GeForce Kepler 104 (GK104) Packs 256-bit GDDR5 Memory Bus, 225W TDP

[btarunr]

NVIDIA GeForce Kepler (GK104) will be NVIDIA's first high-performance GPU launched, based on its Kepler architecture. New reports suggest that this GPU, which will succeed GF114 (on which the likes of GeForce GTX 560 Ti are based), will continue to have a 256-bit wide GDDR5 memory interface. An equally recent report suggests that NVIDIA could give the front-line product based on GK104 as much as 2 GB of memory. We are also getting to hear from the INPAI report that on this product based on the GK104, the GPU will have a TDP of 225W. What's more, NVIDIA is gunning for the performance crown from AMD Radeon HD 7900 series with this chip, so it suggests that NVIDIA is designing the GK104 to have a massive performance improvement over the GF114 that it's succeeding.



View at TechPowerUp Main Site

 

[Arrakis9]

heres hoping for another ATI killer

 

[eidairaman1]

heres hoping for another ATI killer

zzzzz

 

[Wyverex]

I'm not sure what to think of this.
If Nvidia's "middle" card would to end up as powerful as AMD's top cards - that would make for a serious advantage for Nvidia's flagship cards. Which would in turn make those cards terribly expensive.
On the other hand, it does bring back memories of HD 4800 cards slacking behind GTX 200, but being super cheap and good buys.

Bring on the competition!

 

[1c3d0g]

So if I'm understanding this correctly, NVIDIA's mid-range GPU is supposed to compete or even topple ATI's high-end?!? Now this is what I'm talking about! Let the games begin!

 

[HTC]

heres hoping for another ATI killer

I hope not.

Neither ATI nor nVidia are allowed to have a significant lead one over the other because, if one of them do, there will be no price wars which will mean pricier cards for us consumers.

 

[reverze]

good news for those few people who still buy Nvidia

 

[entropy13]

heres hoping for another ATI killer

ATI's already dead. Killed by...AMD.

 

[EastCoasthandle]

Don't get caught up in the marketing. The 560 ti (GF104), a 40nm part, has a tdp of just 170w. What they are suggesting for the GK104 is a tdp of 225w for a 28nm part! Something is not right with that, to say the least. Even if it's implied they are trying to overclock a 560ti replacement for a mid/high part.

 

[Shihab]

good news for those few people who still buy Nvidia

"few" ?

 

[entropy13]

What they are suggesting for the GK104 is a tdp of 225w for a 28nm part! Something is not right with that, to say the least.

GTX 580 (GF110) has a TDP of 244W for a 40nm part. If the GK104 is indeed a challenge for the performance crown (and therefore is matched with the HD 7970) then a slight decrease in the TDP (225W v. 244W) with an increase in performance of at least roughly 20% (to make it match against AMD's 7970) is "not right"?

 

[Completely Bonkers]

GF104 to GK104.

Sounds like a revision to me, not a new design. And 225W is enormous. Shrink to 28nm and new design should result in similar performance at half the power, ie 560Ti performance at 90W, or similar power envelope 150W but double to triple the performance.

Somehow 225W seems all wrong, unless, they are aiming at performance of 4x, which I would say is physically and technically impossible unless their transistor count has gone through the roof, and this chip is the size of a football pitch! (And essentially SLI on-chip).

 

[Zubasa]

Enough talk nVidia just release the damn thing.
People can see for themselves, right now its all speculation.

 

[entropy13]

Enough talk nVidia just release the damn thing.
People can see for themselves, right now its all speculation.

Nvidia have actually yet to say anything. But in your second sentence you imply that, stating "its (sic) all speculation." lolwut?

 

[EastCoasthandle]

I'm not sure I believe this. The reason why I didn't suggest a higher transistor count was because if that were true why is it still limited to just a 256-bit memory bus? If that were the case and they know the GK100 (or whatever they want to call it) is not ready then why not go for memory bus higher then 256bit? Lets be honest, with a tdp of 225w (55w higher then the GF104) do you think they would care about efficiency at that point?


To me, going from GF104 to GK104 with a 55w tdp increase and still using 256bit just doesn't add up for me.

 

[entropy13]

F**kin' hell people here are taking "which will succeed GF114" too literally, when all it could mean is that the GF114 was their latest chip they made, and thus it's simple logic that a new chip (GK104) would obviously be "succeeding" its predecessor.

 

[Steevo]

Perhaps its another Fermi make and bake oven?

 

[iLLz]

GF104 to GK104.

Sounds like a revision to me, not a new design. And 225W is enormous. Shrink to 28nm and new design should result in similar performance at half the power, ie 560Ti performance at 90W, or similar power envelope 150W but double to triple the performance.

Somehow 225W seems all wrong, unless, they are aiming at performance of 4x, which I would say is physically and technically impossible unless their transistor count has gone through the roof, and this chip is the size of a football pitch! (And essentially SLI on-chip).

The F you bolded = Fermi and the K you bolded = Kepler, so I do believe this would be a new design. My guess is they kept the numbering scheme similar for easier comparison which is a welcome move considering the past transgressions.

And if the power numbers are to be believed, then I would think this card to be very fast. I thought Nvidia released slides that said Kepler and the subsequent releases would double or triple speed. Maybe they are following through on this.

 

[EastCoasthandle]

If this were true and the GK104 was challenging the 7900 series wouldn't it have been displayed or told at CES ? As far as I know nothing was mentioned about their next gen parts.

 

[Fourstaff]

Ah, the predicted "but wait, we have something too" response from Nvidia. I wonder how good Kepler is going to be, given the massive power consumption.

 

[Vancha]

If that's the TDP despite moving to the 28nm process, there's no excuse for this not being monstrous.

 

[DarkOCean]

Isn't 225w tdp a little scary? gtx570 has official tdp of 215 and in reality according to wizz reviews its 100w over that!

 

[Stephen.]

I'll believe it when i see it. a TDP of 225 on 28nm GK104 damn that sounds wrong, compared to 170 on 40nm GF104/114. it should be lower TDP and more performance or equal TDP and more performance.

Don't they see what Intel does with their processors.

 

[phanbuey]

excited to see this card... NV cards game so smooth.

 

[EastCoasthandle]

If that's the TDP despite moving to the 28nm process, there's no excuse for this not being monstrous.

Going from 170w using 40nm to 225w using 28nm does not necessarily mean monstrous performance. It could be the opposite.