Computer processor architectures (G5, G4, Intel and AMD)

[tech4all]

I was thinking about Apple going Intel, and just had a thought today.

Currently the G5 Macs are on a PowerPC architecture, right? And Intel and AMD are on the x86 architecture, right? Then what architecture is the G4 processor on? I mean the G5 and G4 are made by different companies, so do they both use the same architecture or different ones?

 

[mad jew]

Still PowerPC, just as G3s were too.

 

[iGary]

I was thinking about Apple going Intel, and just had a thought today.

Currently the G5 Macs are on a PowerPC architecture, right? And Intel and AMD are on the x86 architecture, right? Then what architecture is the G4 processor on? I mean the G5 and G4 are made by different companies, so do they both use the same architecture or different ones?

G3's and G4's are PPC.

 

[tech4all]

Oh so PowerPC isn't just specific to IBM but rather Macs (their logic/mother board) in general? All the companies that produce processors for Apple all run on the PowerPC architecture?

Hehe, and hear I though "PowerPC" was the technical name to the G5 processor So I guess the G5, G4, and G3 processor are really only called G5, G4, and G3....learn something new everyday


Thanks!

 

[mad jew]

Think of PowerPc as an x86 equivalent and then G5, G4 and G3 and P4, P3 alternatives if it helps.

Also, to the best of my knowledge, only Motorola and IBM have the licence to build PPC chips. This, I think, stems back to an alliance with Apple called AIM but it was a bit before my time as an Apple fanatic so I'm not positive on this one.

 

[superbovine]

http://en.wikipedia.org/wiki/Powerpc

 

[ravenvii]

You're so wrong on so many levels...

The architecture is the "type" of CPU it is - x86/CISC, PowerPC/RISC, MIPS, and several others. The PowerPC is not specific to Apple - many embedded devices use PowerPC chips. The G3, G4 and the G5 is just Apple's marketing name for those processors. The G5 is the PPC 970, the G4 are actually several chips, including the PPC 7447 and 7457. The G3 is actually several chips, originally made by Motorola, then later made by IBM. And Apple was not always using PowerPC chips either - before PowerPC, Apple was using Motorola's 680xx chips. I'm not sure what architecture it was, but it wasn't PowerPC- PowerPC didn't even exist at the time.

 

[jcgerm]

And Apple was not always using PowerPC chips either - before PowerPC, Apple was using Motorola's 680xx chips. I'm not sure what architecture it was, but it wasn't PowerPC- PowerPC didn't even exist at the time.

The 68k series was CISC and I believe the architecture is just 68k.

Currently the G5 Macs are on a PowerPC architecture, right? And Intel and AMD are on the x86 architecture, right? Then what architecture is the G4 processor on? I mean the G5 and G4 are made by different companies, so do they both use the same architecture or different ones?

Among other things, what makes two different models of processor the same architecture (x86, PPC, etc.) is really the instruction set. There's usually a base set of instructions and different sets added on to that, such as MMX and SSE on x86, per generation. Past that, the processors different in hardware aspects such as bus sizes, register sizes, cell placement, pin out, etc.

So actually (I'm being somewhat anal retentive here), what everyone here refers to as x86 is actually x86-32 or x86-64. Plain old x86 died out when Intel introduced the 386 CPU in the 80's.

 

[mad jew]

So actually (I'm being somewhat anal retentive here), what everyone here refers to as x86 is actually x86-32 or x86-64. Plain old x86 died out when Intel introduced the 386 CPU in the 80's.

Thanks for that jcgerm. Is the x86-64 about as similar to the x86-32 as the G5 is to the G4? I'm pretty ignorant when it comes to this kind of stuff. Is the only difference the 64-bit part or is there much much more to it?

 

[im_to_hyper]

This is kinda deviating here....

Prior to the G3 there were PowerPC processors as well. I know that there were the 603e and the 604 PowerPC processors. Computers with these CAN run OS X with the help of XPostFacto since they are based on the same architecture as the G5, etc.

I think of the 603x processors being G1 and the 604 being G2... after that came the G3, which I believe literally stands for Generation 3... it was the third generation PowerPC processor.

Someone correct me if I am way off here... its half past midnight

 

[ravenvii]

This is kinda deviating here....

Prior to the G3 there were PowerPC processors as well. I know that there were the 603e and the 604 PowerPC processors. Computers with these CAN run OS X with the help of XPostFacto since they are based on the same architecture as the G5, etc.

I think of the 603x processors being G1 and the 604 being G2... after that came the G3, which I believe literally stands for Generation 3... it was the third generation PowerPC processor.

Someone correct me if I am way off here... its half past midnight

Nope you're not way off - you're right on the money.

To MadJew: yeah, that's pretty much the idea. The (slight in reality) difference is that the G5 is a 64-bit processor that handles 32-bit code, and the x86-64 processors are 32-bit processors that can handle 64-bit code. Intel was pushing the Itanium, or IA-64 processors which are true 64-bit descentant to x86, but unfortunately for Intel, it handles 32-bit code with crapcular results, and guess what 99% of the world uses. 64-bit doesn't really matter yet, and won't for a long while - except for the added memory addressing, of course. Overhype to the max.

 

[whocares]

Prior to the G3 there were PowerPC processors as well. I know that there were the 603e and the 604 PowerPC processors. Computers with these CAN run OS X with the help of XPostFacto since they are based on the same architecture as the G5, etc.

I think of the 603x processors being G1 and the 604 being G2... after that came the G3, which I believe literally stands for Generation 3... it was the third generation PowerPC processor.

Don't forget the 601. IIRC twas the chip in the very first PPC, or the "G1".

 

[Nermal]

Yeah the 601 was the "G1" and the 603 and 604 were the "G2". 750 series is G3, 7400 series is G4 and 970 series is G5.

If I recall correctly, there were only 3 or 4 models released with 601, before Apple jumped to 603. Also, while we're talking about history, the "Power" in "PowerBook" doesn't mean PowerPC, as there were 68K PowerBooks too.

 

[Platform]

G1=PPC 601 series [IBM/Motorola]
G2=PPC 603/603e and 604/604e series [IBM/Motorola]
G3=PPC 75x series [IBM]
G4=PPC 7xxx series [Motorla/Freescale]
G5=PPC 970 series [IBM]

 

[sacear]

G(x) is Apple's moniker in reference to generation of Power Macintosh

"Talking 'bout Mac Generations..."

The Gx series moniker as used by Apple refers to the generation of "Power Macintosh," describing the computer model generation. Apple secondarily named the CPU (not the PowerPC chip) after the machine. "Power Macintosh G3" means third-generation Power Macintosh, not G3 CPU Power Macintosh, nor Power Macintosh with a G3 CPU. The CPU is named after and according to the Power Macintosh. The Power Macintosh is not named after and according to the CPU.

An Apple Power Mac G3 CPU uses a Power Performance Computing 750 processor chip. The chip was not the Gx. Seems the PowerPC 970 chip is the first to actually be labeled G5, along with CPU, and the Power Mac.

Apple's current generation moniker (G3, G4, & G5) marks the generations of Power Macintosh computers, then the PowerPC processor based CPUs were named to follow suit.

The moniker Gx is Apple's and refers to the Power Macintosh model first and predominantly, and secondarily to the CPU. The CPU was named after the Power Mac model moniker. At this particular time the fifth generation Power Mac just happens to coincide with the fifth generation of PowerPC.

After the Power Mac G3, the naming of the CPU seems to have taken over importance as the Gx and gone full circle with PowerBooks and iMacs now being named after the CPU, which was originally named after the generation of Power Mac.

Similar to how the "Power" Performance Computing chip was named after Apple's already existing Power moniker and planned Power desktop models, yet now people think Power named Macintoshes are named after the PowerPC chip.
Actually, that was the other way around.

Gx is Apple's moniker, not IBM's or Motorola's. The naming convention for Power Mac models (Gx or otherwise) is independent of the actual processor. Gx is the sub-model (generation) name of the machine and the name of the CPU (not the name of the chip itself as manufactured) as given by Apple, not IBM or Motorola (now Freescale).

The first PPC (601 series) Mac models were called Power Macintosh (1994-1998). The sub-models being 6100, 7100, 7200, 7500, 8100 and 8200. The PPC 601series was the first generation PowerPC. The second generation consisted of the PPC 603 series and the PPC 604 series, Mac models 44xx, 5xxx, 6xxx, 7xxx, 8xxx, and 9xxx. PPC 750 (Apple's G3) then being the third generation.

G3 was the third generation of the PPC chipset CPU and the third generation of Power Macintosh, now called Power Mac (please note the space in the name). Then continuing accordingly.

"...said Mac generations baby."

 

[robbieduncan]

Thanks for that jcgerm. Is the x86-64 about as similar to the x86-32 as the G5 is to the G4? I'm pretty ignorant when it comes to this kind of stuff. Is the only difference the 64-bit part or is there much much more to it?

Sort of. The PowerPC architecture was alwasy designed to scale to 64 bits so the 64bit instruction set on PowerPC is very natural. x86 was not reall designed for 64bit so it's kind of shoehorned in there and a bit of a hack.

Both acheive the same thing: the ability to address more RAM than you are likely to ever own!

 

[jcgerm]

Thanks for that jcgerm. Is the x86-64 about as similar to the x86-32 as the G5 is to the G4? I'm pretty ignorant when it comes to this kind of stuff. Is the only difference the 64-bit part or is there much much more to it?

64-bit is the defining factor. The gist of the transition to the next generation of CPU in both cases is 64-bit addressing (being able to access more than 4 GB of RAM) and compatibility with 32-bit applications (i.e. everything that runs 32-bit). But, at the same time, the G5 and x86-64 aren't just the previous generation's processor + 64-bit addressing. Lots of optimizations happened between generations (i.e removal of some assembly instructions, change in CPU layout, register sizes, etc). It's best to think that the G5 is based on of the G4 and x86-64 is based on x86-32. On the surface, they're very similar, but under the hood, they're very different (x86-32 vs. x86-64 and G4 vs. G5). That way, the transition between processor generations doesn't give everyone a headache .

 

[robbieduncan]

It's best to think that the G5 is based on of the G4 and x86-64 is based on x86-32.

This is not a good way to think about anything. This is taking comparing a specific processor (well familiy of processors) made by IBM with another made by FreeScale (so not that similar) and saying it's simialar to comparing on instruction set (x86-64) with another that is a strict subset of it.

The G5 is not based on the G4, although they both implement the PowerPC architecture.

 

[jcgerm]

This is not a good way to think about anything. This is taking comparing a specific processor (well familiy of processors) made by IBM with another made by FreeScale (so not that similar) and saying it's simialar to comparing on instruction set (x86-64) with another that is a strict subset of it.

The G5 is not based on the G4, although they both implement the PowerPC architecture.

True. I should have said the PPC 970 is based on IBM's POWER4 architecture. Saying that the G5 is based on the G4 would be the same as saying that the Athlon 64 is based on the Pentium 4. Sorry about that.

That said, does anyone know is there is actually a name for the instruction set that the G5 implements, or is it just referred to as PPC with 64 bit extensions?

 

[robbieduncan]

True. I should have said the PPC 970 is based on IBM's POWER4 architecture. Saying that the G5 is based on the G4 would be the same as saying that the Athlon 64 is based on the Pentium 4. Sorry about that.

That said, does anyone know is there is actually a name for the instruction set that the G5 implements, or is it just referred to as PPC with 64 bit extensions?

They are not really extensions. The PPC architecture was designed from the ground up (all those years ago) with the 64-bit instructions in place. I believe (although could be very wrong) that chips with only the 32 bit instructions implement PPC32 and the G5 implments PPC64. The Power ISA is a little different but is similar to PPC64.

Good History of PowerPC

 

[jcgerm]

They are not really extensions. The PPC architecture was designed from the ground up (all those years ago) with the 64-bit instructions in place. I believe (although could be very wrong) that chips with only the 32 bit instructions implement PPC32 and the G5 implments PPC64. The Power ISA is a little different but is similar to PPC64.

Good History of PowerPC

Interesting. I hadn't realized PPC was designed that way.

 

[Gizmotoy]

Interesting. I hadn't realized PPC was designed that way.

Yes, and really it is one of the reasons the PPC is viewed by many in the community (I'm a microprocessor designer) as a far superior chip. You see, over the years the x86 instruction set has been hacked onto and built up to maintain backwards compatibility at all costs. The x86 architecture is a complete mess of outdated and mostly unused components. Hell, it used to be a 16-bit chip! This is its second major architecture change. A major problem with the x86 spec is that it allows for "buffer overflow" exploits, which you may be familiar with, that allow the execution of data instead of code. This is expressly prohibited in the PPC design, a major security advantage. Fortunately, I believe both AMD and Intel have either recently or plan to release a solution for this problem. I have not been paying much attention to that area lately.

Intel, for their part, did a commendable job in attempting to jettison the legacy support... but industry did not follow due to the AMD64 providing backwards compatibility. Without AMD, the 64-bit chips would have been starting from scratch (and incompatible with all current chips). Someone earlier mentioned the IA-64, which is a truly remarkable chip. I had a chance to do some analysis on its design as it was developed almost entirely by ex DEC employees who graduated from the University of Cincinnati, my college. Unfortunately, it was just too radical and didn't catch on. Ah, backwards compatibility - so useful, yet such a pain.

 

[punkbass25]

Someone earlier mentioned the IA-64, which is a truly remarkable chip. I had a chance to do some analysis on its design as it was developed almost entirely by ex DEC employees who graduated from the University of Cincinnati, my college. Unfortunately, it was just too radical and didn't catch on. Ah, backwards compatibility - so useful, yet such a pain.

unless your computer exsists in a solely 64bit environment (which to my knowledge is extremely rare) then the itanium sucks, it is totally incapable of ddoing simple things if the coding is 32 bit..and worse it doesnt just not run teh file, instead it craps out and crashes.... i don't know how you could quallify this as "good"... but hey to eacht here own.. and im pretty sure the DEC designers did not make itanium, but ARE making a new 64 bit chip likely to appear in powermacs, in two years or so.

please correct me if i'm wrong

 

[robbieduncan]

A major problem with the x86 spec is that it allows for "buffer overflow" exploits, which you may be familiar with, that allow the execution of data instead of code. This is expressly prohibited in the PPC design, a major security advantage. Fortunately, I believe both AMD and Intel have either recently or plan to release a solution for this problem. I have not been paying much attention to that area lately.

Current Intel and AMD chips feature NX (No eXecute) bits which fix this when used with a supporting OS (Win XP SP2 for example).

 

[Gizmotoy]

unless your computer exsists in a solely 64bit environment (which to my knowledge is extremely rare) then the itanium sucks, it is totally incapable of ddoing simple things if the coding is 32 bit..and worse it doesnt just not run teh file, instead it craps out and crashes.... i don't know how you could quallify this as "good"... but hey to eacht here own.. and im pretty sure the DEC designers did not make itanium, but ARE making a new 64 bit chip likely to appear in powermacs, in two years or so.

please correct me if i'm wrong

You're definately wrong, because we had access to the designers during the process. Anyway, the point of my post had nothing to do with your rant. The Itanium is a superbly-designed chip and was not designed to run 32bit code. It's supposed to reside in a solely 64bit environment. Sure you can complain that it can't run 32bit code properly, but why? That's like complaining the PPC can't run x86 code properly because the endian notation is opposite. It is good at what it was designed for, and is in fact pretty elegant in silicon. It's an impressive chip. That said, being an impressive chip has nothing to do with real-world usability. You're looking at the chip from the angle "What can this thing do for me" and I was talking about it from the "What is its design like" angle. I'm not surprised it was a commercial flop, because it didn't provide what the customer wanted. That in no way changes my opinion of it. The chip was merely ahead of its time, and was brought down by the lack of 64bit software and a competitor's pledge for backwards compatibility.

As a note, yes, the original Itanium development team has been moved onto the main Pentium roadmap. It is likely one of the chips they assist on now will make it into a future Mac (they're still doing the 64 work, no 32bit chips).

Current Intel and AMD chips feature NX (No eXecute) bits which fix this when used with a supporting OS (Win XP SP2 for example).

Ah, thanks for the update. I knew they were working on something, but I didn't know if it was in consumer hands yet or not. I remember seeing a sticker for "Virus Protection" on the newer AMD64 components, which now that you bring it to my attention is probably what it was referring to. The sticker said it requires XP SP2. As you may have notice, my area of expertise is not the x86 That was a big hole. I'm glad its finally patched.