> Do you know if this also works if you replace i386 packages with > recompiled i686 ones? If so has anyone actually benchmarked a system > with a standard install vs the same packages recompiled to i686. > The reason I ask is that my requirement for linux is for intensive > number crunching and so to get a 5% to 10% performance increase would be > sufficient justification for recompiling - on the other hand no good if > you can't kickstart from it. I would guess yes. I would compile one and test before going to recompile all. But some of the experts might answer immediately... Anne
