NVIDIA Kepler speculation thread

Could it be 8800 Ultra?

 

Rather pointless argument when you can buy a console for the money you can spend on an high end GPU, and still play the same games

Note: I have a GTS450, which I choose over the HD5770 because of brand loyalty and PhysX So If I would be biased at all, it would be towards nVIDIA But I prefer to face reality

 

So if Fermi is a 9, kepler seems to be a 15.8! Nvidia seems to be promising a 75.55% performance increase over Fermi! I will eat my hat if true!

I really do hope this is true though. That means the lower specced card's, ie the X70 or X60 parts, could possibly deliver my personal much desired +50% performance increase compared to my 570s, in order to upgrade and possibly at a reasonable price, since they will not be the top dogs.

 

And yet you think they might be doing a console comparison to somehow sell more PC GPUs than AMD? Consoles are a major source of competition for both nVidia and AMD, that chart won't look any different if it used AMD GPUs.

In general the current state of affairs for PC gaming is pretty miserable. We get excited for crap that really doesn't improve our gaming experience at all - a few more fps, IQ settings that you need a magnifying glass to notice. Sadly there's nothing AMD or nVidia can really do to change that. nVidia tried with PhysX and got crucified for it. AMD brought Eyefinity which is great but still very much a niche feature both in terms of hardware and software adoption. We need better games and better input devices on the PC far more than we need new GPUs.

 

Then the questions is back at you
Why are they doing this then?

 

yeah this is first series of 28nm so we can expect +15-20% for mid-life kicker..

i would put GK104 to 11.75-12.25x area.. GTX560ti is 30% faster and yet consumes less power than GTX285

 

I already answered that question.

 

The Kepler bar is almost exactly 2x GTX 480. Would be nice.

 

That's not very impressive. The graph is based on theoreticals, not measured performance.

 

Well the other bars seem to be somewhat properly aligned for real gaming performance also, but yeah better to wait and see how things turn out.

 

Here's another slide to look at:



Looks plausible to me.

 

Is there any reason we keep posting unofficial slides that anyone with half a brain and Excel can create?

 

Lack of anything official perhaps? Hopefully that will change today.

 

One thing I've learned about nVidia over the years is they are very good at keeping quiet when a product is going to be really good. Very little was leaked prior to GTX5xx, 8800gtx, GTX280, etc. They were hard launches and surprised everyone.
Now... if you guys remember the 6 months prior to GTX4xx.... when TSMC was having horrible yields and nothing was going right... all sorts of hype, BS, and excuses were coming from nVidia.

Staying silent is also good for your third party partners that have lots of product on store shelves and would like to sell some of it before you make an announcement that causes any potential buyers to hold off spending money.

 

Im worried that there is something terribly wrong with Kepler because AMD 7000 is practically out and we havent heard a peep from NV. CES was the ideal venue to unveil it, what other event could it show up at? Is there anything before E3? And E3 2012 is (likely) going to be an orgy of new hardware debuts so that would be a terrible venue because it would get lost in the crowd.

I don't know, it just looks ominous. I was/am really dissatisfied with my HD5850 (shitty drivers, hardware defect causes BSOD when playing video files) and desperately want to go back to NV but its time for an upgrade and I dont want to wait a whole lot longer.

 

But likewise, Nvidia have also been the kings of running spoilers against competitor launches, but we've not had a peep out of them for the AMD 7xxx launch.

 

Somewhat yes. Would that imply 2 dual-chip cards? One with 2 8-pin and one with 3 8-pin connectors or what?
The fastest single-chip card would not be all that fast if those scores were for real (accounting for physx bonus ~20% over HD7970), and probably more importantly if nvidia is going for big chip again the second fastest chip (which would thus likely be a bit smaller than Tahiti) would have no chance (most likely) catching a HD7950.

 

next event is Cebit March.. then GTC May

 

It seems like the launch of Kepler should be prior to GTC May. Maybe just at the Cebit March