AMD vs Intel

Arthur_Tuxedo · Post by **Arthur_Tuxedo** » 2003-04-10 04:14pm

I sure would like to see an x86 version of the PPC chip.

Lord Pounder · Post by **Lord Pounder** » 2003-04-10 04:18pm

my 1st PC was an Intel. I leaned since and every chips since, including this lovely 2200XP i got today, is AMD.

phongn · Post by **phongn** » 2003-04-10 04:29pm

Arthur_Tuxedo wrote:I sure would like to see an x86 version of the PPC chip.

That's an oxymoron.

Durandal · Post by **Durandal** » 2003-04-10 05:08pm

Arthur_Tuxedo wrote:I sure would like to see an x86 version of the PPC chip.

And I'd sure like to see a car version of a train.

phongn · Post by **phongn** » 2003-04-10 06:10pm

Durandal wrote:
Arthur_Tuxedo wrote:I sure would like to see an x86 version of the PPC chip.
And I'd sure like to see a car version of a train.

Well, to be nice to Arthur, there was a planned dual-core version of the 604e that would have had an x86 core on it as well, but that never went anywhere.

Arthur_Tuxedo · Post by **Arthur_Tuxedo** » 2003-04-15 04:04am

As usual, I've made myself a victim of not being specific enough. I'd like to see Motorolla design an x86 chip using the principles of the PPC. That better?

aphexmonster · Post by **aphexmonster** » 2003-04-15 11:25am

im happy with my p1 133

... okay, well maybe not really.

phongn · Post by **phongn** » 2003-04-15 11:57am

Arthur_Tuxedo wrote:As usual, I've made myself a victim of not being specific enough. I'd like to see Motorolla design an x86 chip using the principles of the PPC. That better?

I'm not sure if that's really possible.

Slartibartfast · Post by **Slartibartfast** » 2003-04-15 12:02pm

PPC is RISC, which is a completely different architecture.
In fact PPC and 68k (previous Macs) aren't even the same tech, there's some kind of ultra-efficient emulator there, or a subroutine of 68k instructions in the PPC chip or whatever. If you really wanted to make a RISC PC, you would have to do basically the same: screw backwards compatibility (which could be good... with Intel it would definitely be an improvement

) and at the same time make a very good emulator of the old 386.

phongn · Post by **phongn** » 2003-04-15 12:07pm

Slartibartfast wrote:PPC is RISC, which is a completely different architecture.
In fact PPC and 68k (previous Macs) aren't even the same tech, there's some kind of ultra-efficient emulator there, or a subroutine of 68k instructions in the PPC chip or whatever. If you really wanted to make a RISC PC, you would have to do basically the same: screw backwards compatibility (which could be good... with Intel it would definitely be an improvement ) and at the same time make a very good emulator of the old 386.

PPC has a 68k emulator onboard but it isn't perfect (there are some issues).

The whole RISC/CISC thing is kinda blurred now. PPC has CISC instructions (i.e. AltiVec), while the whole K7 line is really closer to a RISC design (it breaks down x86 instructions).

Arthur_Tuxedo · Post by **Arthur_Tuxedo** » 2003-04-15 05:12pm

So why is it not possible if not for the RISC issue? (besides, that shouldn't be relevant anyway, Alpha has been designing x86 RISC chips for a long time, have they not?)

phongn · Post by **phongn** » 2003-04-15 05:55pm

Arthur_Tuxedo wrote:So why is it not possible if not for the RISC issue? (besides, that shouldn't be relevant anyway, Alpha has been designing x86 RISC chips for a long time, have they not?)

The design philosophy of PPC chips are geared to a much different markets than x86. Motorola concentrates mostly on parts for integrated applications, such as routers. The chips that Apple use are a byproduct of this. IBM concentrates mostly on high-end parts for servers (POWER4, for example) and also competes with Motorola for integrated stuff. Neither of them are exactly where x86 is.

IBM actually did make a x86 chip before - it would serve as the basis for AMD's K7, and it did not resemble the PPC.

DEC did not make x86 chips. Alpha was Alpha; it ran x86 code in emulation. Windows NT 4 had an Alpha port to it (back when Microsoft was targetting a bunch of different platforms, including PPC).

Durandal · Post by **Durandal** » 2003-04-15 06:15pm

Arthur_Tuxedo wrote:As usual, I've made myself a victim of not being specific enough. I'd like to see Motorolla design an x86 chip using the principles of the PPC. That better?

Motorola's design philosophies are almost totally different from those which have shown to be successful in the desktop market. While the desktop market is quick to adopt new technologies, Motorola designs CPU's for embedded and networking applications. In other words, applications which demand that the CPU be good enough, low-power and working. The desktop market wants the CPU to be the fastest and isn't terribly concerned with power usage unless it uses so much power that the core melts.

If you want to know why the G4 has been stuck on a 133 MHz frontside bus while making slow progress up the clockspeed scale, the answer is simply that Motorola's chief PowerPC customers just don't need the chip to be any faster than it is. They need it to be cheaper and cooler; it's fast enough. Apple were stupid enough to lock themselves in with Motorola and it's downward spiral in the desktop CPU sector because of their hard-on for AltiVec, and now we have Apple trying to make embedded CPU's work for the desktop and compete with Intel and AMD, who are locked in a speed competition with each other and stepping up the GHz ladder at a rate that makes El Stevo cream his pants. So, you have Steve Jobs pumping more and more power into the Reality Distortion Field to make it seem like a dual 1.42 GHz machine with a 167 MHz frontside bus for $3000 is anything but a complete joke in the year 2003.

Arthur_Tuxedo · Post by **Arthur_Tuxedo** » 2003-04-15 06:30pm

I was under the impression that PPC's were the fastest chips since they were so much better pound-for-pound even though clockspeed was lower. Anyway, that was the situation a couple years ago and I just assumed it hadn't changed. Guess I need to keep up with the times, eh?

phongn · Post by **phongn** » 2003-04-15 06:35pm

Arthur_Tuxedo wrote:I was under the impression that PPC's were the fastest chips since they were so much better pound-for-pound even though clockspeed was lower. Anyway, that was the situation a couple years ago and I just assumed it hadn't changed. Guess I need to keep up with the times, eh?

In terms of IPC, the G4 has generally kept up with the K7. However, in terms of clockspeed it has been grossly falling behind.

Durandal · Post by **Durandal** » 2003-04-15 07:16pm

Arthur_Tuxedo wrote:I was under the impression that PPC's were the fastest chips since they were so much better pound-for-pound even though clockspeed was lower. Anyway, that was the situation a couple years ago and I just assumed it hadn't changed. Guess I need to keep up with the times, eh?

As I said earlier, the performance/clockrate ratio is now irrelevant. The G4 is, by far, the most efficient chip out there. It's reached 1.4 GHz with 7 pipeline stages and a 0.18µ fabrication process. This is certainly nothing to scoff at, considering the low power usage of the chip; you'll notice that both AMD and Intel have special versions of their flagship processors for laptops, while Apple uses the same processors in their laptops as they do in desktops. Add to that the fact that the die still has room for a L3 cache and SIMD unit, and it's a pretty remarkable piece of engineering, and even the most virulent PC modder/Mac-hater can't deny that. Neither AMD or Intel could take the core as far as Motorola has.

The problem? Pentium 4's are at 3 GHz (more than enough to eclipse their horrid efficiency), on faster buses and are being sold for much cheaper. When the Pentium 4 launched, people were on the floor laughing at how horridly slow it was, even at 1.5 GHz and predicting that AMD would have no problem consistently pounding the Pentium 4 into the ground. Well, now Intel is having the last laugh. They designed the Pentium 4 to scale preposterously far, and it's paid off in a big way. Compare a 1.4 GHz Pentium 4 to a 1.4 GHz G4, and the Pentium 4 will get its ass handed to it. A 3 GHz Pentium 4, on the other hand, will stomp the G4, and do it while being half the price. Price/performance does the talking. Efficiency means jackshit to the end user.

phongn · Post by **phongn** » 2003-04-15 07:18pm

Hey, you forgot about Banias, which is designed from the ground up to be a relatively low-power laptop processor. It's quite the design.

Durandal · Post by **Durandal** » 2003-04-15 08:51pm

phongn wrote:Hey, you forgot about Banias, which is designed from the ground up to be a relatively low-power laptop processor. It's quite the design.

Yeah, I neglected Centrino mostly because we're comparing Pentiums to G4's.

Slartibartfast · Post by **Slartibartfast** » 2003-04-16 01:12am

phongn wrote:
Slartibartfast wrote:PPC is RISC, which is a completely different architecture.
In fact PPC and 68k (previous Macs) aren't even the same tech, there's some kind of ultra-efficient emulator there, or a subroutine of 68k instructions in the PPC chip or whatever. If you really wanted to make a RISC PC, you would have to do basically the same: screw backwards compatibility (which could be good... with Intel it would definitely be an improvement ) and at the same time make a very good emulator of the old 386.
PPC has a 68k emulator onboard but it isn't perfect (there are some issues).

I know its not perfect but what it *does* manage to emulate (most System 7.something stuff and upwards) it does very well, much better than software emulation.

The whole RISC/CISC thing is kinda blurred now. PPC has CISC instructions (i.e. AltiVec), while the whole K7 line is really closer to a RISC design (it breaks down x86 instructions).

Well I'm not that well versed on that, just that x86 instructions with their retarded use of expanded memory and too big amount of opcodes makes it inefficient. If K7 are sorta like a PPC for PCs, then so be it.

Slartibartfast · Post by **Slartibartfast** » 2003-04-16 01:14am

phongn wrote:IBM actually did make a x86 chip before - it would serve as the basis for AMD's K7, and it did not resemble the PPC.

I had one of those, I think they were called 486SLC, which was sortof like a 486 but not as fast...

phongn · Post by **phongn** » 2003-04-16 09:20am

Slartibartfast wrote:
phongn wrote:IBM actually did make a x86 chip before - it would serve as the basis for AMD's K7, and it did not resemble the PPC.
I had one of those, I think they were called 486SLC, which was sortof like a 486 but not as fast...

NexGen was IBM's spinoff - AMD acquired them later.

EDIT: nevermind, I'm wrong. I'll look up what happened to IBM's x86 division.

phongn · Post by **phongn** » 2003-04-16 09:24am

Slartibartfast wrote:I know its not perfect but what it *does* manage to emulate (most System 7.something stuff and upwards) it does very well, much better than software emulation.

Well, it does do it fast, though for awhile we had to deal with fat binaries with both 68k and PPC code.

The whole RISC/CISC thing is kinda blurred now. PPC has CISC instructions (i.e. AltiVec), while the whole K7 line is really closer to a RISC design (it breaks down x86 instructions).
Well I'm not that well versed on that, just that x86 instructions with their retarded use of expanded memory and too big amount of opcodes makes it inefficient. If K7 are sorta like a PPC for PCs, then so be it.

K7 is not 'sorta' like PPC for PC. It's simply a RISC-like processor (which a bunch of complex instructions added on).

GrandMasterTerwynn · Post by **GrandMasterTerwynn** » 2003-04-16 02:30pm

Arthur_Tuxedo wrote:As usual, I've made myself a victim of not being specific enough. I'd like to see Motorolla design an x86 chip using the principles of the PPC. That better? :)

And I would like Kawasaki to design a motorcycle that weighs three tons and runs on train tracks. The x86 architecture is the epitome of CISC. It's carrying a massive amount of deadweight that extends clear back to the original 8088/8086 chips. It's got instructions that do just about everything except the laundry.

If you want lean-and-mean RISC performance, then you should purchase a lean-and-mean RISC chip. Trying to completely redesign the x86 architecture is going to result in a lot of issues with software incompatibility and whatnot.