NVIDIA Kepler speculation thread

720mhz being the boost clock ('up to').

So gtx670 should have 50% faster core and 57% faster memory ... I somewhat doubt the performance claim, but ofcourse it may perform closer to the 720 mhz than 670 does to the 1080 most of the time.
I don't see it neccesary being <100W either - maybe in the original GK106 based design, but performance crown is more important..
 
Last edited by a moderator:
680m based on the same chip as the desktop 680? Now they're just rubbing it in our faces. 720Mhz is quite a drop in clocks though.
 
680m based on the same chip as the desktop 680? Now they're just rubbing it in our faces. 720Mhz is quite a drop in clocks though.

Taking in account it is similar of a GTX670 but with 4gb and 195mhz lower on core ( GTX670 is at 915 )...

So the GTX675M is still a Fermi revamped with 384cores 627mhz... i cant imagine the gap with this one.


I dunno, maybe it was more easy to decrease clock speed and play on turbo ( need see what is the initial clock speed and if 720mhz is the turbo ), instead of use lower CC counts and maintain an higher clock.
speed.

Well for the performance graph, need to see the res, the settings used ofc.
 
It should be 1920 X 1080. Everything maxed out. :D

80% faster than 580M is a really nice step forward. Big like and hope the other 600 series members will be placed accordingly.
 
720mhz being the boost clock ('up to').
Is that necessarily the case though, since the memory clock also says "up to" 1800 MHz, and unless I'm missing something, memory doesn't boost on Keplers. (Not that I'd be surprised if the base core clock was < 720 MHz.)
 
Is that necessarily the case though, since the memory clock also says "up to" 1800 MHz, and unless I'm missing something, memory doesn't boost on Keplers. (Not that I'd be surprised if the base core clock was < 720 MHz.)
"Up to" is notebook PR speak for the fact that OEMs commonly set their own clocks lower.
 
http://www.geforce.com/hardware/desktop-gpus/geforce-gt640

GT 640 DDR3 retail is pretty much starved for bandwidth, why won't Nvidia release the GDDR5 version?
Toms Hardware review is up: http://www.tomshardware.com/reviews/geforce-gt-640-review,3214.html
I've no clue neither why the ddr3 version is launched first. If nvidia doesn't have enough chips why not only launch the gddr5 version instead? Maybe nvidia went for the all-important shiny "2GB" sticker first instead of reasonable product... It should also do ok in the office I guess when all you need is more display outputs.
Performance for a chip with those capabilities (that is, based on the chip it should probably be slightly faster than gts 450) performance is abysimal, but OTOH for a card with so little bandwidth performance is pretty good. As expected power draw is ok too (in particular compared against the ridiculous gts 450 power hog) though supposedly perf/w would be better with gddr5.
 
Wanted to address a lower end market segment first?
Well AMD doesn't address this segment with Cape Verde at all. Granted to nvidia's credit they have a pretty big hole in their lineup between the anemic GF108 and the monster-sized (for this segment) GF116 so it makes sense they'd address that with gk107 instead of some downcut GF116 whereas AMD just can sell turks instead of ddr3 variants of Cape Verde.
Maybe nvidia would have launched the gddr5 version first if it could compete with hd7770...
 
GTX 560 Ti 820MHz @170W -> GTX 580M 620MHz @ 100W : 25% clock reduction
GTX 670 ~1000MHz @170W -> GTX 680M 720MHz @ 100W : 28% clock reduction

btw.
Is there any information, that GTX 680M is a 100W part?
 
Back
Top