...well not really.

http://www.apple.com/pr/library/2005/jun/06intel.html

The deal is that Apple officially announced that between 2006 and 2007 it will make the transition of all Mac models from IBM Power PC processors to Intel CPUs.

Basically the whole architecture, hardware and software will be x86 based (x86 version of OS X too).

Kinda interesting isn't it?

My interpretation is that within two years the hardware will be basically the same as any brand PC, the difference will be only the design, quality of components and the operating system (OS X is a UNIX implementation, for those who aren't familiar with it).

While I don't consider myself an Apple fan (more of an Intel fan) I still don't get why on earth they wanna switch to x86 platform when the whole idea behind Macs is stability, immunity to viruses and high level of hardware-software compatibility due to the in-house built hardware.

There are plenty of speculations on the web, from the idea that IBM is too console happy (Sony PS 3) for Apple and ignores Apple's requests to the idea that Apple wants the DRM (Digital Rights Management) features of the new Pentium D processors in order to make a boom in the home entertainment market.

Intel's new plan for integrated hardware-DRM just made me even more a fan of AMD (assuming they don't follow suit).

I doubt Apple, or anyone else with a brain really, cares about the DRM stuff in any way other than as a checkbox thing for idiot suits caught on the meme. It's more likely that Apple simply wasn't a big enough customer for IBM to give them the sort of attention or pricing they needed. It does provide a nice "ah-ha!" with regard to what happened to Rhapsody on Intel back when.

I'm pretty bummed about it on principle, at risk of sounding like a RISC zealot PowerPC was sort of like the last guardian of the Roman Empire against the barbarian PC horde since Intel killed off Alpha. I know instruction-level architecture is fairly irrelevant these days but doing low-level stuff on Alpha and Power was simply a dream compared to x86 and that underlying design elegance really showed through in how with 1/50th the development budget they were faster than the Pentium family and far more power efficient.

Will be interesting to see how this plays out, if they can convince PC makers in general to support OpenBoot that would be the biggest system advance in the Wintel world since PCI and god knows BIOS is 25 years overdue for a bullet to the head.

RISC meant something versus CISC when RISC chips executed their instructions in fewer cycles than their CISC counterparts. Since nearly every instruction now executes in a single clock cycle (or multiple instructions per clock cycle), RISC doesn't have any sort of legitimate edge.

To illustrate, back in the 8086 days, it took something like 177 clock cycles do integer division. Now it takes one cycle. And, at the same time you're DIV AX BX, you can MOV BX CX too, making an effective HALF cycle if there's something the machine can do that's not dependant on AX. Early on, CISC machines took fewer than 177 cycles to do the same DIV. If a RISC chip could divide two numbers in a mere 50 clock cycles, your PC was 3x faster at math. Now, they're pretty much all 1 cycle. You get the picture.

Excuse my ignorance but aren't Pentium 4s basically RISC processors with a CISC layer on top? I could have sworn that P4s break down CISC instructions to 'RISC like' instructions (micro-operations or something) before executing them....

Yeah, I understand that, but simply having oh, 32 or 64 real registers, better integrated FP instructions, proper long word and BLT-type instructions plus a cleaner addressing system did add up. In reality the 8086 family of processors died with the Pentium, since the Pentium Pro and its descendants are all entirely blackboxed Pentium emulators, but it is noticable when you can cut through the 5 layers of abstractions gluing together 10 different design periods covering 15 years of hacking an 8-bit micro environment into something almost, but not quite approximating a Real Computer (tm) and watch what a coherently imagined system can do. RISC wasn't so much any given design as a design paradigm. The Alpha was probably its most pure offspring, I think Alpha was the closest I ever got to getting religion over technology :> Man could that thing fly.

From an end-of-the-day practical standpoint it doesn't really matter now, but I'm talking aesthetics here.


Excuse my ignorance but aren't Pentium 4s basically RISC processors with a CISC layer on top? I could have sworn that P4s break down CISC instructions to 'RISC like' instructions (micro-operations or something) before executing them..

More or less. The internal workings of the PPro/P4 family have never really been elaborated on to the public, as least as far as I know, but yes, what is actually executing in the core has no resemblance at all to the instructions stream you're feeding it, it just gives the same result. To a certain degree with out-of-order and speculative execution, multiple instruction pipelines, etc you could say that of a lot of things, but with most RISC systems it was far more evident to the programmer or compiler what was going to happen. I used to have a collection of great Alpha hacks that did unholy things with conditional evaluation to take biggest advantage of the 16 or whatever pipelines it had at that time, was funny stuff that hurt your head if you thought about it linearly :>

In reality the 8086 family of processors died with the Pentium, since the Pentium Pro and its descendants are all entirely blackboxed Pentium emulators, but it is noticable when you can cut through the 5 layers of abstractions gluing together 10 different design periods covering 15 years of hacking an 8-bit micro environment into something almost, but not quite approximating a Real Computer (tm) and watch what a coherently imagined system can do.


Well the name of the beast is "backward compatibility" which is quite misleading since half of the software written 3 platform generations ago will simply refuse to work properly on the current one.

Excuse my ignorance

The line between RISC and CISC is entirely blurry these days. I think every major CPU has both RISC and CISC attributes.

Again, it only mattered so much when your reduced instruction set meant that you could process those limited instructions more efficiently - and your compiler (or your talented programmer) could take advantage of it.

Now now, it also mattered when ridiculing Mac zealots.

The line between RISC and CISC is entirely blurry these days.

Not at all. RISC is not a how-to-make-a-chip thing, it's a design paradigm. There is no blur at all between RISC principles and CISC ones, either you have uniformly structured instruction and operand sets, or you don't. You have large numbers of general purpose registers or you have special purpose ones. Ad nauseum. In effect you either have a processor conceived before or after 1989, or one from later that's faking it. All of the nifty new stuff in processors has been the children of RISC -- out of order execution, multiple pipelines, speculative execution, all of these depend on being able to keep the instruction stream as atomic and predictable as possible. You'd be taken out back and shot for an instruction with a variable number of operands in hard-coded registers now.

In terms of actual implementations though, the distinction has become largely irrelevant since the only real CISC candidate left is in fact an emulator with a theoretically RISC core. Likely the only true CISC processors still made are 680x0 variants in various embedded domains, but even those have likely been replaced with even more power/space efficient ARM and POWER chips in anything designed in the last ten years. That RISC was fundamentally superior is taken for granted now because nobody designs with the CISC mindset anymore -- graduates of processor design schools today would probably have a heart attack if they saw a VAX.

But it's like anything else political, any sort of logical understanding of the issue has long been brushed aside for punditry.

(speaking of punditry, I'd sell out my own mother to see BIOS get axed in favor of OpenBoot though)

If it's of any consolation, I'd sell Malse' mother too for OpenBoot.

In terms of actual implementations though, the distinction has become largely irrelevant

heh. You disagreed with me then you agreed with me. As I said: http://en.wikipedia.org/wiki/CISC