POST A COMMENT

15 Comments

Back to Article

  • ltcommanderdata - Monday, May 21, 2012 - link

    I just wanted to thank Manju Hegde for taking the time to respond to all those questions. Thank you Anand as well for organizing this. I'm definitely looking forward to the role-out and adoption of HSA. Reply
  • Lucian Armasu - Monday, May 21, 2012 - link

    That sounds kind of like what ARM wants to do with big.Little and Mali GPU's:

    http://www.cnx-software.com/wp-content/uploads/201...

    http://www.cnx-software.com/2011/11/22/midgard-arc...
    Reply
  • CeriseCogburn - Wednesday, May 23, 2012 - link

    The first page sounds to me like AMD is going for Homeland Security dollars.
    " We have done extensive analysis on several workloads and have obtained significant performance per joule savings for workloads such as face detection, image stabilization, gesture recognition etc"

    I see.
    Don't be evil amd.
    Reply
  • CeriseCogburn - Wednesday, May 23, 2012 - link

    Then I thought maybe they are going for gaming console capabilities - but it appears on page2 they are going for logging on to Windows:

    " OSes are moving towards providing some base functionality in terms of security, voice recognition, face detection, biometrics, gesture recognition, authentication, some core database functionality. All these benefit significantly from the optimizations in HSA described above. With the industry support we are building this should happen in the next few years."

    So, it's probably all three.
    Reply
  • Loki726 - Monday, May 21, 2012 - link

    I'll need some time to read over all of the responses, but I want to thank Manju Hegde for taking the time to respond in such detail. Reply
  • SanLouBlues - Monday, May 21, 2012 - link

    I'm not sure he understood the question of the guy with the Intel CPU and AMD GPU who wanted GPU accelerated winzip. Reply
  • c0d1f1ed - Monday, May 21, 2012 - link

    None of the tough questions were answered, like how they're hoping to compete against AVX2 and AVX-1024.

    I'm sure HSA is an improvement, but it's not developed in a vacuum. AMD has to be willing to compare it against homogeneous high throughput computing. They won't gain the respect of developers if they dodge questions. It doesn't show much confidence in their own technology.
    Reply
  • texasti89 - Monday, May 21, 2012 - link

    +1 Reply
  • B3an - Monday, May 21, 2012 - link

    Thank you Manju for answering my questions! :) And a interesting article, still reading it... Reply
  • oldguybt - Tuesday, May 22, 2012 - link

    Sounds like for AMD it was surprise...?Just like they find out just now, that they have usualy x2 x3 or x4 more steam processors than Nvidia cuda cores.
    Probably local "hacker" told them that than u trying to decrypt in Backtrack WPA/WPA2 internet password AMD GPU do that 4 or 5 times faster.
    Well i think i find out that faster than AMD :D

    ok sorry for stupidity....
    Anyway AMD saying that their GPU`s made for gaming is actualy is like TESLA or i`m wrong?? :)

    Just buy Nvidia,
    Reply
  • spaceyyeti - Tuesday, May 22, 2012 - link

    You, sir, make no sense to me. But that just as well could be me.
    ;)
    Anyways; this article was a very interesting read.

    Loved it.
    Reply
  • tipoo - Tuesday, May 22, 2012 - link

    Kepler has more stream processors but at half the clock speed of Fermi (hot clocks), they aren't directly comparable. Reply
  • oldguybt - Tuesday, May 22, 2012 - link

    Ok, its not just steam processors but what i`m saying sounds like AMD findout that their cards is working much better in programing, counting and so on.
    I`m noob here but with linux i could do that before and they saying i could do that in future..
    Ok. Good job AMD anyway. First + in your side in 10 years.
    Reply
  • ullix - Wednesday, May 23, 2012 - link

    What remains unclear to me: in order to take advantage of e.g. OpenCL must I use an APU with CPU and GPU on the same die, or can I also use a discrete CPU and a GPU on a plugin card? Apparently the latter would always offer more power to begin with, which then could (or could not?) be enhanced with OpenCL and the like?

    Or is it that an APU has the advantage, as it is already designed to use memory common to its CPU/GPU?

    I am thinking primarily of x264 encoding - will I be better off with a "slow" A10-xxxx then with a FX8150+Radeon card 7xxx?
    Reply
  • TC2 - Thursday, May 24, 2012 - link

    big words, big plans aaand big red\green power point presentations, but 1 or 2 generations behind Intel & Nvidia :)))
    about sw support, they just haven't idea what is the meaning of this!!!
    Reply

Log in

Don't have an account? Sign up now