first of all this thing probably doesn't take advantage of the dual MHz, then certainly no altivec, third it probably even uses some little endian stuff etc etc.
Claiming that a P3 667 is as fast as a Dual G4 1000 is simply ridiculous. Sure, claiming it's 70% faster than a P4 2000 is equally ridiculous.
Wake up my friend, c't aren't gods either.
I remember they compared WinZip vs Zipit in the review of the G4 500 and were surprised that Zipit (a shareware tool that hasn't gotten updated in about 10 years) was about 5x slower than winzip on the PC.
I can only laugh at those benchmarks, and not because they imply that the G4 sucks...
After reading the article (which is German) you'll note that (if you KNOW German that is).
The compiler (C++) used for Spec int is slower than CodeWarrior.
SPECint_base on the G4 1000 (only one chip used) and the P3 1000 are 306 and 309 respectively. That is equal.
as for SPECfp they admit that fortran is not common on Macs, that the compiler probably sucks etc.
They also note that the G4 also performed worse in C++ FP specs, but that is no surprise, as the 604 was the last Mot chip that had competitive FP scores. Would the compiler use altivec for optimisations, it would probably look very different.
Now if you take all this, add teh second chip for the 1 GHz machine, take a decent compiler, and optimize to a certain degree, it's quite clear that the Dual 1 GHz G4 can theoretically be more than twice as fast as a 1 GHz P3. Add to that that we know the P3 has better IPC than the P4, and we're where everyone else sees the Dual 1GHz machine at the moment: About 90% of the fastest single CPU PC offerings.
In case you didn't see that:
The comparison they made isn't that bad after all, but then again there are so many things you have to take care of in order to make it an even remotely platform independent benchmark, it's almost impossible, especially with Intel having stirred a lot of optimisations into SPEC over th past 5 years, SPECfp being based on a lot of scientific Fortran code etc.
The way things look now, there are many ways to cross-platform-benchmark easier and more accurately than with SPEC2000CPU atm.
Blame Apple and Moto for that (not supporting SPEC, or releasing similarly "SPEC-whored" compilers as Intel does), but in the end, real world benchmarks still beat them all, and I'm not talking about Photoshop.
read those articles carefully, and in your native language before you jump to conclusions.
I didn't read the article in German, I didn't have to. I had already mentioned in my post that compilers were a problem for the G4. I personally don't think a G4 1 Ghz is slower than a Pentium 3 at 700 Mhz.
But you can say all you want about Intel having too much control over spec or optimising their compilers for it and you still won't be able to change the fact that a number of other non-Intel processors score much higher.
People comparing the performance of a dual 1Ghz G4 to a single P4 or Athlon seem to always conveniently ignore the fact that dual Athlons exist too and so too dual P4s, they just call them Xeon.
The fact is, we all knew that the G4 especially lagged behind in FP performance, but still many people blamed Alias|WaveFront or Newtek for not properly optimising their code when PowerMacs turned in rendering times twice as long as the competition. I think it's now clear that this is not the case.
Oh, and taken from <a href="http://www.spec.org,"; target="_blank">www.spec.org,</a> you can search for it in the database if you wish:
Who said anything about keeping AMD? And who cares if AMD is willing or not? As long as Apple has the cash, it can engauge in a hostile takeover.
I proposed that Apple buy AMD to kill it and eliminate the x86 competition Intel faces, not switch to x86. Sure, Apple would lose a few billion, but so what?
I'm sure Hammer's technology can incorporated into the G5 and other future PPC's.</strong><hr></blockquote>
Wow, that is by far the dumbest idea I've ever heard. I'm impressed you were able to come up with something so completely illogical.
How in the world would eliminating competition for Intel in any way benefit Apple?
You realize that almost all the technology developed by AMD relates to x86 and ways to improve x86 technology? This would be little to no use to anyone who wanted to design a PPC chip. And what would Apple do with this technology? They aren't even a CPU manufacturer. You would have Apple squander it's entire cash reserve to make a strategic blunder. Nice, very nice. <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
If anything you should be kissing AMD's ass for giving you 1GHz G4s. If it wasn't for their help Moto would still be stuck at 500MHz. Luckily for AMD they saw the error in getting involved with Motorola and recently broke off their alliance.
Of course I know its not twice as fast, thus the "theoretically".
As you can read in the maccentral thread I linked to above, spec is actually designed to compare chips on the same platform, but is being abused for penis comparisons by companies like IBM, Intel, AMD etc etc.
The article also states that using an optimized compiler, you can get up to 20-30% more score out of the same chip, which c't luckily didn't.
Thus comparing the P3 700 on the spec submissions to the benchmarking done at c't is impossible, it an apple vs oranges comparison even more than the c't article already is.
I never said the G4 was the fastest chip in the world, that's probably the Power4 right now, but even that is mixed with a lot of "my penis is bigger than yours" with Sun's UltraSPARC.
I frankly don't give a shit what SPEC scores say or better claim about the G4.
It's obvious that the G4 is a fast chip for many applications, while seriously lagging behind in other areas, but that is true for almost every chip out there.
I'd still like to see a lay person comparing a 2.2GHz P4 vs a 1GHz DP G4 without benchmarks, just using apps. I'm actually quite sure you couldn't determine a speed difference.
The same thing will be true for the G5, if it ever comes out. Stupid people will compare using MHz.
Stupid people who think they're smart will compare using FLOPs. Smartasses will compare using SPEC. People with a clue will try to compare using real apps and fail as well. And the really smart people won't start to compare in the first place, they'll just use it.
<strong> Smartasses will compare using SPEC. People with a clue will try to compare using real apps and fail as well. And the really smart people won't start to compare in the first place, they'll just use it.
G-News</strong><hr></blockquote>
You realize that SPEC is based off of source code for real apps don't you? Applications such as word processors, compilers, games, circuit simulation, database, compression, chemistry, 3D, advanced physics, image processing, and even a weather tool. It is written by some of the best programmers and computer engineers in the world.
SPEC is not some silly little benchmark, it is the industry benchmark. And it is perfectly valid for comparing one platform to another. Sun SPARC vs. IBM Power4 vs. Intel Xeon. Or whatever combo floats your boat. And I wouldn't say that the P4 is that crappy of a chip when the 2.2GHz Northwood chip with 512K of cache is just about as fast in SPECInt as IBM's 1.3GHz Power4 using all 1.5MB of L2 cache (normally shared between 2 processors) and 128MB of L3 cache.
Yeah I know that, but the point is that it's PARTS of real world software, namely the msot computing intense parts. All SPEC proves is that a chip can handle processing intensive tasks generating huge amounts of data, while probably performing quite different in one of the other tasks an application may need to be executed.
SPEC could be the benchmark of all benchmarks, but it's like Linux: everyone has to fiddle it together himself, optimize here and there a bit, cheat a few instructions here and there and in the end what you have is the reincarnation of the anti standard.
Power4 vs UltraSparc is probably the worst example of them all, because every other week Sun or IBM releases a new score that beats the previous record by 10 points, adding some more code optimisation or another chip or two. It's the best example to show that you can basically have SPEC generate whatever score you want, simply depending on how keen you are to demonstrate YOU have the biggest penis.
If it was possible to have a benchmark that came compiled for a specific platform (say P4 or Athlon or G4) and you'd run that, basically independent from the rest of the machine, that would be a viable cross-platform benchmark.
No cheating, everyone would have to stick to the same standard and it would really allow for an Apples to Apples comparison.
What we have now is just a sad farce of a good idea. Too many cooks still spoil the soup.
Also I didn't say the P4 was a crappy chip, it has some amazing ideas in there and a lot of future potential. Given it has to deal with so much legacy shit, it does an amazing job. It's a quite cleverly designed chip, compared to AMD's brute force approach.
<strong>SPEC could be the benchmark of all benchmarks, but it's like Linux: everyone has to fiddle it together himself, optimize here and there a bit, cheat a few instructions here and there and in the end what you have is the reincarnation of the anti standard.
</strong><hr></blockquote>
The way you say it, you make it sound like one can tinker with the spec source code, or write special compilers that are only designed to produce good spec binaries. But the spec benchmarks have a lot of rules that must be adhered to. First, you cannot touch the source code, you cannot optimise it in any way shape or form. You CAN use compiler optimisations, but they're only allowed as far as they improve overall application performance in general, and are not just profiled for spec sources. The main rules are:
[quote] Hardware and software used to run the CINT2000/CFP2000 benchmarks must provide a suitable environment for running typical C, C++, or Fortran programs.
Optimizations must generate correct code for a class of programs, where the class of programs must be larger than a single SPEC benchmark or SPEC benchmark suite. This also applies to assertion flags that may be used for peak compilation measurements (see section 2.2.4).
Optimizations must improve performance for a class of programs where the class of programs must be larger than a single SPEC benchmark or SPEC benchmark suite.
The vendor encourages the implementation for general use.
The implementation is generally available, documented and supported by the providing vendor.
<hr></blockquote>
There's also tons of other regulations on the details of what compiler flags can and cannot be used.
As I've already said, even if, as some people claim, Intel were to somehow find a way to cheat a little, this wouldn't change the fact that many other CPU's from various MPU manufacturers both with lower and higher Mhz numbers than the 1 Ghz G4, score much higher than it. Overall the G4 looks like a desktop CPU from yesteryear, at best. And that is sad...
[quote]A P3 at 4GHz would most likely out perform a P4 at the same 4GHz by 50%.<hr></blockquote>
I doubt it. Without more memory and FSB bandwidth, a PIII would likely be spinning its wheels long before even 3GHz, much less 4GHz. As multipliers increase, bandwidth and memory speed become ever more paramount. This is why in many tests a Pentium 4 2.4GHz using RDRAM on the i850 can outperform a Pentium 4 3GHz using DDR-SDRAM on any of the desktop market chipsets. A Pentium III at anywhere near 4GHz using plain jane PC133 would be utterly useless.
No, strike that. It would be useful for one thing: laughing at how pathetic it would be.
[quote]I'm not all that concerned about a procssor that's not shipping for some time, and it's P4 at that. The P3 is still a faster processor if you live in reality (the Mac world) intel, and their users have always based their processors on x-Hz. But we base ours on actual performnance. The P3 will still out pace the P4 at an = clock cycle (MHz), and if I was looking for a processor to take to 4GHz I would taken he P3 up to 4GHz. A P3 at 4GHz would most likely out perform a P4 at the same 4GHz by 50%. <hr></blockquote>
If I wanted to give others the impression that I didn't what the hell I was talking about, I don't think I could come up with better words myself.
Well, well, well if it isn't my nemesis, johnsonfromwisconsin...
heh.
Don't bother with some of the people on this forum, who like to knock the P4 because its demonstrably faster than a dual Gig.
People who think that the northwood is a shitty processor, just because its a P4, need to bite the head off the dog next to them, because they're lving in Candyland and the dog is made of chocolate. (recycled joke, I know).
Besides, comparisons between a PIII and a P4 at 4 GHz are fallacious. The P4 has been demonstrated at 4 GHz, whereas the PIII will never get past 2, in its current, and any future, form.
Listen, I'm all for the G4, and I really think that with the HIP-7 process transistion, Motorola can really deliver a kickass chip, especially if Apple gets its **** together and couples it with a northbridge that even comes close to circa 2000 memory specs. But really, its foolish to even talk about AMD or Intel, because they don't make processors for the Mac platform...its a waste of breath and a waste of time.
[quote]Listen, I'm all for the G4, and I really think that with the HIP-7 process transistion, Motorola can really deliver a kickass chip, especially if Apple gets its **** together and couples it with a northbridge that even comes close to circa 2000 memory specs. But really, its foolish to even talk about AMD or Intel, because they don't make processors for the Mac platform...its a waste of breath and a waste of time<hr></blockquote>
I think the G4 is a decent processor, it just can't match the Athlon or P4's at scalar Int or FP. I'm well versed in uArch, but I'm not all that familiar with Mot's process tech. So when I prognosticate about a processors future performance like this, I usually find that recent history is a pretty good indicator. In this case, I don't think it's going to bring scalar INT or FP equity to the powerPC, especially with AMD and Intel as moving targets. The only way the PPC will catch up is if AMD flounders with 'Hammer.
The future of the PPC chips is in the altivec unit. The altivec unit prevent the G4 to be an has been chip. All the photoshop outrageous benchmarks demo from Jobs vs intel works upon the altivec stuff.
The future of the ppc chips for Apple computers will be a 256 bit Altivec unit.
The chips will need too , a much better gestion of the memory in order to feed the new altivec stuff, with a larger memory bandwitch (DDR ram 266 will not be sufficiant).
It would be a good idea if Mot can copy the P4 for the integer unit , who runs at double speed (but it needs a far deeper pipeline).
Most of the FP stuff (if i remember well what i read here on others threads) can be run on the altivec unit with recompiling works.
actually the P3 never surpassed 1133MHz, they shipped it at that speed just to notice it didn't run stable, had to pull it, big fiasco etcetc until the P4 came out.
The P3 was basically an evolved P2, and at 1133 it had simply reached its last markstone, that architecture was DEAD right there. We'd think the G4 would also reach such a dead zone one day, but so far moto think it has a lot of life left in it, I wonder whether that will prove true.
Comparing the P3 against the P4 is like comparing the P1 vs the P2: a nono.
<strong>actually the P3 never surpassed 1133MHz, they shipped it at that speed just to notice it didn't run stable, had to pull it, big fiasco etcetc until the P4 came out.
</strong><hr></blockquote>
They indeed had problems with the 1.13GHz version initially, and had to pull it, but it was later reintroduced. Also, the P3 and P3-M do ship in versions up to 1.2GHz right now (using a 130nm process).
Comments
<strong>
<a href="http://www.hitachi.com"; target="_blank">hitachi</a> doesn't seem to think theres anything wrong with a <a href="http://www.pcworld.com/news/article/0,aid,86363,00.asp"; target="_blank">water cooled notebook.</a></strong><hr></blockquote>
That's nothing new. That concept has been around forever, just never really caught on.
I'd say Mot never released the scores because there are no usable fortran compilers out there.
(see <a href="http://maccentral.macworld.com/storyforum/forums/2001/04/17/myth/?read=131)" target="_blank">http://maccentral.macworld.com/storyforum/forums/2001/04/17/myth/?read=131)</a>
first of all this thing probably doesn't take advantage of the dual MHz, then certainly no altivec, third it probably even uses some little endian stuff etc etc.
Claiming that a P3 667 is as fast as a Dual G4 1000 is simply ridiculous. Sure, claiming it's 70% faster than a P4 2000 is equally ridiculous.
Wake up my friend, c't aren't gods either.
I remember they compared WinZip vs Zipit in the review of the G4 500 and were surprised that Zipit (a shareware tool that hasn't gotten updated in about 10 years) was about 5x slower than winzip on the PC.
I can only laugh at those benchmarks, and not because they imply that the G4 sucks...
G-News
btw: you might want to take a look at this:
<a href="http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=03M9430304504 67M98653" target="_blank">http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=03M9430304504 67M98653</a>
One more:
Explain me why these guys here get so very different scores?
<a href="http://www.cpuscorecard.com/"; target="_blank">http://www.cpuscorecard.com/</a>;
[ 02-28-2002: Message edited by: G-News ]
[ 02-28-2002: Message edited by: G-News ]</p>
The compiler (C++) used for Spec int is slower than CodeWarrior.
SPECint_base on the G4 1000 (only one chip used) and the P3 1000 are 306 and 309 respectively. That is equal.
as for SPECfp they admit that fortran is not common on Macs, that the compiler probably sucks etc.
They also note that the G4 also performed worse in C++ FP specs, but that is no surprise, as the 604 was the last Mot chip that had competitive FP scores. Would the compiler use altivec for optimisations, it would probably look very different.
Now if you take all this, add teh second chip for the 1 GHz machine, take a decent compiler, and optimize to a certain degree, it's quite clear that the Dual 1 GHz G4 can theoretically be more than twice as fast as a 1 GHz P3. Add to that that we know the P3 has better IPC than the P4, and we're where everyone else sees the Dual 1GHz machine at the moment: About 90% of the fastest single CPU PC offerings.
In case you didn't see that:
The comparison they made isn't that bad after all, but then again there are so many things you have to take care of in order to make it an even remotely platform independent benchmark, it's almost impossible, especially with Intel having stirred a lot of optimisations into SPEC over th past 5 years, SPECfp being based on a lot of scientific Fortran code etc.
The way things look now, there are many ways to cross-platform-benchmark easier and more accurately than with SPEC2000CPU atm.
Blame Apple and Moto for that (not supporting SPEC, or releasing similarly "SPEC-whored" compilers as Intel does), but in the end, real world benchmarks still beat them all, and I'm not talking about Photoshop.
read those articles carefully, and in your native language before you jump to conclusions.
G-News
But you can say all you want about Intel having too much control over spec or optimising their compilers for it and you still won't be able to change the fact that a number of other non-Intel processors score much higher.
People comparing the performance of a dual 1Ghz G4 to a single P4 or Athlon seem to always conveniently ignore the fact that dual Athlons exist too and so too dual P4s, they just call them Xeon.
The fact is, we all knew that the G4 especially lagged behind in FP performance, but still many people blamed Alias|WaveFront or Newtek for not properly optimising their code when PowerMacs turned in rendering times twice as long as the competition. I think it's now clear that this is not the case.
Oh, and taken from <a href="http://www.spec.org,"; target="_blank">www.spec.org,</a> you can search for it in the database if you wish:
[quote]Intel SE440BX-2 (700 MHz Pentium III):
specintbase 310 specfpbase 213 <hr></blockquote>
<strong>
Who said anything about keeping AMD? And who cares if AMD is willing or not? As long as Apple has the cash, it can engauge in a hostile takeover.
I proposed that Apple buy AMD to kill it and eliminate the x86 competition Intel faces, not switch to x86. Sure, Apple would lose a few billion, but so what?
I'm sure Hammer's technology can incorporated into the G5 and other future PPC's.</strong><hr></blockquote>
Wow, that is by far the dumbest idea I've ever heard. I'm impressed you were able to come up with something so completely illogical.
How in the world would eliminating competition for Intel in any way benefit Apple?
You realize that almost all the technology developed by AMD relates to x86 and ways to improve x86 technology? This would be little to no use to anyone who wanted to design a PPC chip. And what would Apple do with this technology? They aren't even a CPU manufacturer. You would have Apple squander it's entire cash reserve to make a strategic blunder. Nice, very nice. <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
If anything you should be kissing AMD's ass for giving you 1GHz G4s. If it wasn't for their help Moto would still be stuck at 500MHz. Luckily for AMD they saw the error in getting involved with Motorola and recently broke off their alliance.
As you can read in the maccentral thread I linked to above, spec is actually designed to compare chips on the same platform, but is being abused for penis comparisons by companies like IBM, Intel, AMD etc etc.
The article also states that using an optimized compiler, you can get up to 20-30% more score out of the same chip, which c't luckily didn't.
Thus comparing the P3 700 on the spec submissions to the benchmarking done at c't is impossible, it an apple vs oranges comparison even more than the c't article already is.
I never said the G4 was the fastest chip in the world, that's probably the Power4 right now, but even that is mixed with a lot of "my penis is bigger than yours" with Sun's UltraSPARC.
I frankly don't give a shit what SPEC scores say or better claim about the G4.
It's obvious that the G4 is a fast chip for many applications, while seriously lagging behind in other areas, but that is true for almost every chip out there.
I'd still like to see a lay person comparing a 2.2GHz P4 vs a 1GHz DP G4 without benchmarks, just using apps. I'm actually quite sure you couldn't determine a speed difference.
The same thing will be true for the G5, if it ever comes out. Stupid people will compare using MHz.
Stupid people who think they're smart will compare using FLOPs. Smartasses will compare using SPEC. People with a clue will try to compare using real apps and fail as well. And the really smart people won't start to compare in the first place, they'll just use it.
G-News
<strong> Smartasses will compare using SPEC. People with a clue will try to compare using real apps and fail as well. And the really smart people won't start to compare in the first place, they'll just use it.
G-News</strong><hr></blockquote>
You realize that SPEC is based off of source code for real apps don't you? Applications such as word processors, compilers, games, circuit simulation, database, compression, chemistry, 3D, advanced physics, image processing, and even a weather tool. It is written by some of the best programmers and computer engineers in the world.
SPEC is not some silly little benchmark, it is the industry benchmark. And it is perfectly valid for comparing one platform to another. Sun SPARC vs. IBM Power4 vs. Intel Xeon. Or whatever combo floats your boat. And I wouldn't say that the P4 is that crappy of a chip when the 2.2GHz Northwood chip with 512K of cache is just about as fast in SPECInt as IBM's 1.3GHz Power4 using all 1.5MB of L2 cache (normally shared between 2 processors) and 128MB of L3 cache.
SPEC could be the benchmark of all benchmarks, but it's like Linux: everyone has to fiddle it together himself, optimize here and there a bit, cheat a few instructions here and there and in the end what you have is the reincarnation of the anti standard.
Power4 vs UltraSparc is probably the worst example of them all, because every other week Sun or IBM releases a new score that beats the previous record by 10 points, adding some more code optimisation or another chip or two. It's the best example to show that you can basically have SPEC generate whatever score you want, simply depending on how keen you are to demonstrate YOU have the biggest penis.
If it was possible to have a benchmark that came compiled for a specific platform (say P4 or Athlon or G4) and you'd run that, basically independent from the rest of the machine, that would be a viable cross-platform benchmark.
No cheating, everyone would have to stick to the same standard and it would really allow for an Apples to Apples comparison.
What we have now is just a sad farce of a good idea. Too many cooks still spoil the soup.
Also I didn't say the P4 was a crappy chip, it has some amazing ideas in there and a lot of future potential. Given it has to deal with so much legacy shit, it does an amazing job. It's a quite cleverly designed chip, compared to AMD's brute force approach.
G-News
[ 02-28-2002: Message edited by: G-News ]</p>
<strong>SPEC could be the benchmark of all benchmarks, but it's like Linux: everyone has to fiddle it together himself, optimize here and there a bit, cheat a few instructions here and there and in the end what you have is the reincarnation of the anti standard.
</strong><hr></blockquote>
The way you say it, you make it sound like one can tinker with the spec source code, or write special compilers that are only designed to produce good spec binaries. But the spec benchmarks have a lot of rules that must be adhered to. First, you cannot touch the source code, you cannot optimise it in any way shape or form. You CAN use compiler optimisations, but they're only allowed as far as they improve overall application performance in general, and are not just profiled for spec sources. The main rules are:
[quote] Hardware and software used to run the CINT2000/CFP2000 benchmarks must provide a suitable environment for running typical C, C++, or Fortran programs.
Optimizations must generate correct code for a class of programs, where the class of programs must be larger than a single SPEC benchmark or SPEC benchmark suite. This also applies to assertion flags that may be used for peak compilation measurements (see section 2.2.4).
Optimizations must improve performance for a class of programs where the class of programs must be larger than a single SPEC benchmark or SPEC benchmark suite.
The vendor encourages the implementation for general use.
The implementation is generally available, documented and supported by the providing vendor.
<hr></blockquote>
There's also tons of other regulations on the details of what compiler flags can and cannot be used.
As I've already said, even if, as some people claim, Intel were to somehow find a way to cheat a little, this wouldn't change the fact that many other CPU's from various MPU manufacturers both with lower and higher Mhz numbers than the 1 Ghz G4, score much higher than it. Overall the G4 looks like a desktop CPU from yesteryear, at best. And that is sad...
[ 02-28-2002: Message edited by: timortis ]</p>
I doubt it. Without more memory and FSB bandwidth, a PIII would likely be spinning its wheels long before even 3GHz, much less 4GHz. As multipliers increase, bandwidth and memory speed become ever more paramount. This is why in many tests a Pentium 4 2.4GHz using RDRAM on the i850 can outperform a Pentium 4 3GHz using DDR-SDRAM on any of the desktop market chipsets. A Pentium III at anywhere near 4GHz using plain jane PC133 would be utterly useless.
No, strike that. It would be useful for one thing: laughing at how pathetic it would be.
[ 02-28-2002: Message edited by: TheAlmightyBabaramm ]</p>
If I wanted to give others the impression that I didn't what the hell I was talking about, I don't think I could come up with better words myself.
Bravo! teh W1nn@r is YUO!
[ 02-28-2002: Message edited by: johnsonfromwisconsin ]</p>
heh.
Don't bother with some of the people on this forum, who like to knock the P4 because its demonstrably faster than a dual Gig.
People who think that the northwood is a shitty processor, just because its a P4, need to bite the head off the dog next to them, because they're lving in Candyland and the dog is made of chocolate. (recycled joke, I know).
Besides, comparisons between a PIII and a P4 at 4 GHz are fallacious. The P4 has been demonstrated at 4 GHz, whereas the PIII will never get past 2, in its current, and any future, form.
Listen, I'm all for the G4, and I really think that with the HIP-7 process transistion, Motorola can really deliver a kickass chip, especially if Apple gets its **** together and couples it with a northbridge that even comes close to circa 2000 memory specs. But really, its foolish to even talk about AMD or Intel, because they don't make processors for the Mac platform...its a waste of breath and a waste of time.
I think the G4 is a decent processor, it just can't match the Athlon or P4's at scalar Int or FP. I'm well versed in uArch, but I'm not all that familiar with Mot's process tech. So when I prognosticate about a processors future performance like this, I usually find that recent history is a pretty good indicator. In this case, I don't think it's going to bring scalar INT or FP equity to the powerPC, especially with AMD and Intel as moving targets. The only way the PPC will catch up is if AMD flounders with 'Hammer.
[ 03-01-2002: Message edited by: johnsonfromwisconsin ]</p>
The future of the ppc chips for Apple computers will be a 256 bit Altivec unit.
The chips will need too , a much better gestion of the memory in order to feed the new altivec stuff, with a larger memory bandwitch (DDR ram 266 will not be sufficiant).
It would be a good idea if Mot can copy the P4 for the integer unit , who runs at double speed (but it needs a far deeper pipeline).
Most of the FP stuff (if i remember well what i read here on others threads) can be run on the altivec unit with recompiling works.
The P3 was basically an evolved P2, and at 1133 it had simply reached its last markstone, that architecture was DEAD right there. We'd think the G4 would also reach such a dead zone one day, but so far moto think it has a lot of life left in it, I wonder whether that will prove true.
Comparing the P3 against the P4 is like comparing the P1 vs the P2: a nono.
G-News
<strong>actually the P3 never surpassed 1133MHz, they shipped it at that speed just to notice it didn't run stable, had to pull it, big fiasco etcetc until the P4 came out.
</strong><hr></blockquote>
They indeed had problems with the 1.13GHz version initially, and had to pull it, but it was later reintroduced. Also, the P3 and P3-M do ship in versions up to 1.2GHz right now (using a 130nm process).
Bye,
RazzFazz
The P3-M doesn't quite compare though, it's quite a bit modified, and the process change of course allows for more than 1133.
G-News