NVIDIA Begins Sampling First Kepler GPUs

[btarunr]

Graphics major NVIDIA has begun sampling the first GPUs of its next-generation Kepler family. Kepler succeeds the present generation Fermi family of GPUs, that make up the GeForce 400 and 500 series. Members of the GeForce Kepler family will follow the codename nomenclature of GK1xx, like GeForce Fermi followed GF1xx. The first GPU being sampled is codenamed GK107. The codename suggests that this isn't the top-end part, it's more like a lower-mainstream or value segment offering.

It is reported that GK107 has a 128-bit GDDR5 memory bus, which also supports inexpensive DDR3 memory. It is built in small quantities on TSMC's 28 nanometer fab process (as the foundry isn't ready for 28 nm bulk manufacturing. NVIDIA will follow a "bottom-to-top" strategy, when dealing with a new fab process technology. It will first design the smallest, simplest GPUs in the lineup, and then gradually move on to larger ones. The first GK107-based SKUs will succeed the GeForce GT 500M series. NVIDIA will carve out four SKUs, internally, NVIDIA will refer to those as N13P-LP, N13P-GS, N13P-GT and N13E-GE, with N13E-GE being the "enthusiast" part. Its market SKU will likely succeed the GeForce GT 560M.

View at TechPowerUp Main Site

 

[Lionheart]

Using AMD strategy now aye ^_^

 

[NC37]

Well..104 was the 460, 106 was the 450. Yeah 107 would be midrange if they are following that scheme now.

Bottom up strat...hmm. Isn't that what they did in the 3 series? I remember it seemed like the 3 series was just all low end parts. Course ATI stole a lot of the show with the 4870s.

Kinda prefer it when they show some of the good stuff first. But I'm sure this way they can tweak things as AMD launches 7 series.

 

[DarkOCean]

Using AMD strategy now aye ^_^

Good they finally learned something good, but they also doing it their own way releasing first 28nm gpus for laptops so they can ask for them many $$$ like they usually do with the mobile cards compared to amd.

 

[phanbuey]

Is the enthusiast GPU still scheduled to come out Q4 2011?


I thought I had read that somewhere.

 

[Casecutter]

Sounds like a Pipe-cleaner to verify process... good move!
Though instead of running it with the current GT5XX number (like most pipe-cleaners), they'll roll with the new nomenclature, while in all probability this will be very low volumes (Aka pricy). But this is good for Nvidia as they’ve been exceptionally weak on 128-Bit entry level (<$100) for a while now… the last "good one" was a GT240 DDR5.

 

[Sihastru]

It depends on TSMC really, as they are to blame for the 40nm fiasco (well, mostly), but 3 months is a bit of a tight squeeze. If it does come out, it's going to be very close to the new year's eve.

I just hope we won't have the incremental upgrades like we had from GTX280 to GTX285 a few weeks/months after they release it. It's also rumored that we will get a 670, 680 and a 690, all single GPU cards at the high end. How would you feel if you bought a 680 or 690 in December and at the begining of February they come out with 685's and 695's...

I myself am waiting to buy two of the higher end 600 series cards for an SLI setup, but I'm going to wait a bit before taking the plunge. Any disabled SP's in any of the chips or not good enough efficiency can be an indication of a quick GPU "facelift".

 

[phanbuey]

It depends on TSMC really, as they are to blame for the 40nm fiasco (well, mostly), but 3 months is a bit of a tight squeeze. If it does come out, it's going to be very close to the new year's eve.

I just hope we won't have the incremental upgrades like we had from GTX280 to GTX285 a few weeks/months after they release it. It's also rumored that we will get a 670, 680 and a 690, all single GPU cards at the high end. How would you feel if you bought a 680 or 690 in December and at the begining of February they come out with 685's and 695's...

I myself am waiting to buy two of the higher end 600 series cards for an SLI setup, but I'm going to wait a bit before taking the plunge. Any disabled SP's in any of the chips or not good enough efficiency can be an indication of a quick GPU "facelift".

+1 or if they come out with a 570/590 debacle where they dont put enough VRM's on the cards. One of the reasons I am asking becuase I just popped my GTX570 last night playing Dragon Age 2

I think i will go back to get a refund and just go back to my 470 until the newer generation comes out.

No OCP override and a mild overclock...

 

[WarraWarra]

Confused, TSMC is not ready for large scale 28nm but they are ready for ?? 20~nm "25nm" ?? so why not just jump to what they have ready.
I can imagine that 20nm "25nm" if correct size is more beneficial for Nvidia / end users.

All this sounds too much like "Nvidia" again have no clue what they are doing and like blind mouse in a maze are trying to smell their way out of trouble.

Come on NVidia we buy your shares / products because + for desktop use not your most lucrative notepad / mobile phone something that no one cares or uses.

Nvidia if your run out of idea's why not just glue 1200x nvidia e-top/notepad quad core cpu's together, stick it on a gpu board and call it your next great gpu for gaming. That stuff that is the same as a intel (atom/i3/i5) cpu.

If NVidia can not then just get 10x Chinese to do it, they can and likely will or have done so already.

 

[[H]@RD5TUFF]

Well..104 was the 460, 106 was the 450. Yeah 107 would be midrange if they are following that scheme now.

Bottom up strat...hmm. Isn't that what they did in the 3 series? I remember it seemed like the 3 series was just all low end parts. Course ATI stole a lot of the show with the 4870s.

Kinda prefer it when they show some of the good stuff first. But I'm sure this way they can tweak things as AMD launches 7 series.

As long as they run cooler than the 4 and 5 series, so we can get some passive loving I'm down with it.

 

[Casecutter]

The more I read this... it has me thinking its' about OEM SKU's, it doesn’t ever explicitly call them aftermarket. And looking at it that way it makes sense, and a win-win for the players.

Intel’s soon to be Ivy Bridge needs good graphics to sell, really above Llano A8 (if AMD ever really start to show; let’s say production does sort its' self out about the same time as I-B). Intel will need to have the latest and greatest mainstream graphics that bumps them above the A8 (6550D)... but low power.

Nvidia has only a GT545 and that wasn't going to be enough to bump the 6550D, or do it with low power. So Intel "makes" Nvidia (can’t go to AMD) to get a discrete card of low power (≤70W) that will be more like a 6670. Nvidia see's the issue also, if folk buy a Llano because of good graphics' they’ll never see a upgrade sale. Those folks will just take advantage Hybrid-Crossfire. If they release this now all Intel OEM boxes will get a GK107, which will provide a "Aura upshot" for all Kepler’s just as they come out in the aftermarket. Next if they (Nvidia) negotiated it right with Intel to let let go with 2x PCI-E slots mobos (maybe only for the first couple of months) and the OEM’s to install a adequate PSU to position Nvidia with the upgrade (sale for another) to SLI.

That's a "Win-Win" because as good as Ivy Bridge can be... without class the leading graphics and holding down TPD and price they would've had a tough hill to climb in the OEM boxes. The GT545 was all they had and that wasn’t going to make the grade performance/watt verse the Llano A8 no matter how you sliced it. Right now Intel/Nvidia and OEM’s are gearing up for the next generation Ivy Bridge machines and they all saw the writing on the wall. It doesn't matter how good either "could be" by themselves, they weren't anything if the OEM's didn't have a viable option. Now they should be able to give great I-B CPU and graphics close to a GTS250/6670, which trumps an A8 at an attractive price.

 

[bear jesus]

This is a very smart move for nvidia, apparently AMD is doing the same thing and with TSMC's issues it would not surprise me if it was the only logical path as trying to build monster multiple billion transistor chips right now could be a big mistake leading to the kind of issues Nvidia had with the first fermi chips as it was not so much the chip but more the flaws in the process node.

Is the enthusiast GPU still scheduled to come out Q4 2011?


I thought I had read that somewhere.

I have mainly read Q1 2012, i had seen Q3 2011 floating around for AMD's cards but apparently they will also not be out until at least Q1 2012, but i admit all this comes from "news" articles with no official confirmation as, at least as far as i know there is no official word on release date yet.

*edit*
Although as i read more i keep seeing Q2 rumored.

But it seams as it is AMD is still ahead as this is saying that Nvidia is sampling their first 28nm mobile chips and AMD has already been showing off mobile 28nm chips running, but either way it seams like both company's are getting delayed by issues with the 28nm process.

 

[NC37]

The more I read this... it has me thinking its' about OEM SKU's, it doesn’t ever explicitly call them aftermarket. And looking at it that way it makes sense, and a win-win for the players.

Intel’s soon to be Ivy Bridge needs good graphics to sell, really above Llano A8 (if AMD ever really start to show; let’s say production does sort its' self out about the same time as I-B). Intel will need to have the latest and greatest mainstream graphics that bumps them above the A8 (6550D)... but low power.

Nvidia has only a GT545 and that wasn't going to be enough to bump the 6550D, or do it with low power. So Intel "makes" Nvidia (can’t go to AMD) to get a discrete card of low power (≤70W) that will be more like a 6670. Nvidia see's the issue also, if folk buy a Llano because of good graphics' they’ll never see a upgrade sale. Those folks will just take advantage Hybrid-Crossfire. If they release this now all Intel OEM boxes will get a GK107, which will provide a "Aura upshot" for all Kepler’s just as they come out in the aftermarket. Next if they (Nvidia) negotiated it right with Intel to let let go with 2x PCI-E slots mobos (maybe only for the first couple of months) and the OEM’s to install a adequate PSU to position Nvidia with the upgrade (sale for another) to SLI.

That's a "Win-Win" because as good as Ivy Bridge can be... without class the leading graphics and holding down TPD and price they would've had a tough hill to climb in the OEM boxes. The GT545 was all they had and that wasn’t going to make the grade performance/watt verse the Llano A8 no matter how you sliced it. Right now Intel/Nvidia and OEM’s are gearing up for the next generation Ivy Bridge machines and they all saw the writing on the wall. It doesn't matter how good either "could be" by themselves, they weren't anything if the OEM's didn't have a viable option. Now they should be able to give great I-B CPU and graphics close to a GTS250/6670, which trumps an A8 at an attractive price.

That would all depend on if Intel is really feeling the pressure from the APU yet. Even if they were, going to NV would admit right away that their IB doesn't have the stuff to match even the current gen APUs. Which I wasn't expecting it to do anyways.

But on the plus side if this was true, it would mean a lot more decent Intel based laptops out there with NV chips. But they'd have to make the companies market for it too. I really don't think they can build an Intel system with discreet graphics for sub $400 that can beat an APU. Intel charges too much of a premium and NV does too at times.

Asus has an APU model at Best Buy right now that I am just shocked by. A6 3400, plus it comes with a discreet 6650 that when Crossfired goes up to a 6720. All for $450. Really this is what I was hoping APUs would do, bring decent GPU performance to the masses who can't afford gaming laptops. Can Intel match that? Without NV, it can't. But I really have to wonder how they can, without cutting their profits. For the same machine on an Intel build, I'd expect them to price it at about $550 minimum just because it has the Intel name.

 

[WarraWarra]

Talking about APU's seems more likely that nvidia is trying with Kepler to get a place in line or at least trying to get ready for the normal Apple IGP/APU graphics for the next laptop release to try and knock ATI out as current APU/IGP graphics supplier for Apple with this move.

NVidia lost a lot of money on the current Mac Book Pro Apple ATI IGP graphics cards that is labels as video cards. "ATI6750m, current ATI APU equivalent" A NV560/570m would have been the lowest specs Apple should have done.

If NVidia gets their ducks in a row this time with Kepler they might even be allowed to have desktop pc GPU's in Apple Desktops instead of the current APU equivalent GPU's in Apple Desktops pc's.
Just hoping NVidia Kepler can save Apple from themselves, if this is what is really going on at NVidia.

Sry Apple but no one else is stupid enough to put a AMD APU as gpu in a $2200 laptop and call it a GPU or make all the spelling mistakes on a website.

http://www.apple.com/macbookpro/specs.html
AMD Radeon HD 6490M graphics processor with 256MB of GDDR5 memory on 2.0GHz configuration; or AMD Radeon HD 6750M graphics processor with 1GB of GDDR5 memory on 2.2GHz configuration

Note the ; and missing . on that Apple's website link.

 

[bencrutz]

Talking about APU's seems more likely that nvidia is trying with Kepler to get a place in line or at least trying to get ready for the normal Apple IGP/APU graphics for the next laptop release to try and knock ATI out as current APU/IGP graphics supplier for Apple with this move.

NVidia lost a lot of money on the current Mac Book Pro Apple ATI IGP graphics cards that is labels as video cards. "ATI6750m, current ATI APU equivalent" A NV560/570m would have been the lowest specs Apple should have done.

If NVidia gets their ducks in a row this time with Kepler they might even be allowed to have desktop pc GPU's in Apple Desktops instead of the current APU equivalent GPU's in Apple Desktops pc's.
Just hoping NVidia Kepler can save Apple from themselves, if this is what is really going on at NVidia.

Sry Apple but no one else is stupid enough to put a AMD APU as gpu in a $2200 laptop and call it a GPU or make all the spelling mistakes on a website.

http://www.apple.com/macbookpro/specs.html
AMD Radeon HD 6490M graphics processor with 256MB of GDDR5 memory on 2.0GHz configuration; or AMD Radeon HD 6750M graphics processor with 1GB of GDDR5 memory on 2.2GHz configuration

Note the ; and missing . on that Apple's website link.

there's no APU on 15" MBP, it's GPU.
if you take the 2GHz core i7 configuration, it comes with Radeon 6490M GPU.
if you take the 2,2GHz one, it comes with Radeon 6750M GPU.
which part does it say it uses AMD APU? it would be impossible to use intel processor with amd APU in one single motherboard

 

[Casecutter]

That would all depend on if Intel is really feeling the pressure from the APU yet.

I think Intel knows their GMA graphics has become their Achilles heel for low end systems (desktop or laptops). At least in a desktop they can plug in a GT520 and get by, but then OEM’s are working with a CPU that has graphics but isn’t providing any value. So Intel has lost any bargaining chip while OEM need to get a card from Nvidia (so costing goes out of kilter). Intel needs to stop worrying about raw compute power and start finding some serious improvement with real integral graphics. Every other netbook and notebook OEM are extensively already on or gearing for the APU. I could see Apple considering AMD and APU in notebooks sometime when Trinity hits. Why have two vendors and a more complex mobo for discrete graphics, it improve packaging and TPD which are what interest Apples engineering.

Sry Apple