longofest said:
Granted it's Apple's own tests, but they do show a HUGE difference between G5's and Athlon-64. If you are able to find competing scores that show a reversal between a dual-G5 and AMD-64, post it to this list so we can actually use some info rather than unsubstantiated claims. The Athlon64 may mature into a better chip, but as of now, it is underpowered for a 64-bit chip.
Apple very smartly only focuses on benchmarks that take advantage of the PPCs strong DSP capability. Clock for clock, the AMD64 cores are slightly faster than PPC970 cores on average in every general purpose high-performance computing benchmark I've seen conducted. They have different strengths though where one will clearly outshine the other.
The PPC chips are very good at DSP type workloads, and therefore give superior performance for audio, video, and graphics processing, nominally the primary target market for Apple. Anything DSP will run faster on a PPC.
The AMD64 chips have one of the best memory subsystems you can buy and therefore give superior performance for any memory intensive application. Unlike PPCs strong DSP performance which is only useful in narrow application spaces, most applications can take advantage of a high-performance memory subsystem. In fact, this is the primary reason that AMD64 looks better clock-for-clock in most cases.
So it really boils down to picking the best tool for the job. In the GENERAL case, AMD64 is slightly faster. In the case of multimedia processing and other DSP-ish apps, PPC is slightly faster. For my work, we use Apple workstations as visualization front-ends to monster AMD64 back-ends. Because our codes exercise memory vigorously, the superior memory performance of the AMD64 means it runs rings around the G5s at the same clock speed for our most intensive workloads.
So when people say the G5 is faster than AMD64, they need to qualify that with "for DSP-ish codes". Otherwise, it isn't true.