+ Reply to Thread
Results 1 to 6 of 6

Thread: Fusions GPU. Will it be open source friendly?

  1. #1
    Join Date
    Nov 2007
    Posts
    450

    Default Fusions GPU. Will it be open source friendly?

    The title basically says everything that I want to know. I know you guys arent allowed to speculate on future products, but I was just wondering whether it would fall under the R600\R700 class, or if it would be R800 or newer?

    If it is R700 then by the time it's released it should have decent open source support. But really I'm mostly concerned about general purpose processing. It would be nice to have a general purpose ISA, or API that developers could use for things like video codecs. Or maybe physics libraries. It'd take DRI out of the loop, and would allow the GPU to be used far more liberally.

    What generation will the core be? How open source friendly will it be And what about an open source general purpose API?

  2. #2
    Join Date
    Jul 2008
    Posts
    275

    Default

    IMHO, this is exactly why AMD started down the open source path with their current GPUs - to lay the foundations for when GPU-type functions are an integral part of enterprise-class products, where open source is often demanded. They've said that future cores will be developed with open source in mind from the start, so R800 onwards should see shorter and shorter times before mature open-source drivers are available.

  3. #3
    Join Date
    Nov 2007
    Posts
    621

    Default

    R800 may not be any better. Like the AMD folks have been saying a lot (and I know you guys are reading them), the design process is around 3-4 years. R800 will probably be out late next year, but the super Open Source friendliness only started a year ago. It's just speculation on my part, but I wouldn't bet on R800 being perfectly Open Source friendly. I would expect it to have a lot faster turn around time on the documentation, though.

    I also can't possibly imagine that Fusion -- currently expected to debut in late 2011, so it probably won't be out until 2012 -- would be based on a graphics chipset released in 2008.

  4. #4
    Join Date
    Nov 2007
    Posts
    450

    Default

    Quote Originally Posted by elanthis View Post
    R800 may not be any better. Like the AMD folks have been saying a lot (and I know you guys are reading them), the design process is around 3-4 years. R800 will probably be out late next year, but the super Open Source friendliness only started a year ago. It's just speculation on my part, but I wouldn't bet on R800 being perfectly Open Source friendly. I would expect it to have a lot faster turn around time on the documentation, though.

    I also can't possibly imagine that Fusion -- currently expected to debut in late 2011, so it probably won't be out until 2012 -- would be based on a graphics chipset released in 2008.
    I've been under the impression it'll be a shanghai derivative with some unspecified GPU sometime in the 2009 timeframe, perhaps even q1 or q2... That'd be stage1 of the fusion roadmap. I've also heard that AMD might skip stage1 altogether and release stage2 as bulldozer in the q42010 - q1 2011 time frame.

    If your talking about stage2, then I think AMD is making a massive mistake. From the benches I've seen there is no way in hell that AMD is gonna be able to ride Deneb for the next 2 years.... AMD needs Fusion ASAP.

  5. #5
    Join Date
    Oct 2007
    Location
    Toronto-ish
    Posts
    6,087

    Default

    I've been under the impression it'll be a shanghai derivative with some unspecified GPU sometime in the 2009 timeframe, perhaps even q1 or q2... That'd be stage1 of the fusion roadmap. I've also heard that AMD might skip stage1 altogether and release stage2 as bulldozer in the q42010 - q1 2011 time frame.
    I think the conclusion was that Fusion needed to be 32nm in order to be a truly successful product in the mainstream segments. It's a tricky tradeoff -- you save money on packaging and even some logic by going to a single die, but larger dies tend to have lower yield and higher per-transistor cost once they get above a certain size.

    FYI, one of the reasons for starting the open source graphics project when we did was to make sure that we were in great shape in time for Fusion.
    Last edited by bridgman; 11-27-2008 at 11:47 PM.

  6. #6
    Join Date
    Apr 2007
    Location
    Mexico City, Mexico
    Posts
    877

    Default

    Quote Originally Posted by duby229 View Post
    And what about an open source general purpose API?
    Are you referring to OpenCL [1], by any chance?

    THAT actually makes a LOT of sense.
    Last edited by Thetargos; 11-28-2008 at 02:31 AM.

+ Reply to Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts