On Jun 9, 2005, at 3:17 PM, Kynan Shook wrote: > Why don't you consider adopting the wait-and-see attitude that so many > others have adopted? An Intel Mac will probably beat a PowerPC Mac at > many tasks, and a PowerPC Mac will probably beat the Intel Mac at many > tasks. But as time wears on, the Intel Mac will improve more rapidly. Until we find out from Jordan Hubbard and his team what they've done to improve kernel threading in Darwin, I'm going to remain quite pessimistic as to the future of OS X on x86 hardware because of fundamental differences in how threads and vector code is handled on the two architectures. x86 Darwin runs like a 3-legged dog compared to PPC Darwin on comparable hardware. x86 Darwin runs like a dog gasping for air in its final moments of life compared to linux on the same x86 hardware. Slapping Aqua on top and calling it "Mac OS X" doesn't make it magically run better. The problem with OS X is that Mach kernel threads are only available for kernel level programs, not applications. Apps have to make use of a POSIX thread - slower user-level threads, and not fast kernel threads like linux does for instance. Those x86 dev boxes are not there to only make sure your app *runs* on x86. It has to be *optimized* or you're going to loose the battle. If similar functionality (in your app) is available on Windows or Linux, and either of those two blows your x86 Mac app away in performance doing the same task, you won't be able to sell your app because the mainstream media will benchmark it and label it a "dog". The performance comparisons between the dev boxes and current PowerPC are very real folks, because you're dealing with current-production hardware on both archs. What ever performance increases come down the pipeline on the x86 side will also be enjoyed by x86 linux and Windows, making the idea that "shipping" boxes are going to perform better than the dev boxes a moot point. Some x86 Mac benchmarks have already been leaked from WWDC <sigh> -- Chris