SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?


To: Saturn V who wrote (255203)8/4/2008 11:10:58 PM
From: muzosiRead Replies (2) | Respond to of 275872
 
I think that the Larabee 'blue crystals' are only applicable to a subset of computational problems.

actually we already know that graphics acceleration is a very good application because existing gpus have several hundred execution units already and they're being kept full by existing games over existing interfaces so it's safe to assume larabee can do the same.
i think the real question is the baseline of larabee even if one assumes it can scale. the raw available processing power of a 32 core larabee is not much more than 4870 of today which means it really needs almost perfect scalability if it has to compete. and larabee is a '10 product most probably where as one can buy a 4870 now.

also larabee seems to be committing again the mortal sin of new hardware ie assuming there will be perfect software support for it. all the games' rendering engines which are optimized for current directx 10/11 + opengl have to go through the rendering compiler and be scheduled in there. game programmers understand the hard-wired schedulers of current gpus. it's not that safe an assumption that they'll have an easy time converting their engines to larabee; scheduling for 32 general purpose cores is not easy. if they don't have to do it, it's left to intel to optimize their renderer/scheduler to each and every game which is not an easy task either even with intel's size.