NVIDIA Maxwell Speculation Thread

Ambulance chasers at their worst ....

Lawyers homing in on Nvidia after GTX 970 Memory allocation Claims

A law firm from the US that specializes in class action litigation is now homing in on Nvidia, this was to be expected in some form. In our findings widely explained here the firm is seeking people who bought a GeForce GTX 970 and is asking them how it performs.

Here is that request from the New York based Bursor & Firsher, P.A. Remember guys, this law firm is investigating. Also realize that any money in lawsuit often go to greedy cash-grab lawyers, not so much the customers. But that is just my personal opinion of course.
http://www.guru3d.com/news-story/la...a-after-gtx-970-memory-allocation-claims.html

index.php
 
Lawsuit based on L2 cache and ROP count.. Is this ever going to stick?
 
No lawsuit yet at least according to what's been posted. But you'd expect something like that to happen in the US, no?

As I've heard, lawyers are paid there according to the amount of damages won. So they'll at least try to get something out of the misfortune of nv..
 
How can they possiby progress a case on the basis of asking users how the card performs? Err, compared to what? Has Nvidia ever given guarenteed performance levels for specific games/settings? Nope. All we have are review benchmarks which are clearly going to be totally representitve of the cards performance. Anyone who purchased this thing based on advertised performance (reviews) got exactly what they expected.
 
The bar for filing a suit is not that high, and that's a different thing from winning it. The bar is even lower for a law firm asking "is there anything to this", which is what it appears like so far.
 
Anyone who purchased this thing based on advertised performance (reviews) got exactly what they expected.
Well, yes and no. Real-world performance might differ from advertised claims (do NV even really claim anything in the way of performance figures in their ads - to the extent they even advertise at all? I hardly ever see any NV GPU ads whatsoever), due to the undisclosed deficiencies in the 970 ASIC.

Making fraudulent statements is sue-worthy after all. I'm sure we don't want to condone a situation where a GPU maker knowingly posts false specs and we let them get away with it because we all despise frivolous lawsuits.
 
Apparently this law firm is real sleazy ...

In its lawsuits, AT&T accuses Bursor & Fisher of misleading potential claimants by suggesting they can win $10,000 if anybody prevails. (Under a heading “The $10,000 Payment,” the firm says customers can win that much “under certain circumstances.”)

AT&T also accuses Bursor of recruiting “a familiar cast of characters as the figurehead plaintiffs.” Thirteen of the 26 claimants are either Faruqi attorneys or people who have served as plaintiffs in previous Bursor and Faruqi cases. Plaintiff Richard Colosimo, for example, is identified as a former Faruqi attorney who served as a plaintiff in a 2010 class action against AT&T, while Faruqi represented two other plaintiffs in a labor lawsuit against a health club.

Bursor doesn’t deny any of it.

http://www.forbes.com/sites/danielf...tion-victory-breeds-swarm-of-antitrust-cases/
 
Making fraudulent statements is sue-worthy after all. I'm sure we don't want to condone a situation where a GPU maker knowingly posts false specs and we let them get away with it because we all despise frivolous lawsuits.

I could be wrong but don't recall seeing an Nvidia ad that stated all 4GB would be used to attain highest level of performance when using the 970 GTX, it was more assumed by everyone that 4GB was required to reach the level of performance that Nvidia envisioned for the card. We now know otherwise ...
 
Last edited:
I could be wrong but don't recall seeing an Nvidia ad that stated all 4GB would be used to attain highest level of performance when using the 970 GTX
An omission of fact (that not all 4GB of your card is useable), is equivalent to a lie. Also, could be considered a bait-and-switch, which is quite illegal in many countries.
 
Bait and switch involves luring a customer to a store with the promise of one product with the plan to sell them something inferior instead.
There's only one 970 product, to my knowledge, so nothing is being switched.
Other examples may include some kind of service that incurs surprise fees, which does not seem applicable.

False advertising seems like it is still the most likely angle for any suit.
 
Apparently..Samsung is a bit ahead of TSMC on FINFET (they entered mass production in December) ..and has also achieved lower feature size and lower power. These are all rumours though so take them with a healthy grain of salt.

So this was confirmed by Nebuchadnezzar on another thread, i.e. Samsung 14nm is higher density and lower power. That's a pretty strong incentive for anyone to switch.
False advertising seems like it is still the most likely angle for any suit.

That's the way I see it as well..although as usual its the lawyers who are the real winners in all these suits.
 
I have a hard time believing that a lawsuit will go anywhere. I just went over the boxes my 2 970s came in, as well as their marketing material, and all reviews associated with them that I read. In all of those, the ROP count was mentioned in one table on AnandTech incorrectly (and has since been corrected). My cards still have 4GB of memory each, and while .5GB of that is not ideal, all 4 GBs are usable. The cards perform exactly how the review sites said they would with comparable benchmark scores, and have so far lived up to the marketing claims on the box (cooler, quieter, lower fan power, and increased lifespan were the marketing claims). There doesn't seem to be any false advertising claim, because the marketing material I got all omitted technical specs like ROP count.

I spent several days going over everything I could find on the issue in an attempt to get myself worked up about it. In the end, I bought these cards for GPGPU, so the ROP change doesn't really matter to me. Can't keep them filled anyway. The memory sounds a lot more problematic, but I haven't seen any problems with it. After everything I've found I just can't manage to get myself worked up. The cards still work great, and outperform anything else I had access to at the time. My GPGPU code ported perfectly and has had no problem with the memory partitioning. My games play great, even at higher resolutions. When I take a step back from all the "NVIDIA is lying and evil!" stuff, I guess I feel like I got what I paid for.

I think the switch to Samsung is a much more interesting story at this point. Especially if it pushes out the release of future cards. Seems like a big change to be making right now for NVIDIA.
 
I miss the goold old days of evil like shader replacement and seriously ugly texture hacks. They made some games look awful and hacked up 3DMark in fascinating ways. They didn't get sued back then. They did get sued over all of their dead notebook and desktop chips during the bad solder age though.
 
I have a hard time believing that a lawsuit will go anywhere. I just went over the boxes my 2 970s came in, as well as their marketing material, and all reviews associated with them that I read. In all of those, the ROP count was mentioned in one table on AnandTech incorrectly (and has since been corrected). My cards still have 4GB of memory each, and while .5GB of that is not ideal, all 4 GBs are usable. The cards perform exactly how the review sites said they would with comparable benchmark scores, and have so far lived up to the marketing claims on the box (cooler, quieter, lower fan power, and increased lifespan were the marketing claims). There doesn't seem to be any false advertising claim, because the marketing material I got all omitted technical specs like ROP count.

I spent several days going over everything I could find on the issue in an attempt to get myself worked up about it. In the end, I bought these cards for GPGPU, so the ROP change doesn't really matter to me. Can't keep them filled anyway. The memory sounds a lot more problematic, but I haven't seen any problems with it. After everything I've found I just can't manage to get myself worked up. The cards still work great, and outperform anything else I had access to at the time. My GPGPU code ported perfectly and has had no problem with the memory partitioning. My games play great, even at higher resolutions. When I take a step back from all the "NVIDIA is lying and evil!" stuff, I guess I feel like I got what I paid for.

I think the switch to Samsung is a much more interesting story at this point. Especially if it pushes out the release of future cards. Seems like a big change to be making right now for NVIDIA.
Two that might be mentioned on package, and which are definitely wrong, are membandwidth and membus, as 224GB/s and 256bit simply aren't true (since it's 196Gbps and 224bit or 28Gbps and 32bit at once, never both)
 
Two that might be mentioned on package, and which are definitely wrong, are membandwidth and membus, as 224GB/s and 256bit simply aren't true (since it's 196Gbps and 224bit or 28Gbps and 32bit at once, never both)

Per marketing wiggle room, the bus is physically 256 bits and it physically offers the rated bandwidth. The GPU can access both partitions, in the case of a read operation on one partition and a write on the other.
Did the box specifically state read/write bandwidth?
 
I know here is Australia the online stores I shop at either copy the specs or provide an image of the specs taken from the AIBs websites. As such, every single listing for the 970 I have come accross has listed the memory bandwidth as 224GB/s. I've also found this to be the case on newegg when I briefly checked a few different 970's there just out of price comparison curiosity.

EDIT: Just checked again and Newegg only mention 256-bit memory. Maybe I misremembered or maybe it was only certain AIBs.

Checking back at my local online store it seems like EVGA, Gainward and Galax mention memory bandwidth (image from their website marketing) while ASUS and Gigabyte do not.
 
Last edited:
Those weren't any "rumors" per se. Anything for =/>600mm2 were guessed values based on the GM200 die shots that leaked out. Rumors long before that shot appeared from supposed insider information where in the 560+mm2 range. Die size however doesn't necessarily answer his question. How powerful? My guess is on standard clocks for the full chip =/>35% on top of GTX980 performance on average.
 
Back
Top