NVIDIA Kepler speculation thread

Blazkowicz said:
That doesn't rule out a respin, what if GK114 is just a silicon respin of the very same GPU and launched silently so to speak?
Click to expand...

It's written "In both Haswell notebooks two new graphics chips are used."
How do you interpret it?
And 30% performance difference between 680m and 680MX. Isn't it too much and what's the point of 780M with comparable performance to that of 680MX.

These nvidia names are so screwed up, I wonder if they themselves understand the "logic" behind it...
 
If the 780m is lower power consumption there would be plenty of point to it. Also OEMs like to put new bigger numbers on hardware every year.
 
Huum dont forget, the 680M is a 670 based chips .. who have been replaced then with a full GK104 680 ( in term of shaders, but with lower clock speed ) .. add a bit of core clock speed, and the fact it dont compare it with the 680MX ( 1536CC and 732mhz vs 1344CC and 720mhz) . and you got it. This card could well be just a 680MX with better boost clock ( with turbo 2 you have a check on temperature for increase the turbo boost, + Nvidia could have allready increase the base clock to 800+mhz )
 
Last edited by a moderator:
xDxD said:
How a mobile chip can do what desktop version don't?
Click to expand...

What do you mean mobile chip doing what desktop version don't?
GTX680 desktop is fully enabled GK104 and clocked a lot higher

Only thing that could prevent mobile-GK104 being fully enabled and clocked higher is power consumption
 
Kaotik said:
What do you mean mobile chip doing what desktop version don't?
GTX680 desktop is fully enabled GK104 and clocked a lot higher

Only thing that could prevent mobile-GK104 being fully enabled and clocked higher is power consumption
Click to expand...

And temperature.. and with turbo boost 2.0 linked to temperatures, you have a lot of variation possible.. basically the chip will reduce is clock speed when reaching some temps ( and so impact battery life simultaneously ( TDP ).
 
Kaotik said:
What do you mean mobile chip doing what desktop version don't?
GTX680 desktop is fully enabled GK104 and clocked a lot higher

Only thing that could prevent mobile-GK104 being fully enabled and clocked higher is power consumption
Click to expand...

excuse i don't speak english well...i mean that between gtx 670 and gtx 680 there isn't 30% difference

now, how can a full gk104 mobile, even with higher clocks than actual 680m (but never at livel of desktop version of course, because power, temperature ecc ecc), get 30% more performance?
 
xDxD said:
excuse i don't speak english well...i mean that between gtx 670 and gtx 680 there isn't 30% difference

now, how can a full gk104 mobile, even with higher clocks than actual 680m, get 30% more performance?
Click to expand...

No, the difference isn't 30%, but the clock differences are under 100MHz on core and nothing on memory side.

Now, going fully enabled on the core-side and pushing clocks up by ~100MHz (+13.9%), you'd probably get about 15% increase on performance alone, maybe push it bit more and get the GPU clocks (at least in boost state) to 900MHz, and you're already getting around 25% performance increases if the memories can keep up, if not, be a little less aggressive on the core clcoks and get the memclocks up too, and hitting 30%+ higher performance shouldn't be an issue - GTX680M's memories are really slow for modern GDDR5, only 900MHz.
 
A 700MHz QuadroK5000 (GK104 + 4GB) has an MSRP of $2249 and it won't even be launched with the other models? Boy those GK104s must be hard to make :rolleyes:
 
What i dont understand is why they have not launch thoses models before ( 4000 and under ).. I mean i was really think Nvidia have allready decline Kepler on lower cards of the K5000 .
 
Last edited by a moderator:
DSC said:
http://www.anandtech.com/show/6140/...erbased-quadro-k5000-secondgeneration-maximus

K5000 was launched and selling since last year, not sure what you are talking about.
Click to expand...

Because I apparently had missed it and read the latest Anand part selectively:

On a final note, unlike the K5000 this will be a hard launch, so K4000, K2000, and K600 cards will be available starting today. This goes for both retail add-in card sales and for pre-built OEM systems, with OEMs shipping systems equipped with NVIDIA’s latest Quadro cards as soon as today.
Click to expand...

lanek said:
What i dont understand is why they have not launch thoses models before ( 4000 and under ).. I mean i was really think Nvidia have allready decline Kepler on lower cards of the K5000 .
Click to expand...

They might even release some higher end Quadros based on GK110 later on. NV obviously didn't have endless 28nm capacities booked at TSMC, so they launched the <GK104 cores later on after they've served desktop and OEM markets in the meantime with those.
 
The Quadro 410 has been around for a while, at least.
K5000 was rare I think, not really that available I guess that's what they mean with "not a hard launch" for that one but now you can readily get it (assuming you need it, of course!)

Another reason for delaying a kepler Quadro family I guess : when GK110 Tesla and Titan were (completely) unavailable, you'd have been looking up at GF100/110 Quadro 6000 or 5000 as the top dog, from the point of view of your lowly K600 or K2000.
It's weird to obstensibly have the two gens competing with each other.
 
From Sweclockers: "Nvidia prepares revamped Geforce GTX 650 Ti" (original).

There have been a few rumors of a 660 SE between the 650 Ti and the 660 that was supposedly planned for a Q1/April 2013 release. Maybe they decided to upclock the 650 Ti instead. It makes me wonder two things:
  1. Will this 650 Ti have boost?
  2. Will there be more upclocked versions of existing 600 series cards in the near future, depending on AMD's future cards (Hainan, Curacao)?
 
Nvidia prepares Geforce GT 655

Geforce 650TI Boost Edition

Our birds from the Far East have whispered that Nvidia is working on a new SKU that might hit the market as the Geforce GT 655. The card is nothing more than a Geforce GT 650 TI boosted to higher clocks, with a 192-bit memory interface.

The plain Geforce GT 650 TI has 128-bit memory and the faster memory interface definitely means the faster overall performance. It simply brings more bandwidth.

We don’t know the clock but the GT 655 name sounds very plausible. The chip behind the card is most likely the GK106, but this is also something that we need to confirm. The card should be coming very soon and it sounds like Nvidia thinks that there enough of a gap between Geforce GT 650 TI and Geforce GT 660 to fill with a new SKU.
Click to expand...
http://www.fudzilla.com/home/item/30775-nvidia-prepares-geforce-gt-655
 
Also interesting:
NVIDIA_DEV.1001 = "NVIDIA D15U-60"
NVIDIA_DEV.1003 = "NVIDIA GK110 DT Bringup"
NVIDIA_DEV.1004 = "NVIDIA D15U-50"
NVIDIA_DEV.1005 = "NVIDIA GeForce GTX TITAN"
Click to expand...

D14 was Geforce 600.
GK110 relaunch as GTX 780/770?
 
You must log in or register to reply here.

Similar threads

A
RDNA3 Efficiency [Spinoff from RDNA4]
3 4 5
Replies
85
Views
9K
Bondrewd
Bondrewd
D
The Silent Hand of Driver "Optimization" (spinoff)
2
Replies
31
Views
3K
DavidGraham
D
F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
Shortbread
Nvidia is Now The Second-Largest Company in The US Market
Replies
1
Views
3K
Shortbread
Shortbread
T
GPU Stability testing for Nvidia cards - PowerSweep
Replies
0
Views
451
T2098
T
Back
Top