19-Apr-2024 05:55 GMT.		 UNDER CONSTRUCTION
Anonymous, there are 38 items in your selection
[Web] Analysis: x86 Vs PPCANN.lu
Posted on 09-Jul-2003 21:21 GMT by Hagge38 comments
View flat
View list
It's up, created by Nicholas Blachford, read it all on osnews.
Analysis: x86 Vs PPC : Comment 1 of 38ANN.lu
Posted by Matt Parsons on 09-Jul-2003 19:50 GMT
I would suggest a visit to here both before and after reading the article on OSNews:

http://arstechnica.com/cpu/index.html

Arstechnica is not opinions it's simple facts (though requires quite a bit of technical knowledge). Reading the Arstechnica articles should clear up deficiencies in the OSNews article regarding both the x86 and the PPC.
Analysis: x86 Vs PPC : Comment 2 of 38ANN.lu
Posted by Don Cox on 09-Jul-2003 20:11 GMT
In reply to Comment 1 (Matt Parsons):
This one is worth reading too:

http://www.igeek.com/articles/Hardware/Processors/IBMs970.txt

There is no such thing as "simple facts".
Analysis: x86 Vs PPC : Comment 3 of 38ANN.lu
Posted by takemehomegrandma on 09-Jul-2003 20:24 GMT
It was a very nice article IMO.
Analysis: x86 Vs PPC : Comment 4 of 38ANN.lu
Posted by corpse on 09-Jul-2003 20:41 GMT
In reply to Comment 2 (Don Cox):
I like this "simple fact" from the pegasos website:

"What is Linux?, A derivative of UNIX,"

Hmm I didn't know Linux was derived from UNIX(TM), but I'm sure this is "old news" ;)
Analysis: x86 Vs PPC : Comment 5 of 38ANN.lu
Posted by corpse on 09-Jul-2003 21:03 GMT
"Remember that the x86 only has 8 registers"

The original 8086 has eight 16bit registers (AX,BX,CX,DX,SI,DI etc) the X's being general purpose most of the time, but I'm sure a more were added with the 32bit 8386 and some more with the 8586's (pentium) along with protected mode, FPU integration, MMX etc
Analysis: x86 Vs PPC : Comment 6 of 38ANN.lu
Posted by Elektro on 09-Jul-2003 21:26 GMT
Opteron ain't mentioned...
Analysis: x86 Vs PPC : Comment 7 of 38ANN.lu
Posted by Christophe Decanini on 09-Jul-2003 21:33 GMT
In reply to Comment 6 (Elektro):
Read again.

@ALL
And also remember this quote "We produce PowerPC based systems and are often asked why we use PowerPC CPUs instead of x86 so a comparison is rather useful. "

Obviously when you are asked why using your product over another one you enlight your products advantages. Nicholas did it and it did it well.
Analysis: x86 Vs PPC : Comment 8 of 38ANN.lu
Posted by Lando on 09-Jul-2003 22:30 GMT
In reply to Comment 5 (corpse):
>The original 8086 has eight 16bit registers (AX,BX,CX,DX,SI,DI etc) the X's being general purpose >most of the time, but I'm sure a more were added with the 32bit 8386

EAX,EBX,ECX... etc
Still only 8 regs, just fit more in 'em.
Analysis: x86 Vs PPC : Comment 9 of 38ANN.lu
Posted by Johan Rönnblom on 09-Jul-2003 22:54 GMT
In reply to Comment 4 (corpse):
Read some interview with Linux Torvalds about how and why he started
Linux, and you'll see?
Analysis: x86 Vs PPC : Comment 10 of 38ANN.lu
Posted by Anonymous on 09-Jul-2003 23:00 GMT
In reply to Comment 5 (corpse):
Erm, you are "sure"? You don't know it actually? What is your point in posting then? Just playing the smart-ass?
Analysis: x86 Vs PPC : Comment 11 of 38ANN.lu
Posted by corpse on 09-Jul-2003 23:47 GMT
In reply to Comment 8 (Lando):
ah, so they were just wider again..

I stand corrected :)
Analysis: x86 Vs PPC : Comment 12 of 38ANN.lu
Posted by corpse on 09-Jul-2003 23:50 GMT
In reply to Comment 9 (Johan Rönnblom):
"Read some interview with Linux Torvalds about how and why he started
Linux, and you'll see?"

Because he wanted an OS where men where men and wrote their own device drivers... no Linux is not derived for UNIX, its a clone but in no way distilled from it. If the same rules applied for the Linux paragraph the one above should mark MorphOs as an AmigaOS derivative which it clearly isn't.
Analysis: x86 Vs PPC : Comment 13 of 38ANN.lu
Posted by corpse on 09-Jul-2003 23:58 GMT
In reply to Comment 10 (Anonymous):
"Erm, you are "sure"? You don't know it actually? What is your point in posting then? Just playing the smart-ass?"

I did a little 32bit 86 assembly after doing a long period of 16bit 86 assembly .. Anyone thats done some assembly will know that (on a 86 atleast)you can use the registers in segments instead of as a whole (because of the way the 86 has scaled up 8bit->16bit->32bit) so if you go from 16bit to 32bit it can seem like you have double the registers ... combine this with the fact that my 86 assembly dabbling ended a long time ago and I've since been working with 8051 and 16bit PIC assembly intensively .. well its an easy mistake ;)
Analysis: x86 Vs PPC : Comment 14 of 38ANN.lu
Posted by Chain|Q on 10-Jul-2003 05:18 GMT
In reply to Comment 5 (corpse):
Actually, the 80386 added no more general purpose registers, just added two more segment registers (FS, GS) and extended the existing register set to 32bit. It also removed some limits in how can a register be used. Anyway, i still hardly call any of the x86 registers "general purpose". As using 68k terminology, the x86 has 4 'data' registers (EAX-EDX) and four 'address' registers (ESI,EDI,EBP,ESP). The most important difference is, just like on 68k, you can't work with the lower byte of an address register directly. And there are much more limitations in, like you must use some registers for some instructions. For example move.l d0,(a0)+ is simply not exists on x86. There is an instruction which does pretty much the same (STOSD), but it has no operands, so always uses EAX as source, and ES:EDI as target address. Bye-bye general purpose registers... Or you can use MOV [EDI],EAX of course, but in this case you have to increase the address register by hand, doing an ADD EDI,4... These are small but important differences, when you are talking about the effectiveness of an architecture.

Anyway, the most painful part of the x86 architecture is the FPU, since the x86 has no linear address space, but it's some weird kind of a stack. So on x86, you just can't fmove.l (a0),fp3, and then fmul fp2,fp3, because you only can move values to the top of the stack (all values moves down then) and you can only make operations on the value currently at the top of the stack. Of course there is an instruction, which swaps two values in the stack, but this makes overhead simply not exists on architectures using linear FPU register set. Other difference, compared to the 68k for example, that you cannot move values directly between the CPU and the FPU, you have to store the value to the memory first, and then move to the FPU. Needless to say, this makes unnecessary overhead on the cache and memory interface, and in worst case on the whole bus system. So this makes FPU coding on x86 difficult, and also makes the FPU unnecessarily complex, and ineffective. This is why most architectures beats the sh*t out of x86 at the same clock speeds, when floating point operations came into the picture.

And yes, i'm playing the smart-ass, hehe... ;] But hope this makes the picture more clear for ppl. has no or few programming experiences with 32bit x86. (I've some 10 years or so, but i'm not proud for it, hehe.:)
Analysis: x86 Vs PPC : Comment 15 of 38ANN.lu
Posted by Chain|Q on 10-Jul-2003 05:22 GMT
In reply to Comment 14 (Chain|Q):
bugfix: "the most painful part of the x86 architecture is the FPU, since the x86 has no linear address space"

no linear register set i mean... :)
Analysis: x86 Vs PPC : Comment 16 of 38ANN.lu
Posted by gz on 10-Jul-2003 05:55 GMT
I think it's obvious that a ppc is faster than x86 when it boils down to floating point operations. That's the whole idea upon which a risc is based on. However you shouldn't really call a modern x86 a mere cisc these days as it's more a hybrid between risc and cisc architechture and ideas.

Also the use of floating point operations outside of say, a 3d workstation market. Is becoming less and less irrelevant to the processor used with the introduction of continually more powerful 3d accelerators that handle most of the 3d crunching in a desktop computer these days.

Some ppl seem to think that all a processor needs to do is floating point operations to perform fast. This couldn't be further from the truth, especially nowadays when allkinds of radeon's and geforce's are relieving the cpu from the fp workload in games and 3d applications.

The "megaherz myth" argument used often by apple is true but doesen't tell the whole truth simply because it's used to countermarket the x86. Too often when referred to a ppc we hear the "megaherz myth" as a pro argument but we fail to hear about the limitations of the risc technology.

The reason why risc hasn't become popular outside of scientific or 3d workstation market is not all due to the "ignorance" of the rest of the world, nor because of the higher prices of the processors (altough these are strong affecting factors aswell) There is also the reason of performance when using applications that do not involve ray-tracing and the like.
Analysis: x86 Vs PPC : Comment 17 of 38ANN.lu
Posted by Emeric SH on 10-Jul-2003 06:29 GMT
Very nice article, pointing out the benefits - better architecture for one and lower costs for the other. I liked it much.
Analysis: x86 Vs PPC : Comment 18 of 38ANN.lu
Posted by Fabio Alemagna on 10-Jul-2003 06:55 GMT
In reply to Comment 5 (corpse):
> but I'm sure

You are "sure"? In the sense that you "feel" it has to be like that, or in the sense that you *know* what you're talking about? :)

> a more were added with the 32bit 8386

Not general purpose registers, no.

> and some more with the 8586's

See above.

> (pentium) along with protected mode, FPU integration, MMX etc

There are still only *7* general purpose registers (eax, ebx, ecx, edx, esi, edi, ebp). There's another register, ESP, which is the stack pointer and thus can't properly be used for other things, and ebp is generally used as stack frame, but it's not necessary to use it like that and that's why I included it in the "free list" of general purpose registers.
Analysis: x86 Vs PPC : Comment 19 of 38ANN.lu
Posted by Bill Hoggett on 10-Jul-2003 07:01 GMT
As long as you accept that the writer is not being objective or impartial, the article is fine. It clearly presents half the truth, that which suits the PPC is better than x86 argument, which is what you would expect from someone working for a PPC system manufacturer.

Personally, I'd rather take my comparisons from arstechnica than Genesi.
Analysis: x86 Vs PPC : Comment 20 of 38ANN.lu
Posted by Matt Parsons on 10-Jul-2003 07:18 GMT
In reply to Comment 17 (Emeric SH):
The article was flawed, and in some cases wrong.

Example:
1.
It highlighted the weakness in the x87 (the FPU), it failed to note that these weakenesses have been rectified in the Pentium+ (Athlon+), with it's "free instructions" and multiple pipes etc... It also failed to note that the new PPC 970 has actually cut back the FPU as it is simply not that important on the desktop, Vector units and 3D GPU's are far more important.

2.
It claims that the PPC has a shorter fatter pipe than the Athlon. Not true the Athlon and PPC have similar sized pipes, it's the Pentium4 that has a long thin pipe. Also the article claims that the PPC 970's pipe is long and thin, but for the PPC some how this is better than the x86... It doesn't explain how or why,(which is not surprising since long and thin has benefits for any architecture when scaling performance, but this article couldn't suggest that the P4 has a more advanced design... could it?)

3.
The Term RISC and CISC has not been used by chip designers for a long time. The chips are now classified by their architecture, rather than the "design Philosiphy". Most CPU's are now "Load-Store". The x86 has a less complex instructin set than some RISCs, and the PPC has the most complex RISC instruction set I've ever seen... Both CPU's share RISC/CISC features, as it's more efficient to use both conceepts.

4.
The Article fails to note that the modern PPC and the medern x86 CPU's are infact very similar chips, they use all the same "tricks" and have all been designed by people who have kept up to date with the latest technology.

At the end of the day, I like the PPC, I even own one, and I'm hoping the Motherboard fairy brings me a Pegasos... fingers crossed. But I don't appreciate articles that are just blatent marketing.

You don't need to lie about the PPC to show how good it is!!!
Analysis: x86 Vs PPC : Comment 21 of 38ANN.lu
Posted by John Block on 10-Jul-2003 08:13 GMT
Related and a nostalgia trip - Jim Drew is mentioned

http://www.emaculation.com/ppc.php

I was curious as to whether a future amiga forever or amiathlon could emulate an amigaone.
Analysis: x86 Vs PPC : Comment 22 of 38ANN.lu
Posted by elektro on 10-Jul-2003 10:24 GMT
In reply to Comment 7 (Christophe Decanini):
yeah it does get mentioned later on but i stopped reading after it wasn't mentioned in the 'current' section (first time). It should be there since its out there and one can buy it.

"Obviously when you are asked why using your product over another one you enlight your products advantages. Nicholas did it and it did it well."

That's an advertisement then, not an analysis.
Analysis: x86 Vs PPC : Comment 23 of 38ANN.lu
Posted by corpse on 10-Jul-2003 11:15 GMT
In reply to Comment 18 (Fabio Alemagna):
"you *know* what you're talking about? :) "

I haven't done extensive x86 assembly... I implemented xmodem in 16bit x86 assembly for class and never looked back ;)
Analysis: x86 Vs PPC : Comment 24 of 38ANN.lu
Posted by Rob on 10-Jul-2003 15:46 GMT
Hi,
sorry if arrive now with this question but: The IBM G5 is the long awaited G5?
Because that one i mentioned was from Motorola?
What is the relationship?

Thanks
Analysis: x86 Vs PPC : Comment 25 of 38ANN.lu
Posted by Megol on 10-Jul-2003 19:23 GMT
In reply to Comment 14 (Chain|Q):
"Anyway, the most painful part of the x86 architecture is the FPU, since the x86 has no linear address space, but it's some weird kind of a stack. So on x86, you just can't fmove.l (a0),fp3, and then fmul fp2,fp3, because you only can move values to the top of the stack (all values moves down then) and you can only make operations on the value currently at the top of the stack. Of course there is an instruction, which swaps two values in the stack, but this makes overhead simply not exists on architectures using linear FPU register set."
The overhead is not that great in practice as (using Athlon as an example) the swap instruction is treated as a FNOP(==Floating point No OPeration) and the x86 can use memory operands to replace register accesses.
So your code above would translate to (if 'fp2's value is first on the stack):
fmul dword ptr [eax]

"Other difference, compared to the 68k for example, that you cannot move values directly between the CPU and the FPU, you have to store the value to the memory first, and then move to the FPU."
I thought most "RISC" architectures required this too?

"So this makes FPU coding on x86 difficult, and also makes the FPU unnecessarily complex, and ineffective."
Well the x86 FPUs still are among the most effective for microprocessors. But it really is a PITA to code floating point routines in asm :-/

"And yes, i'm playing the smart-ass, hehe... ;] But hope this makes the picture more clear for ppl. has no or few programming experiences with 32bit x86. (I've some 10 years or so, but i'm not proud for it, hehe.:)"
Be proud ;-)
Analysis: x86 Vs PPC : Comment 26 of 38ANN.lu
Posted by anonymous on 10-Jul-2003 20:00 GMT
Both Genesi and Eyetech can argue until they're blue in the face about why the PPC architecture offers advantages over the x86. But for many of us - including respected Amiga engineers like Dave Haynie, the reality of today's market is that cost and availability are far more compelling than technical elegance.

Let's be real, the decision to go with the PPC was made several years ago and it no longer makes sense in today's market. If the Amiga is to have a resurgence it needs hit the US$500 mark for an entry level system. An x86 CPU and motherboard - either off the shelf or customized is a low-risk investment, especially if you can retain compability with Windows and Linux.

Genesi, Eyetech and Hyperion are afraid that once Amiga users discover Windows and Linux (duh) the market for MorphOS and the AmigaOS will evaporate (duh redux). Someone apparently forgot to inform them that expensive, proprietary solutions don't offer much to a user base that hasn't seen new hardware in a decade.

Bottom line: this is a religious argument about hardware while the true value is in the software.
Analysis: x86 Vs PPC : Comment 27 of 38ANN.lu
Posted by CodeSmith on 10-Jul-2003 21:27 GMT
In reply to Comment 26 (anonymous):
"Genesi, Eyetech and Hyperion are afraid that once Amiga users discover Windows and Linux (duh) the market for MorphOS and the AmigaOS will evaporate (duh redux). Someone apparently forgot to inform them that expensive, proprietary solutions don't offer much to a user base that hasn't seen new hardware in a decade.

Bottom line: this is a religious argument about hardware while the true value is in the software. "

So, what you're saying is that Windows and Linux are both better software than AmigaOS or MorphOS. Otherwise, since the true value is in the software why would Hyperion and Genesi need to worry?

Gutsy statement pal. Too bad you're posting anonymously.
Analysis: x86 Vs PPC : Comment 28 of 38ANN.lu
Posted by CodeSmith on 10-Jul-2003 21:32 GMT
In reply to Comment 24 (Rob):
The PPC970 is not the G5 that's been announced for ever. The name was given to the 970 by Apple's marketing department, so ppl wanting to buy Macs would know that a Mac with a PPC970 is better than a Mac with any PPC7xx. G5 > G4 > G3.
Analysis: x86 Vs PPC : Comment 29 of 38ANN.lu
Posted by Bill Hoggett on 10-Jul-2003 21:59 GMT
In reply to Comment 27 (CodeSmith):
@CodeSmith

> So, what you're saying is that Windows and Linux are both better software
> than AmigaOS or MorphOS.

Actually, what he's saying is that Hyperion and Genesi think Windows and Linux both offer better software platforms that AmigaOS and MorphOS, so users who would have the option would not choose AmigaOS or MorphOS software. And of course, in this respect at least he is completely right: both Windows and Linux offer better software catalogues than AmigaOS or MorphOS, and this will remain the case for the forseeable future. Numbers of developers and resources dedicated to them make it inevitable.

> Otherwise, since the true value is in the software why would Hyperion
> and Genesi need to worry?

Hyperion rely on porting old games to niche platforms, and Genesi have PPC hardware they need to sell. Both those crucial core aspects of their businesses would be hit if AmigaOS and MorphOS were available on the same cheap off-the-shelf hardware that Windows (and Windows software, including games) could run on.

> Gutsy statement pal. Too bad you're posting anonymously.

What difference does it make? He's telling the truth, and what were you going to him if you did know his name, go and smash his face in? Sheesh.

The choice of PPC is in order to create an isolated artificial niche market. Nothing else. Articles like the one in this thread are a pile of misinformation designed to confirm to the converted that they have made the right choice and the extra money they will have to spend is all worth it. One day people will learn to promote their products by telling the truth and highlighting their own qualities rather than telling lies about rival products. We will have to wait a bit longer for that day it seems.
Analysis: x86 Vs PPC : Comment 30 of 38ANN.lu
Posted by DaveP on 11-Jul-2003 06:57 GMT
In reply to Comment 29 (Bill Hoggett):
He he he he.

:-D
Analysis: x86 Vs PPC : Comment 31 of 38ANN.lu
Posted by Rob on 11-Jul-2003 08:32 GMT
In reply to Comment 28 (CodeSmith):
What happened to the "real" G5?
Analysis: x86 Vs PPC : Comment 32 of 38ANN.lu
Posted by Matt Parsons on 11-Jul-2003 09:35 GMT
In reply to Comment 30 (DaveP):
>>>>>>>>>>>>>>>>>>>>>>>>>>>>
He he he he.

:-D
<<<<<<<<<<<<<<<<<<<<<<<<<<<<

What's so funny? Seriously, what's the joke?

Bill has raised a good point, that is totally relavent to the OSNews Article.
While I do believe that the 970 might be a competitive CPU, the Article was just plain of advertising make to like like a proper peice of Marketing.
Analysis: x86 Vs PPC : Comment 33 of 38ANN.lu
Posted by Matt Parsons on 11-Jul-2003 11:08 GMT
In reply to Comment 32 (Matt Parsons):
>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Article was just plain of advertising make to like like a proper peice of Marketing
<<<<<<<<<<<<<<<<<<<<<<<<<<<<

What is wrong with me? :-)

"It's Marketing make to look like an analysis"
Analysis: x86 Vs PPC : Comment 34 of 38ANN.lu
Posted by Chain|Q on 11-Jul-2003 15:31 GMT
In reply to Comment 25 (Megol):
"The overhead is not that great in practice as (using Athlon as an example) the swap instruction is treated as a FNOP(==Floating point No OPeration)"

Yes, on Athlon it takes zero clocks to execute this in the ideal case. Just like on Pentium/PPro/PII/PIII. But it's not the case on older x86's, and other x86-clone CPU's like the K6 & co, and also not true for Intel Pentium 4!!! And i just think this is still a good example of the hack&patches made on the terrible ISA to be somewhat more effective...

"and the x86 can use memory operands to replace register accesses.
So your code above would translate to (if 'fp2's value is first on the stack):"

_IF_ it's there... In most cases it doesn't so you've to swap, which makes code less readable, and results in slowdowns. Also using memory operands generate more load on the bus -> reduces efficiency, since in most cases you'll do more operations on a single value... Also if the required value is not in the cache, you might got some nasty pipe stalls and like that. Simply too much limitations, it's a PITA as you said... :]

"Well the x86 FPUs still are among the most effective for microprocessors."

In fact, i'd rather call them fast, than effective. Since they're fast that's unquestionable, but not effective, because they waste more resources to archieve the same speed levels of other CPU's, just because the ugly architecture and brain-damaged design of the ISA... This is why they produce more heat, and stuff like that, already discussed in the quoted articles.
Analysis: x86 Vs PPC : Comment 35 of 38ANN.lu
Posted by Joe "Floid" Kanowitz on 11-Jul-2003 18:05 GMT
In reply to Comment 31 (Rob):
Rob asked,
> What happened to the "real" G5?

You'd have to talk to Motorola. Presumably they did have a chip in the works, possibly to the point of sealed-Mac evaluation hardware (rumors), but squabbles over the bus (RapidIO vs.. whatever the competing tech is?) helped quash it. On the other hand, perhaps it never got off the drawing board, and any pre-970 sealed-boxes were for the higher-clock G4s Moto did put into production. (Arguably, this could've been Apple being their usual selves- an Apple-derived interconnect means *Apple* makes license bling on all implementations sold; RapidIO would mean Motorola could've made some further change off IBM's designs.
http://www.theinquirer.net/?article=6037 - funny how that name keeps cropping up - reviews the situation, as it stood a while back.)

In the end, you can fret about these things, but what really matters, as always, are: -Existence, -Platform support, -Price/Performance, and -Performance/Watt. We've wound up with a pretty darn fast PowerPC via the 970, and if IBM can ramp it Athlon-style (a relatively simple, modifiable design), then "we"'re pretty well set. It's not like losing Transmeta -- which, arguably, we might've seen some Neat Hacks -- it's more like losing the Cyrix M-II when you've already got Durons on the market. Sure, those guys could do some crazy engineering (and as corrolary, the M-II turns out to have been an interesting little chip, in retrospect), but how much can we care when it wouldn't scale to competetive levels? (Those 'overclocked' Motorola chips certainly don't run too cool in existing Macs.)
Analysis: x86 Vs PPC : Comment 36 of 38ANN.lu
Posted by smp266 on 12-Jul-2003 01:00 GMT
In reply to Comment 29 (Bill Hoggett):
In this day and age the motherboard and CPU are like pieces of Lego. Just two elements of the whole package. A good graphics card is more expensive than both of those. Depending on your need for DVD burners etc. The CPU may only be 5% of the entire system value. (Obviously you need matching drivers and applications).

There is also a window of opportunity because of the transition to 64-bit. If an alternate platform is going to take off... Well let's see who is still standing in 2 years.

AMD got plenty of fanatics in the gaming community by beefing up their FPU. Easy if you have popular games on the platform.

Corporations are not very inventive and seem to prefer clones of big selling products over something 'unknown'.

I will be happy if Amiga/Pegasos is nothing more than an internet/video processing box. At least I will some sanctuary from the Windows 3 Ring Bug Circus.
Analysis: x86 Vs PPC : Comment 37 of 38ANN.lu
Posted by Anonymous on 14-Jul-2003 18:17 GMT
In reply to Comment 21 (John Block):
Early this year /. had an article on some company that supposedly had figured out how to use the Amd Athlon processor to emulate a G4. 1Ghz Athlon got you a 750Mhz G4 or something. However, I've yet to see anything actually be released.
Analysis: x86 Vs PPC : Comment 38 of 38ANN.lu
Posted by Anonymous on 14-Jul-2003 18:18 GMT
In reply to Comment 24 (Rob):
G5 is the Apple name. IBM calls the processor the 970. Motorola supposedly has a processor after the Apple G4. However it won't be called G5 as it's not being used by Apple it'll be 7557 (or whatever the Motorola # is.)
Anonymous, there are 38 items in your selection
Back to Top