By: David Kanter (dkanter.delete@this.realworldtech.com), August 9, 2011 12:12 pm
Room: Moderated Discussions
Josh (josh@penstarsys.com) on 8/9/11 wrote:
---------------------------
>Dense stuff David. Good read though. Interesting to see the tradeoffs between
>the different companies.
Thanks, I'm glad you enjoyed it!
>Too bad AMD's Trinity will use the Cayman architecture,
>and likely the current memory interconnect as Llano, >rather than the next gen core
>architecture from the AMD Graphics group and a more >tightly coupled GPU to the CPU.
Well, I don't see much in the way of alternatives. The next gen architecture is only taping out now, so there sure isn't time to get it ported over to SOI. I think we'll always see a gap between new architectures in discrete ASICs vs. integration. Partially that's due to product cycles and also due to the extra time for integration, etc.
The gap should decrease though.
>I think that the AMD Ontario/Zacate chips were the first to hit market with integrated
>graphics on the same die as the CPU. These were announced >several months before
>Sandy Bridge. Unless of course you consider that most ARM >implementations also
>have integrated graphics on die, in which we can go back >to the NS/AMD Geode series with that particular feature.
As I mentioned down-thread, the definition of which was first is very tricky. I agree that Zacate was about the same time, but it's not really clear to me one way or another. It's a matter of days AFAICT.
David
---------------------------
>Dense stuff David. Good read though. Interesting to see the tradeoffs between
>the different companies.
Thanks, I'm glad you enjoyed it!
>Too bad AMD's Trinity will use the Cayman architecture,
>and likely the current memory interconnect as Llano, >rather than the next gen core
>architecture from the AMD Graphics group and a more >tightly coupled GPU to the CPU.
Well, I don't see much in the way of alternatives. The next gen architecture is only taping out now, so there sure isn't time to get it ported over to SOI. I think we'll always see a gap between new architectures in discrete ASICs vs. integration. Partially that's due to product cycles and also due to the extra time for integration, etc.
The gap should decrease though.
>I think that the AMD Ontario/Zacate chips were the first to hit market with integrated
>graphics on the same die as the CPU. These were announced >several months before
>Sandy Bridge. Unless of course you consider that most ARM >implementations also
>have integrated graphics on die, in which we can go back >to the NS/AMD Geode series with that particular feature.
As I mentioned down-thread, the definition of which was first is very tricky. I agree that Zacate was about the same time, but it's not really clear to me one way or another. It's a matter of days AFAICT.
David