NVIDIA Names Fermi Architecture Successors

[btarunr]

At the GPU Technology Conference (GTC), an annual event hosted by NVIDIA, the company named the next two succeeding GPU architectures to Fermi (the current generation architecture on which are based GeForce 400 series GPUs). NVIDIA's next major GPU design change will come in the form of "Kepler", probably named after the German mathematician Johannes Kepler. The only concrete details about this architecture is that chips will be built on the 28 nanometer silicon fabrication process, and that going by the architecture's double-precision GPU compute performance per Watt represented on a graph, NVIDIA expects Kepler to be 4~5 times faster than Tesla, and over twice as fast as Fermi, again, at double-precision GPU compute performance per Watt.

Kepler is slated for 2011, though which part of the year will it be out (since AMD's answer to Fermi isn't far away), wasn't revealed. Looking much further away into the future, much like Intel mentioned Sandy Bridge's successor (Gesher) way back when unveiling Nehalem, NVIDIA talked about Kepler's successor slated for 2013. This one is called Maxwell, probably in honour of Scottish mathematician James Maxwell, with expectations of no less than three times the double-precision computation power per Watt of Kepler. These chips will be built on the 22 nanometer process.



View at TechPowerUp Main Site

 

[erocker]

Seriously, some good news. Hopefully things go quickly and efficiently for Nvidia in these new designs.

 

[Animalpak]

yeah hope so

 

[Athlon2K15]

oh no!! more pwnsauce on the way!!

 

[LAN_deRf_HA]

I wonder if this will actually have the desired effect. Aside from helping investors feel more confident I'm sure they're hoping this might make consumers reconsider getting a 6000 series. My guess is though that those who are going to wait wouldn't have bought AMD under any circumstance. So I believe AMD is about to majorly bite into their market share for quite a few months, as if they haven't already.

 

[ebolamonkey3]

I dunno, Kepler and Maxwell just doesn't have the ring to it that Fermi does. Oh well it'll probably grow on me over the next several years lol.

Hope they can move on to quickly as well, they need to compete better w/ AMD so we can get video cards that sell for less than original MSRP 9 months after release :\

On the other hand, when do we get Gauss and Euler?

 

[Wrigleyvillain]

Sweet.

 

[PVTCaboose1337]

I overlayed an exponential growth onto the NVIDIA graph to prove utter and complete BS that is that graph. It is an obvious estimation scheme. No real numbers, I promise you!

 

[Kreij]

Does any one have the formula for converting Double-Precision Computational Power per Watt into FPS in Crysis?

 

[PVTCaboose1337]

Does any one have the formula for converting Double-Precision Computational Power per Watt into FPS in Crysis?

W1zzard is coming out with that app next week. Standby.

EDIT: PS, Kriej, I believe the efficiency per watt would be determinant on individual hardware (and software!) of each person, therefore no computer would be exactly the same. It would be neat impossible to convert PpW into FPS for many reasons, one of being which the equation would be longer than Erocker's epeen, which is pretty damn long they tell me.

 

[Kitkat]

they better name something lol

 

[scaminatrix]

Does any one have the formula for converting Double-Precision Computational Power per Watt into FPS in Crysis?

Kepler - Crysis = Aww, must be a bottleneck somewhere...
I like their choices of names; I suppose they're saving the 'Einstein' one.

 

[AltecV1]

i just hope they dont screw up again like they did with gf100.
I WANT PRICE WAR

 

[wolf]

mmm juicy new architecture, still seems like they have time for a decent fermi refresh before kepler

 

[theubersmurf]

I dunno, Kepler and Maxwell just doesn't have the ring to it that Fermi does. Oh well it'll probably grow on me over the next several years lol.

Hope they can move on to quickly as well, they need to compete better w/ AMD so we can get video cards that sell for less than original MSRP 9 months after release :\

On the other hand, when do we get Gauss and Euler?

That's only because Fermi was 20th century. And Kepler doesn't really fit at all, but Maxwell I sort of like.

 

[Easy Rhino]

if nvidia actually makes their desired deadline this time then this new chipset will be a homerun.

 

[phanbuey]

I overlayed an exponential growth onto the NVIDIA graph to prove utter and complete BS that is that graph. It is an obvious estimation scheme. No real numbers, I promise you!

http://forums.techpowerup.com/attachment.php?attachmentid=38083&stc=1&d=1285098900

I mean... yeah obviously its an estimation. Thats 2013. How would they have real numbers on a chip that is barely on paper based on a process that isn't available yet?

Graph is based on Moore's law double the power every 18 mos, if you keep doubling then its an exponential curve. I do agree with you that its a little steep tho.

 

[Mindweaver]

Great paper launch!

 

[wolf]

...How would they have real numbers on a chip that is barely on paper on a process that isn't available yet?...

What I'm wondering is if they are going to keep tweaking/redesigning the CUDA cores themselves and/or the cluster formations they are in, or just keep trying to double up.

 

[the54thvoid]

Kepler....Maxwell....

If it's anything like the GF100 problems, what about...Laurel and ....Hardy.

But seriously. Plans are one thing, it'll be nice to see the design specs when they are able to release them.

And another but seriously... It wont have any effect on AMD 6 series as they will have their own 28nm process too. Let's hope they both (28nm processes) work well and come out at the same time so we, the consumers get a competitive market again.

One last seriously.... I think AMD might screw us all over with the 6 series as NV don't have anything on hand to fight it yet - assuming the hype is true.

Roll on DX15......

 

[phanbuey]

What I'm wondering is if they are going to keep tweaking/redesigning the CUDA cores themselves and/or the cluster formations they are in, or just keep trying to double up.

The very next generation will probably be a double up... after that... no idea. They always fiddle with the cuda cores either way. So I guess its a bit of both. Chances are we will be seeing the signature of Fermi for the next couple of iterations maybe even until the next DX spec.

 

[Steevo]

Date of 480 review.

Mar 19, 2010, 11:30 AM


They show Fermi as 2009, they were how many months late? So to say they plan on 2011, means perhaps they will get product to reviewers in 2012?

 

[CDdude55]

Hope the new architecture is awesome.

Fast and efficient with a nice price is all i ask.

 

[wolf]

Date of 480 review.

Mar 19, 2010, 11:30 AM


They show Fermi as 2009, they were how many months late? So to say they plan on 2011, means perhaps they will get product to reviewers in 2012?

lol what a good point to raise, it was planned for 09 but plauged with delays, hopefully kepler isnt like that. for instance if they transition Fermi to 28nm then they will have a lot of fab issues already sorted for kepler... hopefully.

Hope the new architecture is awesome.

I'm with you on that one, certianly don't hope it sucks

 

[robn]

So, improving performance per Watt ...but surely they will be reducing the number of Watts their next chipset uses vs Fermi, since Fermi is overly power hungry (even the 450 has a 6 pin connector!). That of course means Kepler's performance is not estimated to be 2*Fermi in real terms.

Still I believe they need to reign in power usage just as AMD has. There are knock on effects such as cheaper PCBs, and more systems that can take a card when power is lower. Also less power, means less heat, which inadvertently might allow higher clocks. I think it's their only option.

The market does need continued competition, so I hope things happen faster this time. As I understand it, Fermi was a ground-up redesign so they should just be building on that for a few generations.