[lanek]

hence why i have write " i know Nvidia have do it by the past ".. but not for this type of range and more likey on cards who was released for specific market and price, well at least not as a main "middle range card "...

Just cause it bring a lot of problem in term of efficiency and performance . If this is for differentiate them of the 670, there's different road to explore, starting by the clock speed, and if this is for disable a part of the core ( the memory controller ), starting by just disable 1 SM can be a good solution, specially it can even let you use chips who will have not been qualified for the 680-670..

1SMX disabled on the 670 just bring a decrease of performance of maybe 2% vs the 680 ( hence why you have some 670 overclocked as the Asus DirectCU II who are faster of the 680 stock with just some mhz more ).
Act on the memory controller for limit the bandwith and bring a difference with the 670 dont look like a choice an engineer could do . Its good for place a product in the middle of a range not on a card who will represent the top of his class ( the middle range class ).

I dont think there's many chips who come back from TSMC who have a problem with some memory controller parts, but im sure some can be used with one SMX less when they will not be qualified otherwise. so lasercut the memory controller just for make it a bit different of the 670, is just plenty useless and will even cost you an operation in addition of the production of the chips for nothing but be able to set it in the middle range instead of the "high range" ( so it will cost you money ) . At this point, why dont just make a 670, call it 660TI, a 900mhz turbo and sell it for 80-100$ less ?

If they release this card with thoses spec i can only start to ask me some strange question? have we allready seen a series from Nvidia so badly structured ? some 670 are faster of the 680, some 660 will be faster of the 670, some 650 faster of the 660 maybe ? looking last rumors, i can allready say the pcb of the 670 = 660TI... Is there any value to buy a 680 over a 670?, will it be the same for the 670 over the 660TI ?

[xDxD]

http://videocardz.com/33942/evga-gef...60-ti-pictured

let me see...from gk104 they have gtx 680, 670, 660ti, 680m...Who said anything about poor earnings?

[psolord]

What would be faster? A 670 minus one SMX, with 256bit bus or the current rumored 660 Ti with 1/4 ROPs and bus width disabled?

That being asked, what is the performance ballpark we are expecting from the 660 Ti?

[lanek]

Originally Posted by xDxD

http://videocardz.com/33942/evga-gef...60-ti-pictured

let me see...from gk104 they have gtx 680, 670, 660ti, 680m...Who said anything about poor earnings?

120W with 670 specification its just plenty impossible. I dont understand how the guy who write this article can have copy that.. maybe the 650 or 660 non TI.... but not this one. how do you think they can remove 50W from the 670?

[AnarchX]

http://bbs.expreview.com/thread-54008-1-1.html

120W TDP is probably a GK106 GTX 660 "vanilla", with rumored 980/1032/3000MHz.

[xDxD]

Originally Posted by lanek

120W with 670 specification its just plenty impossible. I dont understand how the guy who write this article can have copy that.. maybe the 650 or 660 non TI.... but not this one. how do you think they can remove 50W from the 670?

i don't write that new

what I wanted to emphasize was that of a single chip have made ​​a multitude of products, too many

[lanek]

Originally Posted by xDxD

i don't write that new

what I wanted to emphasize was that of a single chip have made ​​a multitude of products, too many

lol, yes. this is true.

[lanek]

Originally Posted by AnarchX

http://bbs.expreview.com/thread-54008-1-1.html

120W TDP is probably a GK106 GTX 660 "vanilla", with rumored 980/1032/3000MHz.

Yeah i think some rumors mix different cards or info they have got are from different cards .

For example, some speak about a 660TI, with the same amount of SP of the 670, and with 980 / 1023mhz (turbo), a 192bit bus, 2gb ..

I can imagine the bus will have an impact, maybe the ROP too.. But seriously this card should end as fast of the 670 in some case, and maybe 3-4% slower in other case.. for 100$ less of the 670, many peoples will ofc buy this card over the 670, and peoples who want SLI 670, will even more jump on it.

I had said Nvidia will never let the middle range market to AMD, but at this point it start to be a bit ridiculous to have 3 cards in less of 10% performance window. And ofc we dont even speak about OC models.. (again some 670 are faster of the 680, and it could end the same way with overclocked models, you could even get some 660TI retail OC faster of a 680) a 1150-1200mhz Turbo 660TI should even faster of the standard 680). Good for 660 buyer anyway.

__________________________________

Posted by Olivion on XS:

To take with care ..

http://www.redquasar.com/forum.php?m...=8612#lastpost

English version

@ CTS-MOD NV 660ti leaked, we got the 256bit version
Power port is 6p 6p (offical one is 6p) TDP 120 ~ 130w
default @ 980/6000 BOOST up to 1123
SP1152

Roughly the same pcb as GTX670
Output 2 * DVI DP HDMI
3d mark11 X 27 **
Can be oc to 1.2ghz with air cooling easily

Anyway nvidia is a bit struggling on this, 192bits cannot beat 7870, 256bits is too powerfual will effect their own product such as GTX670.

More infomation is depands on our satuaion

The one leaked before is 192bits, cannot beat the 7870, so nv give it 256bits and removed one group of SP so it comes into 1152SP

Interesting, they said they got a 256bit version with one SMX disabled, cause Nv seems have decided to let down the 192bits version cause it was not beat the 7870.

[xDxD]

i think that a 660ti with 1152sp and 192 bit (or 256 too, i don't think is enormous difference) could be more balanced not only towards 7870, but also gtx 670

[Homeles]

Disabling 25% of the bus results in a ~25% reduction in performance? Ouch.

[Blazkowicz]

bandwith is crucial.
but simply on the market, 1.5GB is not too appealing when you have 2GB cards, over and under and at the same level as it. the number is lower so it doesn't feel like a very good card.
maybe you can have both versions including a cheaper one, as with variants of the gtx 460.

[lanek]

Originally Posted by Homeles

Disabling 25% of the bus results in a ~25% reduction in performance? Ouch.

It seems a little bit too high.. The problem can result of the fact the chips is not made for it.

[xDxD]

http://videocardz.com/33976/nvidia-g...tion-confirmed

make a lot more sense to me

[lanek]

Originally Posted by xDxD

http://videocardz.com/33976/nvidia-g...tion-confirmed

make a lot more sense to me

Agree... anyway lets wait they release them, i think we will know soon enough

[Alexko]

Originally Posted by xDxD

http://videocardz.com/33976/nvidia-g...tion-confirmed

make a lot more sense to me

That would put the 660 Ti somewhere between the 7870 and 7950. Now let's see how much it costs.

[jimbo75]

Something doesn't quite add up with that. I'd expect the higher clocks and especially boost clock of the 660 Ti to have a higher TDP than the 670 even taking the missing shader block into consideration.

[lanek]

Originally Posted by jimbo75

Something doesn't quite add up with that. I'd expect the higher clocks and especially boost clock of the 660 Ti to have a higher TDP than the 670 even taking the missing shader block into consideration.

You will never find a 670 who really work at 980mhz .. they all run at 1054+ mhz.. ( minimum ) ... ( reference cards used in press review was nearly all run at 1084+ mhz in all games ). And ofc the 1033mhz can be from this particular card.

[jimbo75]

Originally Posted by lanek

You will never find a 670 who really work at 980mhz .. they all run at 1054+ mhz.. ( minimum ) ... ( reference cards used in press review was nearly all run at 1084+ mhz in all games ). And ofc the 1033mhz can be from this particular card.

Yes but I'm assuming the same applies to the 660 Ti. The clock increase is enough to make me think there is a voltage increase there.

Can 15% higher clocks and 15% less shaders also lead to 15% lower TDP? The really annoying thing about this is, Nvidia can seed the absolute best of the chips (ie chips that would normally make 680's) to the press as "660 Ti's", when in reality the average 660 Ti chip could be an awful lot worse.

[RedVi]

With 1152SPs surely they could make a specialized chip in a few months time with a smaller die. These specs seem to make sense to me.

[lanek]

Originally Posted by jimbo75

Yes but I'm assuming the same applies to the 660 Ti. The clock increase is enough to make me think there is a voltage increase there.

Can 15% higher clocks and 15% less shaders also lead to 15% lower TDP? The really annoying thing about this is, Nvidia can seed the absolute best of the chips (ie chips that would normally make 680's) to the press as "660 Ti's", when in reality the average 660 Ti chip could be an awful lot worse.

Yes but we dont know from what card is coming thoses GPU-Z screenshots, could be a OC models send from an AIB. On the other way, Nvidia could have set higher turbo clock speed, but this time the 1033mhz is the average, not the minimum "warranty" . Dont forget the turboclock is a good way to set a maximum tdp but trying to get the max performance within this limit.

[xDxD]

Nvidia's gpu on next xbox? http://www.eurogamer.net/articles/di...go-devkit-leak

[eastmen]

Originally Posted by xDxD

Nvidia's gpu on next xbox? http://www.eurogamer.net/articles/di...go-devkit-leak

meh the news writer brought up 8 cores having to be based on atom. Once that happened it lost all credibility.

[gongo]

http://www.facebook.com/photo.php?fb...count=1&ref=nf

660Ti dat fast on 3DMark11...damn!

[xDxD]

http://wccftech.com/nvidia-geforce-g...rates-hd-7950/

[xDxD]

@gongo: amazing timing!