News Nvidia denies GeForce GTX 970 design flaw claims

And if your settings are so high that you need a whole 4GB for texures (for example) you have more to worry about than the paging 0.5GB of those textures. Such as paging the non critical GPU workload to the much slower system RAM.

 

it depends on the game really to how and where the textures are beings stored....

 

Doesn't something like that depend more on the drivers and the OS ?

 

'If this was a "miscommunication" between marketing and engineering how does the fact that GPU-z reports 64 ROPS for this card, did marketing write the BIOS for the cards '

 

GPU-Z does not have a magical card sense. It relies on looking up the available functional blocks, and until now nobody had any idea that Nvidia could disable and reroute components inside a block. Even Nvidia's technical marketing team had no idea. This seems like an advancement that Nvidia's engineering team were holding very close to their chests (because it greatly increases yeilds without as dramatic a loss in performance as disabling an entire block).

 

Isn't some of the GPU-z information based on released tech specs.

 

There must be a backlash. It is false advertising. If this is true about GPU-Z reporting 64 ROPS it is scandalous, someone at Nvidia must have known the whole time.

I hope a consumer with money to burn sues Nvidia's ******s off, there is a landmark ruling and we all get the option for a refund. *hopes and dreams*

 

Good article on anandtech: http://www.anandtech.com/show/8935/geforce-gtx-970-correcting-the-specs-exploring-memory-allocation

As a 970 owner while I am miffed they didn't give the correct specs tbh it doesn't make a lot of difference - the performance is still the same as it was in every review done. It's not like it would have effected my decision to buy the card which was based on performance, price, noise, features and power usage which haven't changed.

 

There is already a petition for refunds. Nvidia are going to need to do something I imagine, if anything for future sales.

 

Yea, someone at Nvidia knew the whole time and thought they could hide it from the tech community, shh they'll never find out.

Don't you have to show intent with regards to false advertising ? If so good luck with demonstrating that Nvidia intentional mislead people and didn't think anyone would notice.

Isn't that what Kickstarter or what ever is for, bound to be some community effort to sue them in a few days.

 

Corky42 and myself is the difference between someone who is directly affected by this issue, and someone who is not.

I am certain your perspective would change if you had over a grand invested in these cards, and a screen to utilize them.

 

Let's sue a GPU company for releasing a cheap, powerful card that in very specific circumstances possibly performs a little slower than we thought. Seriously? It's time to return to the real world now...

If you want raw performance, buy the top-end card. Everything else is going to be a cost-performance compromise.

 

Greetings!

For that? No. Thats not the point.
False advertising is:

 

So when you spent a grand on these cards, and a screen to utilize them what did you base your decision on, reviews and/or benchmarks perhaps ?

If so what has changed since you found out you've got cards with fewer ROP's and less L2 cache, has knowing what we now know magically changed those benchmarks and reviews ?

Care to provide a link to where Nvidia state how many ROP's or how much L2 cache the 970 comes with.

 

It's not the first time that advertising got some technical detail wrong, and I doubt that ROPs actually mean anything more to people than some abstract "more = better" number.

 

Greetings!

This was what nVidia sent to the reviewers:

Accidental nVidia internal miscommunication between departments or on purpose, same effect.

 

Did any of the GPU manufacturers actually put ROP/L2 spec on their spec sheet? Nvidia doesn't even put it on their website, you need to dig into datasheet to get to it.

GPUz misreporting something is nothing new, some of the metrics its reporting are based on guesswork and/or internal database (some older versions are reporting 32 ROPs on 970).

Intel disabled TSX instructions in their CPUs last August, was that also false advertising? Why noone seemed to care then?

Hell, then dealing with ICs these days you usually need to dig through errata longer then an actual datasheet (or errata for that datasheet) to find out whats actually operational/usable.

 

But is it advertising?

Everything on this page is advertising, and everything is correct from day 1.
http://www.overclockers.co.uk/showproduct.php?prodid=GX-259-MS


edit: what Xlog said.

 

And the effect is...?