bucks. But the goal is more so that the companies fear the cost of the class action mechanism and that acts as a deterrent in future cases.

It's a slap on the wrist.

 

[Exophase]

pcper has reported from an nVidia engineer that the two buses can actually satisfy requests simultaneously. The confusion is because there are only seven L2 cache interfaces, but the 7th one can queue multiple requests on the two memory controllers in parallel.

It would have been better off without that split segment or sharing of bandwidth, they would have sold just fine as 3.5GB, at its launch $330 was a bargain compared to what's at the market.

The 3.5GB solution would have still used a 224-bit bus. So no performance benefit vs simply ignoring the last 512MB. In theory it could gain some performance benefit using that RAM, but in practice this would be difficult to realize and it's a lot more likely performance would be worse instead.

Actually it is a society wide mechanism for lawyers to make a lot of money.

Pretty much. Law firms typically get 25% of the collective payout, which can be several dozen or even hundreds of millions of dollars. I don't see why their fee should be proportional to the earnings at all. It would be better if the court awarded the lawyer a fixed amount based on what seems fair and representative of the case they presented.

 

[chimaxi83]

Why do some of you insist, at all costs, on defending Nvidia? You people know damn well that you'd crucify AMD if they did this same exact BS. Give it up already, it's pitiful.

If this lawsuit gets off the ground, hopefully it does more than just line the firms pockets, since that's all these class actions seem to be good for.

 

[MagnusTheBrewer]

Where there's smoke there's lawyers.

 

[frozentundra123456]

Why do some of you insist, at all costs, on defending Nvidia? You people know damn well that you'd crucify AMD if they did this same exact BS. Give it up already, it's pitiful.

If this lawsuit gets off the ground, hopefully it does more than just line the firms pockets, since that's all these class actions seem to be good for.

I dont know if you are specifically directing this to me, but I am not saying nVidia is blameless. I dont think anyone who doesnt get worked up over this has to be classified as an nVidia fanboy. I actually think AMD video cards offer the best value in most segments at the current time. So I am hardly one to go around bashing AMD dgpus.

I think at best nVidia was careless, and in all probability deceptive. My best guess is that they mis-communicated the facts initially and then tried to cover up the mistake. But the chances are no-one will ever really know exactly what happened. I just think some people are over-reacting, and I could just as easily claim that they are so outraged just because it is nVidia, as you are claiming people react towards AMD. I mean it is not like a car defect that is killing people, or a drug that is causing serious harm via side effects. People are getting the same performance they saw in the benchmarks when they bought the card, with the caveat that there are some situations in which you might see a few FPS less than you expected based on the specs. Yes, it is not right, but in the grand scheme of things, I am not sure it merits an expensive lawsuit, which ultimately will probably not really benefit the purchasers anyway.

In any case, I think nVidia should offer a free exchange or some sort of voucher toward say a Steam game or a future purchase. I just think a class action suit is a bit of an over-reaction, and is really not going to help anyone who actually bought a card, because by the time the suit is settled, all current cards will probably be obsolete anyway, and any award to the end user will be basically worthless.

Edit: BTW, I just got a letter in the mail about a class action lawsuit that had been settled to give something like a 10 or 15 dollar refund to purchasers of P4 computers during the early 2000s for some apparently deceptive way the chip was advertised. Ten years plus to settle the case, an award to me that I didnt even bother to send in, and probably no effect on intel or HP (who made the computer) after all these years. Basically a waste of time for everyone except the lawyers who undoubtedly made millions. This is why I am not a fan at all of class action suits, unless there is some sort of egregious situation that is causing serious consequences.

 

[Elixer]

Edit: BTW, I just got a letter in the mail about a class action lawsuit that had been settled to give something like a 10 or 15 dollar refund to purchasers of P4 computers during the early 2000s for some apparently deceptive way the chip was advertised. Ten years plus to settle the case, an award to me that I didnt even bother to send in, and probably no effect on intel or HP (who made the computer) after all these years. Basically a waste of time for everyone except the lawyers who undoubtedly made millions. This is why I am not a fan at all of class action suits, unless there is some sort of egregious situation that is causing serious consequences.

Which is why lots of companies try to pull off deceptive practices. They know that by the time all is said and done, they just pay a small fine that isn't anywhere big enough to prevent them from doing this again...and again.

 

[digitaldurandal]

False advertising has nothing to do with the 970 card.

*Quick Google Search*

"Proof Requirement
To establish that an advertisement is false, a plaintiff must prove five things: (1) a false statement of fact has been made about the advertiser's own or another person's goods, services, or commercial activity; (2) the statement either deceives or has the potential to deceive a substantial portion of its targeted audience; (3) the deception is also likely to affect the purchasing decisions of its audience; (4) the advertising involves goods or services in interstate commerce; and (5) the deception has either resulted in or is likely to result in injury to the plaintiff. The most heavily weighed factor is the advertisement's potential to injure a customer. The injury is usually attributed to money the consumer lost through a purchase that would not have been made had the advertisement not been misleading. False statements can be defined in two ways: those that are false on their face and those that are implicitly false."

I am not a lawyer but it seems to be all 5 requirements can be proven.

Can you explain why you think it does not? Are you allowing bias to manipulate your opinion?

EDIT: to avoid confusion - please note that (5) is not limited to physical injury (see Pentium 4 lawsuit.)

 

[chimaxi83]

snip

Wasn't directed at you at all, just a general statement towards those in this and other threads deflecting this issue. You're right, this is really just a first world problem. In the grand scheme of things, not a huge deal, but it definitely deserves harsh criticism and some kind of slap on the wrist. My comment was about people acting like this is just nothing, when we know that based on their posting history, they'd be judge, jury, and executioner towards AMD.

 

[Exophase]

I am not a lawyer but it seems to be all 5 requirements can be proven.

Did you actually read all five, or do you really believe that not knowing the correct L2 cache sizes or ROP counts can lead to injury?

 

[digitaldurandal]

This is why I am not a fan at all of class action suits, unless there is some sort of egregious situation that is causing serious consequences.

In a capitalist society, these lawsuits are one of the only things keeping companies from deceptive marketing. Without these lawsuits you disagree with, companies would routinely do this without consequence. With these lawsuits, companies are forced to explain to shareholders why they lost millions in a lawsuit and are held to a standard by the shareholders and consumers.

Is it true that lawyers make more money than anyone in these cases and perhaps a system with better reward for consumers should be set up, but to be completely dismissive of them is ignorant.

 

[digitaldurandal]

Did you actually read all five, or do you really believe that not knowing the correct L2 cache sizes or ROP counts can lead to injury?

Injury is not limited to physical injury. Hence why plenty of suits against companies for products have been successful even though the product did not physically harm anyone. It is often a monetary injury.

Do you think the P4s referenced earlier in this thread caused physical injury?

 

[coercitiv]

pcper has reported from an nVidia engineer that the two buses can actually satisfy requests simultaneously.

That sounds really good, sounds as if the GPU has full simultaneous access to both the 7th and the 8th chip. So why not let the 8th chip work in parallel with the others in stripped operations?

 

[Enigmoid]

You don't, because they fused 1 off. The 0.5gb segment shares the lane with the neighbor. It's not 256 bit bus.

If its 256 bit bus, then the first 3.5gb can be fully accessed at 224 gb/s, but it can't because its fused off so its only 196 gb/s.

Because of the XOR nature, it never achieves the advertised 224 gb/s throughput. Hardware.fr and TR actually noted that in the initial 970 launch, its bandwidth was lower than advertised but they didn't follow up, due to taking NV's word on it.

It would have been better off without that split segment or sharing of bandwidth, they would have sold just fine as 3.5GB, at its launch $330 was a bargain compared to what's at the market. This whole situation is stupid because NV didn't NEED TO LIE. They would have been just fine and gamers would have went with the 970 happily if its 3.5gb. I mean the 780/ti sold fine and it was only 3gb compared to the R290/X!

Edit: I think if the 970 was 3.5gb they would have likely made more 980 sales, getting even more profit. Not sure why they went stupid mode then lied about it.

There is a physical 256 bit connection from the GPU to the dram. That entire connection is not usable, but it is there.

The 256 bit bus runs at 7 ghz meaning it has a theoretical bandwidth of 224 GB/sec. That is not completely usable but all in all that is a fairly useless metric given that overhead exists for any interface.

Have you seen all the bandwidth tests from something like the 4770k, a10-7850k, and 8350? With the same DDR3 1600 mhz CL9 RAM they all will have different bandwidths even though the theoretical max is 25.6 GB/s. The efficiency comes into play as well.

I don't support Nvidia on this. However, this is precisely how the lawyers are going to spin it.

 

[Gloomy]

It's got a 256-bit memory interface connected to a 224-bit bus.

 

[Enigmoid]

It's got a 256-bit memory interface connected to a 224-bit bus.

I would draw the parallel of a 8 lane expressway narrowing to 6 lanes before widening to 8. Its an 8 lane expressway. However, that one small 6 lane section is going to hold traffic up like a 6 lane expressway.

Is it still a 8 lane expressway?

 

[SlowSpyder]

I would draw the parallel of a 8 lane expressway narrowing to 6 lanes before widening to 8. Its an 8 lane expressway. However, that one small 6 lane section is going to hold traffic up like a 6 lane expressway.

Is it still a 8 lane expressway?

Is it advertised as an eight lane expressway?

 

[Techhog]

None of that mattered, the card performed exactly like it was suppose to. They won't loose the case, because it will be thrown out before it starts. It still blows my mind that people don't actually understand what false advertising is, yet they continue to throw it around on the internet without even researching it. If you want to know how terrible of a case using false advertising is look no further than the millions of cases stupid people try to sue for and lost. Because they thought false advertising actually means what it says. False advertising has nothing to do with the 970 card.

So, flat-out lying (not being misleading, but actually lying) about the specifications of a product isn't false advertising? Then what is? I'd see your point if it were just the memory config since it actually does have 4GB, but other specs were just plain incorrect. They might as well have said that it was an underclocked GTX 980.

I swear, some of you will defend Nvidia for anything.

 

[Exophase]

Injury is not limited to physical injury. Hence why plenty of suits against companies for products have been successful even though the product did not physically harm anyone. It is often a monetary injury.

Do you think the P4s referenced earlier in this thread caused physical injury?

Okay, I guess my confusion was that I didn't see that it said all five need to be met and thought that just any of them needed to be. I think that injury is a terrible choice of word for more general loss or damage.

 

[Hauk]

Lol. Save the myopic ranting for an OT forum

 

[Techhog]

Did you actually read all five, or do you really believe that not knowing the correct L2 cache sizes or ROP counts can lead to injury?

"The injury is usually attributed to money the consumer lost through a purchase that would not have been made had the advertisement not been misleading."

So...