Geforce Titan - GK110 will be a consumer part

[BenchZowner]

Hold your horses until the 25th ( of Feb )...

[zalbard]

yeah right. the supercomputer named "titan" uses $8000 boards with 14 smx, and nvidia will kindly sell us 15 smx boards for $800. that would be nice of them.

That was months ago. They could have refined the process since then.
Besides, they could've just needed to get those chips made NOW, with little time left for binning.
Or it could be that they disabled one SMX because of power issues similar as they did with GF100 (and it actually is a big issue for supercomputers).
It is a possibility.

[Rollo]

The most important thing - except the possibility to watercool it - is to freely change the Voltage!

http://www.youtube.com/watch?v=sZfZ8uWaOFI

[Rollo]

yeah right. the supercomputer named "titan" uses $8000 boards with 14 smx, and nvidia will kindly sell us 15 smx boards for $800. that would be nice of them.

AFAIK the commercial chips costing thousands have always been basically the same as the much cheaper consumer parts, with difference being drivers.

[bamtan2]

That was months ago. They could have refined the process since then.
Besides, they could've just needed to get those chips made NOW, with little time left for binning.
Or it could be that they disabled one SMX because of power issues similar as they did with GF100 (and it actually is a big issue for supercomputers).
It is a possibility.

it is definitely possible. there is just no reason for them to do this IMO. it would be only kindness on their part! I want a perfect GK110 too, but I'd be surprised if K20x isnt good enough for everybody involved.

[zalbard]

The most important thing - except the possibility to watercool it - is to freely change the Voltage!

http://www.youtube.com/watch?v=sZfZ8uWaOFI

This made me lol, thanks!

[tbone8ty]

NVIDIA GeForce Titan With 6 GB Memory and 235W TDP





http://wccftech.com/nvidia-geforce-t...mples-shipped/

http://www.chiphell.com/thread-665046-1-1.html

[hurleybird]

That's K20, not titan.

[Cordell Cahill]

Yep, with it's lack of video ports

[tbone8ty]

oh ok my bad...but will it look like that?

[SKYMTL]

http://wccftech.com/nvidia-geforce-t...mples-shipped/

http://www.chiphell.com/thread-665046-1-1.html

Some sites are really blowing this all out of proportion with false information.....

[tbone8ty]

lol tell that to a fanboy

[tajoh111]

I wonder if there is a chance titans not being released at all and this really is more so the official release/reviews of tesla/Nvidias pro line. I am surprised there hasn't been any news of the professional line yet. And Nvidia might as well release reviews on the professional line to sell to other customers besides there super computer orders. It makes alot of sense since they can still sell these cards for 4000 dollars or so, rather than sell them as discrete gaming chips.

[Zloyd]

Maybe they are gonna sell most of the GK110 chips as workstation cards and the cherry picked higher clocking chips will go into Titan ?

[Darakian]

Maybe they are gonna sell most of the GK110 chips as workstation cards and the cherry picked higher clocking chips will go into Titan ?

If anything it would be the other way around. Workstation cards carry a higher price tag and thus higher margins.

[AffenJack]

But Workstation parts, atleast Quadro had always not so many Shader Cores, that was also the case with GF100. In Quadro you don't need more Shaders, other things like high geometry throughput is more important. For Tesla more Shaders might help a bit, but you might also be very memory bandwidth limited and if you sell a card with 15 SMX in the professionel space it has to be available good. For Titan at 900$ you just can save the few 15 SM parts you'll get and sell them, while the availability will be pretty bad.

[Lanek]

Maybe they are gonna sell most of the GK110 chips as workstation cards and the cherry picked higher clocking chips will go into Titan ?

Otherway around, They cherry pick the best Asic quality / chips for Tesla and workstation parts, and let the rest for the gamers.. Specially with turbo, they have more flexibility for binn the cores. ( 4 different types reported for the 680 based on Asic quality and who was determine the turbo clock speed. )

But Workstation parts, atleast Quadro had always not so many Shader Cores, that was also the case with GF100. In Quadro you don't need more Shaders, other things like high geometry throughput is more important. For Tesla more Shaders might help a bit, but you might also be very memory bandwidth limited and if you sell a card with 15 SMX in the professionel space it has to be available good. For Titan at 900$ you just can save the few 15 SM parts you'll get and sell them, while the availability will be pretty bad.

They hit the same problem as allways: TDP and cooling ... Tesla cards are constrained by TDP ( max 235W and no turbo ) due to the specification for Cray and other HTPC system makers ( dont forget, there's 2 GPU and
4CPU on a rack ). For computing parts, it was more important to get the Higher amount of Shader instead of speed. they have balance it, do a choice.. Same goes for the memory as ECC remove nearly 1Gb of the 6Gb available. They have do a balance between TDP/shaders/core speed.

I will not say for gaming shader really do a difference, just look at the 670-680.. 1 SMX less and the cards at same clock perform really close and with the higher count of SMX we could seen, im pretty sure one SMX shader do even less difference. specially if you can go to 250W + Turbo for use the difference on Power usage we see in games and save some precious mhz.

Anyway we will know soon enough, as it seems some reviewers have the cards right now in hand.

[zalbard]

http://wccftech.com/nvidia-geforce-t...mples-shipped/

What a useless article. Absolutely nothing new.

If anything it would be the other way around. Workstation cards carry a higher price tag and thus higher margins.

Workstation and server cards have higher power and heat constraints.

[Olivon]

What a useless article. Absolutely nothing new.

Additionally, review samples of the card have already been shipped to various tech blogs including PCPer and TechReport as revealed by News Editor of Rage3D Caveman-jim.

That's new and nobody has posted it on XS. Only thing useless I see is your comment.

[Lanek]

That's new and nobody has posted it on XS. Only thing useless I see is your comment.

huum, the comment from Caveman-Jim is on all sites and forum since yesterday,

[BenchZowner]

Who cares... unless you missed it:

25th of February...

[Tha Last Meal]

yo guys
The result that appeared on 3dmark11 program (GTX 780) was fake!!

look here


http://www.techpowerup.com/179605/Fi...-Revealed.html

and look here



http://www.arabpcworld.com/?p=25899

The result not for GTX 780 it is for GTX 690!(Focused in the picture)


and i see in this Site arabpcworld Other result for GTX 780...do you think fake or not?

http://www.arabpcworld.com/?p=25911

[zalbard]

Hopefully Nvidia changed power connector placement. I don't like the way K20X is set up.

That's new and nobody has posted it on XS. Only thing useless I see is your comment.

That was posted all over the place several days ago (Feb 1). And it still isn't the original source; should be Rage3D instead.

Who cares... unless you missed it:

25th of February...

Isn't that too much time to test a card? 20+ days?

[btarunr]

and i see in this Site arabpcworld Other result for GTX 780...do you think fake or not?

http://www.arabpcworld.com/?p=25911

Hey, we make GPU-Z, and...

[zalbard]

Does anyone have any idea how much power 6GB of GDDR5 consumes? I wonder how big of a clock speed increase G110 could receive if TDP was increased to 300W...