NVIDIA Kepler speculation thread

[Mianca]

As discussed, we cant really compare GK104 with Tahiti, the right comparison would be with Pitcairn. And though the gap has shrunk, AMD still leads in density.

I think that's a very valid statement. Tahiti has a lot of additional GPGPU ballast and (due to time-to-market priority) wasn't even allowed to show its full potential yet.

That being said: what kind of performance lead over Pitcairn would GK104 need to be regarded as "impressive"?

If those numbers posted above are correct, GK104 seems to turn out about 50% bigger and 50% more power hungry than Pitcairn. So it would have to turn out 50% faster, too?

 

[Picao84]

I think that's a very valid statement. Tahiti has a lot of additional GPGPU ballast and (due to time-to-market priority) wasn't even allowed to show its full potential yet.

That being said: what kind of performance lead over Pitcairn would GK104 need to be regarded as "impressive"?

If those numbers posted above are correct, GK104 seems to turn out about 50% bigger and 50% more power hungry than Pitcairn. So it would have to turn out 50% faster, too?

According to my calculations its far from being 50% bigger. The difference is only around 39%, altough power consumption is a bit closer to that, at 45%. And Ive seen somewhere claims of it being 25% faster than HD7970, on average. That would put it at around 40% faster than Pitcairn, right?

 

[Love_In_Rio]

I think that's a very valid statement. Tahiti has a lot of additional GPGPU ballast and (due to time-to-market priority) wasn't even allowed to show its full potential yet.

That being said: what kind of performance lead over Pitcairn would GK104 need to be regarded as "impressive"?

If those numbers posted above are correct, GK104 seems to turn out about 50% bigger and 50% more power hungry than Pitcairn. So it would have to turn out 50% faster, too?

I wish AMD launched a pitcairn-like Tahiti GPU!.

 

[Man from Atlantis]

GF114: GTX 670M/GTX 675M.

This roadmap says GTX 680M 4GiB in Q2 2012: http://forum.notebookreview.com/gam...600-series-not-just-fermi-37.html#post8379171

So its probably GK104 or GK106 with some memory tricks to bind 4GiB@192-Bit.

It could be the first time Nvidia could beat AMD in the core count: 1536 > 1280.

There might be 680M A and B, A is GK104 and B is GK106.. i think there is a 555M A(GF108) and B(GF106) like that.. but hell it's complicated lol:smile:

___

i dont know you believe in fairy tales but i like them.. 25% faster than 7970 without Turbo..



http://we.pcinlife.com/thread-1853062-1-1.html

 

[AnarchX]

According to my calculations its far from being 50% bigger. The difference is only around 39%, altough power consumption is a bit closer to that, at 45%. And Ive seen somewhere claims of it being 25% faster than HD7970, on average. That would put it at around 40% faster than Pitcairn, right?

HD 7870: 175W typical board power
GTX 680: 195W rumored TDP

I wish AMD launched a pitcain-like Tahiti GPU!.

It does not look like AMD reached the scalability of geometry- and raster-processing to do this.
And could be a 384-Bit 6Gbps chip smaller than 365mm²?

There might be 680M A and B, A is GK104 and B is GK106.. i think there is a 555M A(GF108) and B(GF106) like that.. but hell it's complicated lol:smile:

No indications through drivers of a Fermi 680M. 670/675M were know for months.

 

[Picao84]

HD 7870: 175W typical board power
GTX 680: 195W rumored TDP

I was going by the 127W on the table. If 175W it is, then GK104 is even more impressive.

 

[Love_In_Rio]

There might be 680M A and B, A is GK104 and B is GK106.. i think there is a 555M A(GF108) and B(GF106) like that.. but hell it's complicated lol:smile:

___

i dont know you believe in fairy tales but i like them.. 25% faster than 7970 without Turbo..



http://we.pcinlife.com/thread-1853062-1-1.html

In Battlefield 3?.

 

[AnarchX]

I was going by the 127W on the table. If 175W it is, then GK104 is even more impressive.

The reviewed cards were far below this 175W (159W Furmark with PT+20%). Maybe there are some retail cards, which will use this power range.

 

[Man from Atlantis]

In Battlefield 3?.

overall, according to nV numbers..

new AA method ''xx XXAA'', better than MLAA, 8xMSAA?

Spoiler

i see it's starting with M

http://bbs.expreview.com/thread-49888-1-1.html

 

[lanek]

FXAA 3.0 + xxMSAA ? or new CSAA ?

 

[Mize]

how much will that blur a texture?

 

[Love_In_Rio]

overall, according to nV numbers..

new AA method ''xx XXAA'', better than MLAA, 8xMSAA?

Spoiler

i see it's starting with M

http://bbs.expreview.com/thread-49888-1-1.html

It may be the rumoured Nvidia SRAA that implied some hardware prerrequisites. Anyway the texture is quite blurred... I hope the 25% improvement isn´t comparing 4xMSAA with this new AA method...

 

[trinibwoy]

SRAA doesn't require hardware, it requires game developer support. It's basically a less blurry, more accurate shader based AA.

That blanked out name seems to start with an M though.

 

[Mianca]

SRAA doesn't require hardware, it requires game developer support.

And game developer support requires design wins in the console market - or a hell of a lot TWIMTBP money

 

[Blazkowicz]

is there a word on compute features? some new efficiency, GPU multi-tasking?

 

[CarstenS]

It may be the rumoured Nvidia SRAA that implied some hardware prerrequisites. Anyway the texture is quite blurred... I hope the 25% improvement isn´t comparing 4xMSAA with this new AA method...

This is a manual photograph from an image that a beamer cast on a sub-perfect-surface.

 

[trinibwoy]

This is a manual photograph from an image that a beamer cast on a sub-perfect-surface.

And I still don't see the blurring

 

[Man from Atlantis]

And I still don't see the blurring

it looks yellowish with TXAA on

 

[trinibwoy]

it looks yellowish with TXAA on

Nice, they finally brought "Totally eXtreme" AA to market!!

 

[Man from Atlantis]

Nice, they finally brought "Totally eXtreme" AA to market!!

lol, i thought it was Trinibwoy approXimation AA