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: wbmw who wrote (243914)	12/3/2007 11:26:19 AM
From: Petz	Read Replies (2) \| Respond to of 275872

re: Pete: <<Re: Us guys like to argue that you should compare unoptimized code because almost all widely distributed code is not optimized.>> You: <Incorrect. Most widely popular code has received many optimizations. Photoshop, 3D Studio Max, Cinebench, media encoders like DivX and MS Windows Encoder, games like Crysis, Unreal, Half Life, Doom/Quake...> In the context of servers, none of the above matter. Which is why the fact that 2.5 GHz Barcelona was faster than Intel's 45nm Harpertowns on MySQL anandtech.com is 5x more important than performance on your list of titles. (Even 95W 2GHz was <1% behind 3GHz Clovertown.) Petz

To: wbmw who wrote (243914)	12/3/2007 5:42:50 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Wbmw: The game titles are not optimized for being fastest on a given CPU. They are optimized for being fast enough on a value CPU that exists in most of the installed base. Basically a mainstream CPU 3-4 years ago. GPU driver writers (those who work for the GPU makers) do optimize the GPU code for their cards sometimes pulling dirty tricks to do better like silently lowering the quality in order to bench better in the popular titles. Those tricks are usually found by the better review sites. So that takes care of those "popular" titles. Less than 10% of the installed base ever uses Photoshop, 3D Studio Max, Cinebench, etc. These were put into the reviewer's toolbox during the P4 era where it did well on them, but badly on normal office tasks. The biggies like word processors, spreadsheets, web browsers and other office type tasks which most of the installed base uses every day, aren't optimized. As for the encoders, most users do not spend a majority of their time with them. The typical user spends less than 10% in them and again are not optimized for a given CPU. They did not do it for Athlon or K8. In fact one popular benchmark, WME 7.10..., used a highly unoptimal version for Athlons, picked of course by Intel. Later versions stripped out the CPU ID string test. Its hard to find a hard core person who uses these programs 50+% of the time, but not impossible. And servers and the VARs who supply software for them, don't use such programs either. And for Xeons and Opterons that installed base spends less than 0.1% of their time with those "popular" titles. They even usually size the hardware for the software they buy. Somewhat suboptimal software actually makes VARs more money. That turns your arguments on their head, doesn't it? Pete