PowerPC architecture is indeed a much better architecture.
Yes, it is. But if it the
CHIPS are not kept up to date, then it doesn't matter. A Mac Pro running native code blows the doors off a Power Mac G5 running native code. Even at the same clock speed and number of cores, the Mac Pro wins. (See linked article later in this post.) And an eight-core 3.0 GHz Mac Pro completely obliterates a quad-core 2.5 GHz Power Mac G5. And this disparity will only increase. IBM hasn't updated the PowerPC 970
at all since the dual-core models in 2005. Not even a speed increase. Yet Intel is about to release all-new models that are even faster than the present models. Heck, Apple was even 'overclocking' the G5. IBM's max specification calls for a processor bus of 1066 MHz (Apple used 1250 MHz in the dual-core and 1350 MHz in the single-core,) and a max core speed of 2.0 GHz (Apple used 2.5 GHz in the dual-core and 2.7 GHz in the single-core.)
The main difference between x86 and PowerPC is that x86 uses CISC and PowerPC uses RISC. (More information near the bottom.)
No, the main difference between the x86 and PowerPC is that the x86 uses the "IA" (for "Intel Architecture", also called IA-32, also called x86; the 64-bit variety called x86-64, AMD-64, or EM64T,) instruction set, while PowerPC uses the PowerPC instruction set. "Instruction sets" do not necessarily have ANYTHING to do with the core of the processor. All modern "x86" processors are really, at their core, RISC processors. The Transmeta Crusoe, and the original Intel Itanium chips are both "Very Long Instruction Word" (VLIW) cores, that are completely separate from either CISC or RISC, yet both can run "IA" code natively thanks to internal hardware translators. (The same way a Core 2 Duo internally translates IA code into its internal micro-ops. Only the Itanium was designed to natively run "IA-64" code, and the translation layer is only there for compatibility; while the Core 2 Duo has no mechanism for directly using its native instruction set.) And, heck, if Transmeta had wanted to do so, they could have had their processors emulate the PowerPC instruction set just as easily as the x86 instruction set.
If you are wondering why the inferior architecture is currently faster, you can thank your friends at Microsoft for monopolizing the operating system market and only supporting the x86 architecture. This gave Intel and AMD a huge lead in the desktop market in terms of cash flow.
Actually, the
market decided it, not Microsoft. Microsoft was pushing for Windows NT to be the OS of choice, specifically so that Microsoft wouldn't be limited to only running on Intel processors. Windows NT 4.0 was released to MIPS, PowerPC, DEC Alpha, and x86. Microsoft WANTED the variety. The only thing Microsoft cared about was that all computers run Windows NT, regardless of underlying hardware. During the life of the "NT" kernel (NT 3.1 through Windows XP/Server 2003,) it has been released in versions that run on
SIX different architectures. The above mentioned four, plus Itanium (Windows 2000, and XP/Server2003,) and x86-64 (XP/Server2003.) Heck, the NT kernel was originally designed for
Intel's RISC processor, the i860! (That processor was delayed significantly, so Microsoft also compiled for x86, and ended up never actually releasing an i860 version.)
Just to note, I am not trying to argue that the Intel Architecture instruction set is better than the PowerPC instruction set. I fully agree that PPC is technically superior to IA. The problem is that IA
CHIPS are so vastly superior technically to PPC chips that any instruction set superiority is rendered moot. (I don't care if a diesel engine is technically superior to a gasoline engine, the fact is that gasoline engine manufacturers have overcompensated to the point that diesel engines just don't compete, except in a few very tightly controlled circumstances. Although in this comparison, diesel is MUCH closer to matching and surpassing gasoline than PPC is to matching x86.)
When Mac OS X takes away enough of the market share from Microsoft's x86-only operating systems, hopefully there will be enough potential revenue for IBM/Motorola to continue PowerPC development for Apple computers. The more Mac users there are, the more money IBM/Motorola can make from selling PowerPC chips. This will surely lead to much faster chips and hopefully we can make the last major CISC architecture obsolete.
It is. Intel has known that since 1996, when they released the RISC-core Pentium Pro. (And all of its successors, which have also been RISC-core.) Mac users didn't lead IBM or Motorola to make faster chips before, why would it now? Motorola found PowerPC to be so minor a business that they spun off that division as Freescale Semiconductor, who doesn't even market a single PowerPC chip as being for a 'computer'. They're all marketed as being for 'appliances' or embedded uses. IBM gave up on general PC PowerPC (they don't even make one of their own anymore,) in favor of POWER-architecture big iron and video game consoles. Whey else do you think IBM wasn't able to deliver a 3.0 GHz single-core G5 for Apple when they were able to deliver a G5-derivative with THREE cores at 3.2 GHz for Microsoft for the Xbox 360?
Yes, RISC chips are indeed a significant improvement over CISC chips. Why do you think that RISC chips are the ones in the top super computers?
Have you actually
looked at the supercomputer rankings? The only reason RISC are on top any more is because of sheer number of processors. Compare processor count, not just total speed. An x86 system is 3rd, and has significantly fewer processors than the PowerPC system in 4th place.
Also look at SPEC Rates. These are the speed of the individual processor at Integer and Floating Point code. The fastest one right now for a single thread is the Intel Core 2 Duo E6850 (3.0 GHz,) with an Integer score of 22.6 vs. a
4.7 GHz IBM POWER6 at a score of 21.6. Next come Xeons at 3.0 GHz. For Floating Point, the POWER6 (again, at 4.7 GHz,) does win, with a score of 22.4, while the Core 2 Duo E6850 scores 19.3. (By comparison, a 1.6 GHz Itanium 2 scores 18.1! On a per-GHz basis, the Itanium wins.)
In the "Rate" tests, which can take advantage of multiple cores, the 4.7 GHz POWER6 does win, with a score of 122 compared to the 3 GHz Xeon at 116 in Integer, in floating point, it's 4.7 GHz POWER6, followed by 2.0 GHz Opteron, followed by 3.0 GHz Xeon.
Also take a peak at a direct comparison of
Power Mac G5 to Mac Pro, and notice that at equal clock speeds, the Mac Pro beats the Power Mac G5. (Even a 'crippled' single-socket dual-core 2.0 GHz Mac Pro beats a dual-socket 2.0 GHz G5.)
You do realize that the linked article is more than a decade out of date, right? It appears to have been written in 1994! (Before Intel released the Pentium Pro, which had, at its core, a
RISC core.)
will release and sell to the public.
