Nvidia BigK GK110 Kepler Speculation Thread

Interesting. I wonder what kind of workstation workloads require that much memory. Unless this is meant to be an all-purpose card, aimed at compute just as much as typical Quadro workloads.
 
Interesting. I wonder what kind of workstation workloads require that much memory. Unless this is meant to be an all-purpose card, aimed at compute just as much as typical Quadro workloads.

NVIDIA’s other big market for Quadro K6000, like Tesla K10 and K20, is the oil and gas industry, who use GPU products both to analyze seismic data and to display it. In this case the more that can be analyzed and the more that can be displayed at once, the more useful the resulting product is

What comes to my mind is a supercomputer like loads, Earth Simulator type should benefit quite a lot, no?

http://www.anandtech.com/show/7166/nvidia-announces-quadro-k6000

Is this a new GK110 GPU spin?

Should be, given the time frame of releasing the product as late as fall 2013
 
I was under impression that the Titan uses 225W or less itself but you're right its official TDP is 250W.
That's maybe just some extreme cherry picking, you can do it when the margin's so big. It costs more than a Tesla K20X. No idea if a spin is needed and how much it costs (does production just improves even with no respin?)

The memory is dumbfunding. Can you basically render a movie on it?, or at least visual effects / CGI scenes for television or indie production.
 
NVIDIA’s other big market for Quadro K6000, like Tesla K10 and K20, is the oil and gas industry, who use GPU products both to analyze seismic data and to display it. In this case the more that can be analyzed and the more that can be displayed at once, the more useful the resulting product is

Ah, all-purpose it is, then. I wonder if this is a sign that NVIDIA intends to drop Maximus.

What comes to my mind is a supercomputer like loads, Earth Simulator type should benefit quite a lot, no?

Sure, but a Tesla card would make more sense for that kind of thing, as they're designed for denser environments, typically with scalable coolers and few or no display outputs. And they're also quite a bit cheaper.
 
http://www.anandtech.com/show/7166/nvidia-announces-quadro-k6000

Quadro K6000 is fully enabled, all 2880 cores clocked higher at 900MHz and has double the memory of TITAN yet is only 225W TDP vs 250W TDP on TITAN. Is this a new GK110 GPU spin?
Possibly/Probably given the time the GPU has been in production, although I'd note that even though the Titan is lacking in core count compared to the K6000, its TDP must be taking into account the boost function. As Anandtech noted, the Titan is boosting close to 1000M
52LJVpd.jpg
 
Ah, all-purpose it is, then. I wonder if this is a sign that NVIDIA intends to drop Maximus.
No, Maximus is alive and well (I specifically asked about that). There's less of a need to use it with K6000, but you still have all the same reasons to use it with K6000 as you did the Fermi 6000: separating rendering and compute tasks to further improve performance.
 
No, Maximus is alive and well (I specifically asked about that). There's less of a need to use it with K6000, but you still have all the same reasons to use it with K6000 as you did the Fermi 6000: separating rendering and compute tasks to further improve performance.

OK, thanks. Do you have any idea how successful it is? And how people feel about it, as opposed to AMD's relatively unified approach with FirePros?
 
OK, thanks. Do you have any idea how successful it is? And how people feel about it, as opposed to AMD's relatively unified approach with FirePros?
Sorry, nothing in particular. Enough Maximus systems seem to be selling via OEMs that it has to be reasonably successful, but I couldn't tell you more than that. However AMD's approach does not preclude doing the same thing; with the right software packages you can split up workloads over multiple Firepros in exactly the same manner.
 
I think that's just binning.

Plus I doubt it has a Boost function, so it can operate near Idle-voltage (0,875 Volt on Titan) - maybe 0,925 Volt. That'll save a lot of power compared to running at 1.125 Volt or so for 1006 MHz.
 
Plus I doubt it has a Boost function, so it can operate near Idle-voltage (0,875 Volt on Titan) - maybe 0,925 Volt. That'll save a lot of power compared to running at 1.125 Volt or so for 1006 MHz.

I think that is a bit too optimistic. One of my Titans uses 1.062V for 902 MHz. And I have a pretty high ASIC, around 80%.
 
Sorry, nothing in particular. Enough Maximus systems seem to be selling via OEMs that it has to be reasonably successful, but I couldn't tell you more than that. However AMD's approach does not preclude doing the same thing; with the right software packages you can split up workloads over multiple Firepros in exactly the same manner.

Got it, thanks.
 
I want to see a Titan Ultra with 2880 core @1.0GHz clock speed (just like GTX 680) ,I think it will not exceed 250W TDP and I am sure it will rock!

By the time Nvidia really release such a card at an affordable price instead of some $6000 quardo, I expected you would have already been wanting a Maxwell.
 
This was expected after the Quadro K6000.. what worry me is all of this look like a reaction against AMD newcomers.. What interest me there, is to know more about this workload boost, ANSYS; AMBER.. ( if you do 2D/ 3D modelisation since 1991, you will understand why i ask this. ).

I hate when a slide is trying to make them look better with " oppurtinity and lead over the competition "... this said they are maybe in the right to say that.
 
Last edited by a moderator:
GK180 sku name is a bit strange... why not GK200 or 210 ? they do what they want anyway...

This was expected after the Quadro K6000.. what worry me is all of this look like a reaction against AMD newcomers.. What interest me there, is to know more about this workload boost, ANSYS; AMBER.. ( if you do 2D/ 3D modelisation since 1991, you will understand why i ask this. ).

I hate when a slide is trying to make them look better with " oppurtinity and lead over the competition "

GK2xx would imply a new generation. I think this is just GK110, or possibly a respin of it.
 
Back
Top