Although we're barely into August, both AMD and NVIDIA are already making their first moves for their next generation products due at the end of this year and into next year. AMD has recently told investors that it intends to release products using 28nm GPUs this year, and now NVIDIA is telling developers something similar

At their GTC Workshop Japan event, NVIDIA's Senior VP of Research Chris Malachowsky let developers know that the company is expecting that Kepler "should start shipping by the end of the year." Kepler will be the basis of NVIDIA's next generation of GPUs, and like AMD's forthcoming products is expected to be fabricated on TSMC's 28nm process. Though the focus at an event like GTC is primarily on workstation (Quadro) and server (Tesla) products, historically NVIDIA has shipped consumer (GeForce) products first, so it's likely that we're looking at the first shipping date for NVIDIA's newest GeForce parts.

Of course it goes without saying that shipping a GPU is not the same as a video card being available for sale - NVIDIA has made their sale once they ship a GPU to a partner. So NVIDIA shipping Kepler GPUs this year does not necessarily mean that the resulting video cards will also be available this year.

As for performance, it shouldn't come as any surprise that NVIDIA is offering few details - never mind the fact that GTC isn't the venue for gaming performance. For GTC NVIDIA is talking about double precision performance, which looks like it will become a major race between AMD and NVIDIA for this upcoming generation. For Kepler NVIDIA is expecting "about 3x improvement in [double precision] performance per watt," which would be a combination of the die shrink and architectural changes

Update: NVIDIA sent Xbit a clarification that puts the brakes on things.

"Although we will have early silicon this year, Kepler-based products are actually scheduled to go into production in 2012. We wanted to clarify this so people wouldn’t expect product to be available this year,"

So technically Kepler GPUs would still ship to partners this year, but "early silicon" normally signifies GPUs that aren't quite production worthy yet. In any case finished video cards definitely won't be out this year given NVIDIA's latest statement.

Source: Xbit Labs

POST A COMMENT

13 Comments

View All Comments

  • wifiwolf - Wednesday, August 03, 2011 - link

    Nvidia's claims are a bit cheap at the moment since it's twice already that they didn't deliver in time. So we must assume whenever marketing puts out a launch date, we just add 6 months to it.

    I assume this is just to hold people from buying AMD early as last times.
    Reply
  • DanNeely - Thursday, August 04, 2011 - link

    And all AMD launch dates need to be held to the level of fail they suffered with the 2xxx series... Reply
  • IvanChess - Sunday, August 07, 2011 - link

    The HD 2000 series was Ati, the HD 5000 series and up is AMD. Reply
  • HOOfan 1 - Tuesday, August 09, 2011 - link

    How about Bulldozer? Wasn't that originally slated to release back in March or April? Reply
  • EnsilZah - Thursday, August 04, 2011 - link

    Are there any plans for an answer to Eyefinity?
    I'd like to get one card to run my three monitors but I still need CUDA support.
    Reply
  • aguilpa1 - Thursday, August 04, 2011 - link

    Uh, hello? 590GTX!

    I have been running 3 24" Alienware of this one card for some time now and works great even when gaming at 6000x1080. Yes it is a dual card but before the 590 you actually had to have two cards regardless and is smaller then one of my old 295's I used to have, is much quieter and draws less power then the two 295's it replaced.
    Reply
  • DanNeely - Thursday, August 04, 2011 - link

    A product aimed at the top.1% (SLI) or 0.01% (595) of the market isn't an proper answer; merely a stopgap. Reply
  • JKflipflop98 - Thursday, August 04, 2011 - link

    A product aimed at the top .1%? You mean like triple monitor users? Reply
  • DanNeely - Thursday, August 04, 2011 - link

    Not everyone who uses multiple screens is a gamer. For anything less (or even casual games) a 5450 is more than adequate. Nor do all multiple screen users who game do so on all three. If ATI/nVidia ever add support for PLP I'd be willing to give it a try at 4960x1600, until then I'm quite happy gaming on my center screen. Otherwise I find 3 screens very useful for everything else, and the 30" is a nice extravagance but hardly necessary; I'd rather have 3x20 than 1x30 given the choice (sadly at work I can't have either). Reply
  • aguilpa1 - Friday, August 05, 2011 - link

    The OP wants a single Nvidia card that supports the use of 3 monitors and CUDA for that reason the 590 is a proper answer.

    Technically, it is an expensive high end product and Nvidia should be supporting the use of multiple monitors on cards less than a 590 no question. However from experience the card is incredibly fast and works flawlessly with 3 monitors. I do a great deal of photography besides gaming and actual gaming is probably more like 10% of use. I greatly appreciate the use of 3 screens in Photoshop and the CUDA accelerated features. Gaming on 3 screens at the highest settings is a very nice bonus and breathtaking on auto racing sims.
    Reply

Log in

Don't have an account? Sign up now