On Sat, Feb 25, 2012 at 5:22 PM, Walter Dnes <waltd...@waltdnes.org> wrote: > On Fri, Feb 24, 2012 at 03:13:07AM +0200, Nikos Chantziaras wrote > >> The speed gains of building for specific submodels of CPUs might >> be there, but they're minimal. Benchmarks have shown (can't find >> the article, it was on Phoronix) that after -march=i686 you get >> diminishing returns. > > In that case, the benchmarks are useless. From my personal > experience... a fresh i686 install on a 4 and 1/2 year old Dell with > onboard Intel GPU was not able to keep up with the slowest available > speed on NHL Gamecenter Live. Ditto for 1080i TV from my HDHomerun > tuner box. After rebuilding system+world+kernel with "march=native", > it works just fine for the above tasks.
Playing video is one of few situations in which optimisation makes a lot of difference though, thanks to the mmx/sse stuff, which is post i686. So a video benchmark will should show that up, but the boost may be lost in a more general suite of weighted benchmarks. Also, different versions of gcc optimise differently. Usually the optimisation gets better but there a quite a few cases where performance regresses.
