Silicon Investor (SI) -- The First Internet Community

STOCKTALK

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: milo_morai who wrote (74157)	3/10/2002 3:25:02 PM
From: Dan3	Read Replies (2) \| Respond to of 275872

IPP - The Intel Performance Penalty I dared to post some things moderately critical on the Intel thread today, and while responding to the barrage of personal attacks the Intel cheeleaders throw at anyone who dares such an affront, I started thinking about something. On almost any code that hasn't been carefully hand tweaked to work around P4's fussy architecture, Athlon's robust architecture provides better performance. Intel's design teams always start from "if the compiler will produce perfect code, how do we design the chip." While AMD's design teams start from "what's the best design to run the code produced by real world compilers." The result: AMD's chip is better on 95% of software, while Intel's chip is better on the 5% (or less) of the code out there that has been tweaked to be "perfect" for that particular chip. By the time more than 5% of the code base is suitable for a given Intel architecture, Intel is shipping a new design, with new weaknesses, that needs different optimizations to perform well. And that's why AMD's chips are now, and will continue to be, better for 95% of the software out there, while Intel's chips are great for a handful of specially optimized programs and mediocre on everything else. watch.impress.co.jp But hey, why should Intel care, it's Intel's customers who bear that burden, not Intel. It's also why Intel, and its squads of dis-information specialists, are always so desperate to deny the reporting of any but the most radically tweaked benchmarks. Like the Sysmark benchmark that runs a rare broken-for-Athlon version of Windows Media encoder in the background throughout the suite of benchmarks. It seems that each generation of Intel's designs can run fewer and fewer programs well. Almost every comparison review on the web has a half dozen variations of Quake 3 and memory bandwidth tests, but the more fundamental and less easily distorted Chessgame and SuperPI benchmarkes are almost invisible. Maybe that's a phrase that can be added to the lexicon like IPC (instructions per clock), as in, "on should always consider the IPP (Intel Performance Penalty) when buying a new, Intel based computer unless you expect to re-buy or re-write most of your software.