Ah, jeez...
Dude, this is like the 10th time that stupid "c't" SPEC benchmark thing has come up. Maybe this should be in a FAQ or something.
To start with, ALL benchmarks produced by hardware manufacturers are stacked in favor of the company's own hardware. Duh. Apple uses Photoshop for comparison because Photoshop is really really damn fast on Apple hardware compared to PCs. You expect them to NOT demonstrate their strengths?
Second, without getting into details
again, I've done SPEC professionally, and from their original article, there are some things "c't" simply do not understand about this benchmark, the conditions under which it should be run, and the applicability of the results.
The reason Apple has not posted any significant SPEC results yet is because Motorola hasn't got a compiler division worth squat. PC manufacturers get support from Intel, and Intel has one hell of a compiler group. Nobody uses Intel's compiler for REAL code, but it does amazing things for benchmarks on Intel hardware. And even a comparison on two platforms using the same version of gcc says at least as much about gcc's code generation on Intel versus PPC as it does about real performance.
The "c't" test has long ago been demonstrated to be terminally flawed. Your G4 is not a piece of junk, and Apple is not pulling the wool over your eyes any more than any other hardware manufacturer.
Furthermore, there is a LOT of politicking in that article. I like this quote:
The problem is, one former Apple engineer told us, in serializing the twenty five year old BSD layer with the fifteen year old code of the extensions NeXT began to add in the mid 1980s.
The reason UNIX has survived mostly intact for 25 years is that it's well-though-out and efficient. Anyone with any technical knowledge recognizes the internals of UNIX to be relatively clean and consistent compared to the dirty snowball of Win32, which just keeps accumulating more crud with each passing year. And being fifteen years old does not necessarily make the NeXTstep frameworks bad. Honestly, the only OS whose guts I have found more satisfying than OS X has been BeOS, which was just a thing of beauty internally.
Now onto the political side: Surely you're aware of how many times Apple tried and failed to produce a "next generation" OS to supercede the original Mac design. One of the major reasons this happened is that, unfortunately, a lot of the original MacOS engineers (the guys who signed on as Pascal jockies in the mid eighties) were
totally stuck on the original MacOS as a paradigm. It's sad, and it seems ungrateful on Steve Jobs' part, but in order to get you MacOS X, many of those guys just had to be fired. So I ask you whether it isn't possible that the unnamed "former Apple engineer" they cite isn't someone with an axe to grind.
Also problematic is their use of the term "serializing." They seem to be adopting a definition of a technical-sounding term which just doesn't make a lot of sense. "Serialization," they claim, "is one of the toughest problems for an engineer to solve." I've been an engineer for quite a while. I have a working definition of the word "serialization" which in no way relates to any issue in this article. I am aware of no definition of the word which
would apply, much less one which would present itself as one of the toughest problems in my line of work. I don't know
who is feeding
whom a line of crap here, but make no mistake, that's exactly what it is.
So, yes, David, the article is false. I can't claim definitively that Andrew Orlowski, the author of the article, is willfully lying to you, but the information he is presenting to you is not entirely correct.
