NVIDIA Kepler speculation thread

Wow... regardless of the "value" of the 780, the relative value add on that EVGA card for $10 is incredible. You are getting a ~15-25% performance bump for a 1.6% increase in price.

Agreed! :smile:
 
Sounds like a OC'd 7970, even smells like it. From TechSpot review:

7970 GHz Edition in Detail

http://www.techspot.com/review/546-amd-radeon-hd-7970-ghz-edition/

Most other reviews state the same. Don't know what you're smoking but I don't think I want any.:LOL:

I suppose you thought the Radeon 4870 was just an overclocked Radeon 4850 then? They are the same with only the clock and memory speeds being different. I suppose the Geforce 6800 Ultra was just an overclocked Geforce 6800 GT.

Of course, let's ignore the fact that they are all standard reference design clocks and memory speeds on those. And that you can overclock them all individually.

The 7870 Ghz edition isn't an overclocked version. If we used old Nvidia terminology you'd have the 7870 GT (normal) and 7870 Ultra (Ghz). Or if it was old ATI terminology you'd have the 7870 (normal) and the 7870 XT (Ghz). None of which makes any of the cards I just mentioned an overclocked card.

Overclocked cards are those that are running at speeds that are different from the stock/reference clocks. Considering that the 7970 Ghz edition is the reference clock, then it is obviously not an overclock edition.

For an overclocked 7970 Ghz edition look to something like the Gigabyte GV-R797TO-3DG (who comes up with these names) which has a listed feature of "Core clock OC to Base 1050/Boost 1100 MHz" or any of a number of actual overclocked 7970 Ghz edition cards.

Regards,
SB
 
In addition .. the 7970GHZ introduct Turbo from bios ( PTE come from driver ).. Like the TDP board is a lot different of real TDP, its clear the card boost all time at 1050mhz ( AMD have fix the max turbo to 1050mhz ( who at my sense was an error, when looking the TDP margin and voltage increase )) .. ( the 7970 was vary between 150 and 195w tdp for a TDP Board of 250W ).

Is this is resulting to the same level performance of a 7970 overclocked ? Is it just a 7970 overclocked no .

But like said SilentBudda, if you want compare an overclocked card vs an overclocked card, just compare a 7970Ghz or not use a MSI lighting, Asus DCII, or Sapphire, HIS vs an overclocked 680, 780 ..
 
Last edited by a moderator:
browsing the HEXUS review..it tells me with the GTX Titan(aka GK110)...Nvidia comprehensively beat down AMD in the 28nm GPU designs this round... It has the TFlops compute power...the low heat, power and noise signatures...a little sad that AMD failed to continue their success after the HD4,5,6 series....well the HD6 were already struggling a little.....

I wonder if the time for AMD to re-consider their small die thingie....must be sucks to see Nvidia getting away with the GTX7 series unchallenged..
 
AMD don't really have a small die strategy anymore. They can't go as big as Nvidia without the professional market presence to pay for the poor yields and expensive large dies.

[Fairly OT]I think they need to beef up their mid range somewhat to compete with Gx104 at its own game - gaming performance only. The problem with doing this right now is it would eat into their upper end, more compute centered Tahiti based products. So they need to significantly improve them also. For next gen (20nm), they should probably release the x800 series first, perhaps with over 2048SP's, high turbo frequencies and 7Gbs GDDR5. Then they can release the halo 7900 series products a bit later. No, they shouldn't go for 550mm^2 dies, but something around 450mm^2 is probably a good idea to compete with Nvidia in the same class. This way, they don't have to pit a larger, compute enabled device against what is essentially a beefed up, high clocked mid range part.[/Fairly OT]
 
Last edited by a moderator:
browsing the HEXUS review..it tells me with the GTX Titan(aka GK110)...Nvidia comprehensively beat down AMD in the 28nm GPU designs this round... It has the TFlops compute power...the low heat, power and noise signatures...a little sad that AMD failed to continue their success after the HD4,5,6 series....well the HD6 were already struggling a little.....

I wonder if the time for AMD to re-consider their small die thingie....must be sucks to see Nvidia getting away with the GTX7 series unchallenged..
Are you trying to derail this thread?
 
AMD don't really have a small die strategy anymore. They can't go as big as Nvidia without the professional market presence to pay for the poor yields and expensive large dies.
With enough redundancy (GTX780), a very mature process (hello 28nm), and a $650 retail price, nothing prevents you from making a 500mm2+ consumer die a very profitable affair.

Think about it: GK104, 300mm2 give or take, has been selling as a perfect right from the start. Nobody is going to convince me that a die that's less than double the size that can withstand a defect in at least 3 of its SMs won't have a similar yield.

Ask $200 or more for the silicon and watch the money rolling in.

The higher SKU dies (Titan, Tesla) are a mountain of gravy.
 
So according to this Nvidia now have the 3 fastest single GPU's on the market. It's really time AMD stopped messing around and replaces Tahiti.
 
So according to this Nvidia now have the 3 fastest single GPU's on the market. It's really time AMD stopped messing around and replaces Tahiti.


Look the games used, for most, the 680 was allready equal to the Ghz edition in thoses one or really close and with AC3, BL2, Batman, even faster in thoses cases. ( + it seems they have use strange setting, you will never get this fps with a 780 / 7970Ghz in FC3 untill you drop completely the AA ( 15fps difference with review of the 780). Depending the review, the games used, it should end at something like 5-6% average...

its not hard to match and even get higher performance of a card who have been released 1 year ago. just set as high the clockspeed you can ( 230w TDP ) and try get the most of it, you know where you need to aim. Nvidia will be completely stupid to dont success to have some few % faster of the old AMD card.
 
Memories at 8,2ghz (qdr)!!

http://abload.de/image.php?img=223225hiiedy8ds1eh71jrtugc.jpg

Sorry i haven't original link. This is MSI lightning, so 1300mhz on GPU it's no amazing in my opinion, but memories are the same for every 770 and so 8.2ghz it's a great value

Wow if true.... hahaha I remember when the gtx680 came out and some tech sites were saying GDDR5 memory controllers were running at their near theoretical max at 6.0 Gbps.
 
Memories at 8,2ghz (qdr)!!

http://abload.de/image.php?img=223225hiiedy8ds1eh71jrtugc.jpg

Sorry i haven't original link. This is MSI lightning, so 1300mhz on GPU it's no amazing in my opinion, but memories are the same for every 770 and so 8.2ghz it's a great value

That could be on LN2 or other such sub-ambient temperature cooling.

Wow if true.... hahaha I remember when the gtx680 came out and some tech sites were saying GDDR5 memory controllers were running at their near theoretical max at 6.0 Gbps.

I run my reference GTX 680 from Asus at 7GT/s and have since day one, even on the original air cooler, let alone under water as I have it now.
 


Some features of DX11.1 are hardware and software ( only 4 require hardware ), so any DX11.0 cards will be compatible but only by software. ( allready discussed with GTX600 series, as AMD support hardware DX11.1 and Nvidia just by software ).

And like allways the words of Nvidia is " thoses features are useless anyway ".. ( who is not true but thats an other story )
 
Back
Top