Nvidia BigK GK110 Kepler Speculation Thread

nvidia likes number 80 very much, or 8 in general. A pride number, like NV38 and NV48.
GK200, or GK210 or GK120 would imply there's an architecture or process difference, but there is not, it's either the same thing or very slightly tweaked.
 
They could have launch this Atlas K40 at the same time of the K6000..
It really makes no sense to launch a bunch of new products at the same time, especially in the professional world where budgets needs to be made months ahead of time and where you don't have to deal with entitled children who want the latest and greatest NOW. ;)
 
acjjmlzuc.jpg


by expreview forum
 
Is there particularly bad scaling (as number of SMX's increase) to warrant a 10 SMX part that's boosts to 1200Mhz?

Had anyone done any analysis to show NVidia SMX and AMD CU scaling?

Or is this just a filler between 770 and 780?
 
Last edited by a moderator:
Is there particularly bad scaling (as number of SMX's increase) to warrant a 10 SMX part that's boosts to 1200Mhz?

Had anyone done any analysis to show NVidia SMX and AMD CU scaling?

Or is this just a filler between 770 and 780?

It looks more like a replacement for the 780. If those numbers are correct it has the same bandwidth as a 780 with 10% more flops and texturing and 33% higher peak fillrate. It should actually beat the 780.

Interestingly it has the same texturing and shading rate as Titan and 37% more fillrate. Sounds too good to be true unless nVidia found a way to get more clocks out of GK110 without a dramatic bump in voltage and power consumption.
 
Is it possible that the 780i in this GPU-Z screenshot is overclocked (from the 1033 MHz in the other screenshot) and GPU-Z just doesn't know the base clock?
 
This might be NVIDIA's way of bringing Titan-ish performance down to ~$600 without damaging the Titan brand.
 
This might be NVIDIA's way of bringing Titan-ish performance down to ~$600 without damaging the Titan brand.

They have to do something. It looks like AMD is aiming to provide similar performance at a significant discount to nVidia's offerings.
 
Why are the checkmarks missing from the CUDA and PhysX boxes?

Probably just drivers.
nVidia is extremely.. picky regarding the activation of features in their GPUs.

Have an AMD GPU present in the system, even if the GeForce is selected as main graphics card? PhysX misteriously disappears from the list of features. Sometimes, even CUDA goes away.
 
In reference to the "GK180" discussion on page 16 of the Maxwell thread (which I think more properly belongs here), this Newegg interview is with NVIDIA's Quadro Product Manager who mentions that the Quadro K6000 is based on a GK180.

(Thanks AnarchX.)

EDIT: I watched more of the interview and at around 2:37, he says "that's right" when the Newegg person asked, "so GK110?." I'm guessing that, assuming nobody misspoke, the GK110 and GK180 are closely related (which seemed to be the speculation anyway) and maybe the latter is a revision of or some sort of subset of the former.
 
Last edited by a moderator:
In reference to the "GK180" discussion on page 16 of the Maxwell thread (which I think more properly belongs here), this Newegg interview is with NVIDIA's Quadro Product Manager who mentions that the Quadro K6000 is based on a GK180.

(Thanks AnarchX.)

900Mhz fully enabled GK180 under 225w? If that's not a super binned part then it could be the basis for NVidia's high-end refresh.
 
Back
Top