[quote]Apple may be creative in how they use the G5 at first. My guess is that it will be one or two very high end workstation class Macs. The G4 will still be used in most PowerMacs for a while. Once G5 fabrication has gone to the 0.09 process and is tuned for high volumes, we may see G5s in all the PowerMacs.
<hr></blockquote>
LOL, everyone is thinking as if this new G5 is going to be fabbed by Moto. It's time to unlearn what you know about PPC CPUs and get ready to have IBM rock your world.
Remember that IBM is one of the world's most experienced chip manufacturers. They will be fabbing this "G5" in a brand-new, state of the art facility using new equipment and experienced techs. And yes, IBM keeps their fabs clean (so I hear).
This is not your father's PPC fab. I think we should all prepare to be utterly floored by IBM's efficiency, punctuality, and flat-out rocking potential.
[quote] Show us how bad the G4 is. Links please. <hr></blockquote>
<a href="http://www.blanos.com/benchmark/"; target="_blank">Here's a page full of benchmarks.</a> Just pick any one of the Lightwave scenes, and see how badly the G4 gets it's ass kicked. For another, does anybody remember the magazine that benchmarked PCs & Macs in After Effects? The G4 gets it's hat handed to it there, as well.
You can discount this and justify the Mac's poor performance all you want, guys, but these apps are in the market Apple is HEAVILY targeting right now.
[quote] I think the slowness is a perception problem, and it all stems from the GUI. <hr></blockquote>
No. When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. Although, the GUI does need a boost in 10.2. While quickly scrolling down the full length of one of the longer pages here on AI, Chimera can take a good chunk of a second for the thumb of the scroll bar to catch up with the mouse pointer. You can see that lag in other apps, as well.
JYD, I think folks are starting to finally lean away from the notion of Moto fabbing the "G5", and quite possibly accepting the theory of a role switch, as was suggested earlier, between IBM and Moto. All because of IBM's Oct 15th announcement and Moto publicly stating that they are concentrating on the embedded market. I know that is what I would like to see.
You're right about IBM (I have friends that work there), their fabs are clean, and when they release a schedule ... they keep to it. IMO, it's better to have IBM as the PowerMac supplier and Moto as the consumer / laptop supplier. It fits within the capabilities and chosen directions of the companies.
And yes JYD, I'm hold'n on. I want IBM to Rock my Mac World. Only wish I had enough patience.
I'm willing to bet the October announcement is for more than simply releasing the specs on a CPU that going to be fab'd in 12 months. I stated in an earlier post that under a pas antitrust settlement, IBM was absolutely restricted from making any sort of "vaporware" or planned product announcement. This didn't include "roadmaps" though. From what I vaguely remember it stated that IBM cold not announce a product unless is was either VERY near completion or already complete and ready to fab. I'm hoping someone else on this forum can verify that this restriction is still in place. This would provide a better guess as to when this chip is to debut.
Unless they repealed it the restriction is still in place.
\t I know a lot of you don't like <a href="http://www.spymac.com"; target="_blank">www.spymac.com</a> but there is an interesting article on their page. It gives us some more evidence of a new chip coming soon. And surprising enough it's based of evidence we all have access to. Go check it out
<strong>When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. </strong><hr></blockquote>
Great link! So I guess now somebody will stop complaining about Lemon Bon Bon's complaints, since it looks like he's right...
Personally I think Altivec performance is really awesome, but every kind of un-optimized code is simply s-l-o-w, maybe with different degrees of slowness but s-l-o-w all the same.
If IBM can really deliver the next PPC, and if its performance will be anywhere comparable to that of a Power4... well, I might sell my car to buy one of those beasts!
If IBM can really deliver the next PPC, and if its performance will be anywhere comparable to that of a Power4... well, I might sell my car to buy one of those beasts!
ZoSo</strong><hr></blockquote>
As Lionel is saying <a href="http://www.macbidouille.com/niouzcontenu.php?date=2002-09-12#3406"; target="_blank">here</a>, it seems that Apple looks to AMD for a replacement of the G4. Not that MacOS X will be running to any PC...simply the Apple hardware will be powered by some other kind of processor. This reinforces the recent rumor that Apple did not show interest to IBM's proposition for the new PPC derivative of Power4. Time will tell.
<strong>Unless they repealed it the restriction is still in place.
\t I know a lot of you don't like <a href="http://www.spymac.com"; target="_blank">www.spymac.com</a> but there is an interesting article on their page. It gives us some more evidence of a new chip coming soon. And surprising enough it's based of evidence we all have access to. Go check it out</strong><hr></blockquote>
Doesn't it seem likely that Apple had prototypes of both single and dual 1.25 GHz G4s? (and all other clockspeeds as well). Spymac is assuming that a single CPU in the prototype mobo is something other than a single G4 because.....? This isn't really all that exciting.
Thankyou Gamblor and ZoSo. I bow to your superior wisdom.
" I think the slowness is a perception problem."
That appears to be correct for some Apple users on these boards.
Check out Tom's hardware for additional benches on the latest (unlike the Photoshop stats above...) x86 Lightwave benches. These systems are available for half the price of Apple's 'flagship'.
The XP gets it's head handed to it by the P4.
Ouch.
Lemon Bon Bon
PS. Spymac. Interesting 'aftermath' analysis. Add up HUGE heatsinks, a diagonal cpu slot that didn't make it into THIS iteration of 'power'Macs and...well, speculate.
Noted some of the Paris keynote. Jobs was talking about OS X being the only boot when 'faster' Macs or 'faster' versions of current models...turn up in 2003.
He said that with the kind of assurance that means we won't be stuck at 500mhz, sorry, 1.25 gig (shipping yet?), for a year plus.
PPS "No. When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. Although, the GUI does need a boost in 10.2. While quickly scrolling down the full length of one of the longer pages here on AI, Chimera can take a good chunk of a second for the thumb of the scroll bar to catch up with the mouse pointer. You can see that lag in other apps, as well."
Imagine the 'blurb' on Intel's page: 'Macs take 2 days to catch up with our low end processors...'
When I posted this I was not suggesting that the PM's were as fast as any other platform, but that the GUI gives the impression that the machines are much slower than they actually are. The actuall processing power of the Pm's is huge, it's not as fast as the latest Intel offerings, but it is still hugh.
Good God, don't you remember SJ demo'ing DVD software rendering on a 1:1 time basis in software. We all thoght that was a breakthrough.
The GUI makes these machines look as though they are much slower than they actually are. I would love to have a new DP with a fully supported DDR FSB, and until Apple produces such a machine I will work with what I have got. I am sure that Apple is acutley aware of the problem and is doing all it can to resolve the issue, but they really are stuck until IBM or MOTO do something.
I think it is just as Dorsal said months ago, good news in the new year.
<strong>it seems that Apple looks to AMD for a replacement of the G4. Not that MacOS X will be running to any PC...simply the Apple hardware will be powered by some other kind of processor. This reinforces the recent rumor that Apple did not show interest to IBM's proposition for the new PPC derivative of Power4. Time will tell.</strong><hr></blockquote>
My French's a bit rusty, but anyway... I'm not biased against AMD (I am biased against Motorola and Intel) so any new player in this match is welcome! As long as Apple can deliver a truly braektrough pro platform they've got my money, I don't really care if it's powered by an IBM or AMD or Alpha (oh, if it only could be true!) processor. As you said, time will tell. But what IBM is going to announce at the MPF looks to be really awesome, so if they decided to use something else it must be even better! C'mon Stevey-boy, show us what you can do!
<strong>I was suggesting that the GUI gives the impression that the machines are much slower than they actually are. The actuall processing power of the Pm's is huge, it's not as fast as the latest Intel offerings, but it is still hugh.
Good God, don't you remember SJ demo'ing DVD software rendering on a 1:1 time basis in software. We all thoght that was a breakthrough.
The GUI makes these machines look as though they are much slower than they actually are.</strong><hr></blockquote>
Couldn't agree more on that! As I said, Altivec performance is not easy to beat, and DVD encoding is a good example.
I remember reading some time ago on Ars that the OS X GUI was totally based on vectors, so I thought that on a G4 it'd have felt snappy as OS 9, but sadly that is not the case. I got 1 GB of Ram and a 667 MHz G4, I got a QE-supported machine, but still the GUI is so slow it actually reminds me of Win98 under VPC... OS X is not essential for me, but I want to look only at the future of things: this is why I don't dare to install OS 9 on my PB, because I might never want to come back to X!
<digression>I remember the 2nd half of 1996, and all the talk about Apple going to buy Be Inc. It certainly was a less mature OS than OpenSTEP, but goddammit it was FAST! On a ridiculously slow PowerBase 200 (PPC 603 with 32 MB Ram) it could play a QT movie and an OpenGL teapot render without ever dropping a frame. Never seen anything like that ever since! OS X needs to be THAT astonishing: instead we have a bloated GUI full of effects nobody wants, a pitiful file system and an even more pitiful suffix-based file naming scheme... Is this innovation??? <img src="graemlins/oyvey.gif" border="0" alt="[No]" /> </digression>
MacRumors has a blurb on what a Apple rep said at the Expo in Paris, that the G5 exists but it wont appear until the end of 2003. To paraphrase. And light modifications to the G4 in the mean time. I think he may have misinterpreted his information because my feeling is that the processor known as G5 should come a little sooner. Say mid-February.
<strong> because my feeling is that the processor known as G5 should come a little sooner. Say mid-February.</strong><hr></blockquote>
If your talking about an IBM cpu, who knows. But if your talking about a G5 from Motorola, that is based on the MPC 8540/8560 architecture(re: Rapid I/O, Ocean switched web, etc.), I'm inclined to believe the end of 2003 date.
As yet, no indication that even samples of the MPC 8540/8560 have been made for distribution to customer's testing. Supposedly, samples were to be made available in the 2nd half of 2002, they weren't and aren't still.
Also, the MPC8540/8560 are NOT intended for desktops and would be less complicated to design/manufacture than a desktop cpu, therefore, my belief is these processors will appear before any desktop version using the above mentioned tech.
Unfortunately, in my opinion, any cpu's for Apple from Motorola, will be modifications of the G4 using the MPX bus which will not migrate to DDR. Granted a G4 on a 0.13µ or even a 0.09µ process will be significantly faster I don't expect bus modifications. Hope I'm very wrong or IBM actually is going to intro a cpu for Apple soon.
The only loophole for the G4 would be if they integrated the memory controller into the G4 die with the die shrink along with an increase to the L3 cache. I've given this some thought: how would you accomplish this without increasing the pin out on the processor? The POWER4 does this rather ingenously. You have 2 buses coming out of the G4; a 64bit MPX bus and a 64bit L3 cache bus. Put the memory controller behind the L3 controller. Internally the controller can connect to the processor core at 256bit while externally it can connect at a fixed frequency (166MHz or 200MHz * 2) at 64bits wide.
This would comply with their commitment to keep MPX SDR (and Apple will have a dedicated bus for PCI, AGP and peripherals) that doesn't have to share the bus with memory calls (but keep the option open for legacy embedded applications) and yet an integraded memory controller will not only be quite useful for desktop purposes but especially for embedded uses. It all falls in with the SOC concept they are pursuing.
<digression>I remember the 2nd half of 1996, and all the talk about Apple going to buy Be Inc. It certainly was a less mature OS than OpenSTEP, but goddammit it was FAST! On a ridiculously slow PowerBase 200 (PPC 603 with 32 MB Ram) it could play a QT movie and an OpenGL teapot render without ever dropping a frame. Never seen anything like that ever since! OS X needs to be THAT astonishing: ...... </digression>
ZoSo - u r so very right! where oh where is an OS like that? Be was built from th ground up to be modern (as was OS X???) & it shone & it loved 2 cpus & it ate multimedia stuff & it ran tonsa stuff fast &this is 5 yrs ago & yeah howcum OS X isn't that yet?
Motorola has said that they will be fabbing at 90nm by the end of the year (4 months ahwad of Intel). Remains to know what they will be fabbing.</strong>
Well, how much do you want to bet that it will be microprocessors?
<strong>As for performance enhancement, speed normally doubles for each sqrt(2) process shrinkage, so if they were to jump from 180nm to 90nm (2 steps - I know it sounds crazy), you could expect 4-5HGz G4s.</strong>
It sounds crazy to me. 5 GHz is little out there, but I can buy into 4 GHz for a G4 at 0.09u. If they do that, it wound mean HiP 7 has been a catastrophic failure. At least HiP 5 was reasonably useful, but so far HiP 7 has been a bust for Moto processors.
That is interesting. I understand the MPX bus is very efficient, but in the "Altivec Programing Anyone" thread @ Arstechnica some very explicit posts showed that the bus was choked if the data could not be held in L2/3 cache?? correct??</strong>
Yes.
But there is a mythology developing in the Mac community, and the computing masses at large, that if the G4 had a DDR processor bus, it's performance problems would be solved. All of the sudden, that 100% improvement in bandwidth would translate to 100% improvement in performance. It's a boogeyman for the G4's performance problems because it's only true for a few cases. 95% of the time, and for 95% of Apple's users, they'll maybe see 5% improvement, maybe. Nearly everyone on this board won't see anything more than 5%.
The G4 needs a few things to really outshine x86. It needs at least 2 out of the 4 (preferably 3) integer units to be able to do pipelined multiplies, divides, and multiply-add ops, a second FPU that can at least do multiplies, divides, and multiply-add ops, a better bus, and maybe more pipeline stages. The bus is but one component of it's problems. The memory controller can be left to Apple. All this talk of backward compatibility for MPX bus seems bogus to me. Moto is in the process of moving its customers to on-die memory controllers and RapidIO. Leaving its highest performance processor with MPX only bifurcates its market.
looks to me like the Mac seems to do pretty well here. haven't had a chance to check out the other benchmarks, but i think this is a decent start. will check others and get back to you.
Comments
<hr></blockquote>
LOL, everyone is thinking as if this new G5 is going to be fabbed by Moto. It's time to unlearn what you know about PPC CPUs and get ready to have IBM rock your world.
Remember that IBM is one of the world's most experienced chip manufacturers. They will be fabbing this "G5" in a brand-new, state of the art facility using new equipment and experienced techs. And yes, IBM keeps their fabs clean (so I hear).
This is not your father's PPC fab. I think we should all prepare to be utterly floored by IBM's efficiency, punctuality, and flat-out rocking potential.
[quote] Show us how bad the G4 is. Links please. <hr></blockquote>
<a href="http://www.blanos.com/benchmark/"; target="_blank">Here's a page full of benchmarks.</a> Just pick any one of the Lightwave scenes, and see how badly the G4 gets it's ass kicked. For another, does anybody remember the magazine that benchmarked PCs & Macs in After Effects? The G4 gets it's hat handed to it there, as well.
You can discount this and justify the Mac's poor performance all you want, guys, but these apps are in the market Apple is HEAVILY targeting right now.
[quote] I think the slowness is a perception problem, and it all stems from the GUI. <hr></blockquote>
No. When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. Although, the GUI does need a boost in 10.2. While quickly scrolling down the full length of one of the longer pages here on AI, Chimera can take a good chunk of a second for the thumb of the scroll bar to catch up with the mouse pointer. You can see that lag in other apps, as well.
You're right about IBM (I have friends that work there), their fabs are clean, and when they release a schedule ... they keep to it. IMO, it's better to have IBM as the PowerMac supplier and Moto as the consumer / laptop supplier. It fits within the capabilities and chosen directions of the companies.
And yes JYD, I'm hold'n on. I want IBM to Rock my Mac World. Only wish I had enough patience.
--
Ed
\t I know a lot of you don't like <a href="http://www.spymac.com"; target="_blank">www.spymac.com</a> but there is an interesting article on their page. It gives us some more evidence of a new chip coming soon. And surprising enough it's based of evidence we all have access to. Go check it out
<strong>When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. </strong><hr></blockquote>
Great link! So I guess now somebody will stop complaining about Lemon Bon Bon's complaints, since it looks like he's right...
Personally I think Altivec performance is really awesome, but every kind of un-optimized code is simply s-l-o-w, maybe with different degrees of slowness but s-l-o-w all the same.
If IBM can really deliver the next PPC, and if its performance will be anywhere comparable to that of a Power4... well, I might sell my car to buy one of those beasts!
ZoSo
<strong>
If IBM can really deliver the next PPC, and if its performance will be anywhere comparable to that of a Power4... well, I might sell my car to buy one of those beasts!
ZoSo</strong><hr></blockquote>
As Lionel is saying <a href="http://www.macbidouille.com/niouzcontenu.php?date=2002-09-12#3406"; target="_blank">here</a>, it seems that Apple looks to AMD for a replacement of the G4. Not that MacOS X will be running to any PC...simply the Apple hardware will be powered by some other kind of processor. This reinforces the recent rumor that Apple did not show interest to IBM's proposition for the new PPC derivative of Power4. Time will tell.
<strong>Unless they repealed it the restriction is still in place.
\t I know a lot of you don't like <a href="http://www.spymac.com"; target="_blank">www.spymac.com</a> but there is an interesting article on their page. It gives us some more evidence of a new chip coming soon. And surprising enough it's based of evidence we all have access to. Go check it out</strong><hr></blockquote>
Doesn't it seem likely that Apple had prototypes of both single and dual 1.25 GHz G4s? (and all other clockspeeds as well). Spymac is assuming that a single CPU in the prototype mobo is something other than a single G4 because.....? This isn't really all that exciting.
" I think the slowness is a perception problem."
That appears to be correct for some Apple users on these boards.
Check out Tom's hardware for additional benches on the latest (unlike the Photoshop stats above...) x86 Lightwave benches. These systems are available for half the price of Apple's 'flagship'.
The XP gets it's head handed to it by the P4.
Ouch.
Lemon Bon Bon
PS. Spymac. Interesting 'aftermath' analysis. Add up HUGE heatsinks, a diagonal cpu slot that didn't make it into THIS iteration of 'power'Macs and...well, speculate.
Noted some of the Paris keynote. Jobs was talking about OS X being the only boot when 'faster' Macs or 'faster' versions of current models...turn up in 2003.
He said that with the kind of assurance that means we won't be stuck at 500mhz, sorry, 1.25 gig (shipping yet?), for a year plus.
PPS "No. When a Lightwave scene that takes two days to render on a P4 takes four on a G4, it's definitely not a perception problem. Although, the GUI does need a boost in 10.2. While quickly scrolling down the full length of one of the longer pages here on AI, Chimera can take a good chunk of a second for the thumb of the scroll bar to catch up with the mouse pointer. You can see that lag in other apps, as well."
Imagine the 'blurb' on Intel's page: 'Macs take 2 days to catch up with our low end processors...'
Something worthy of Apple, eh?
[ 09-12-2002: Message edited by: Lemon Bon Bon ]</p>
When I posted this I was not suggesting that the PM's were as fast as any other platform, but that the GUI gives the impression that the machines are much slower than they actually are. The actuall processing power of the Pm's is huge, it's not as fast as the latest Intel offerings, but it is still hugh.
Good God, don't you remember SJ demo'ing DVD software rendering on a 1:1 time basis in software. We all thoght that was a breakthrough.
The GUI makes these machines look as though they are much slower than they actually are. I would love to have a new DP with a fully supported DDR FSB, and until Apple produces such a machine I will work with what I have got. I am sure that Apple is acutley aware of the problem and is doing all it can to resolve the issue, but they really are stuck until IBM or MOTO do something.
I think it is just as Dorsal said months ago, good news in the new year.
<strong>it seems that Apple looks to AMD for a replacement of the G4. Not that MacOS X will be running to any PC...simply the Apple hardware will be powered by some other kind of processor. This reinforces the recent rumor that Apple did not show interest to IBM's proposition for the new PPC derivative of Power4. Time will tell.</strong><hr></blockquote>
My French's a bit rusty, but anyway... I'm not biased against AMD (I am biased against Motorola and Intel) so any new player in this match is welcome! As long as Apple can deliver a truly braektrough pro platform they've got my money, I don't really care if it's powered by an IBM or AMD or Alpha (oh, if it only could be true!) processor. As you said, time will tell. But what IBM is going to announce at the MPF looks to be really awesome, so if they decided to use something else it must be even better! C'mon Stevey-boy, show us what you can do!
ZoSo
<strong>I was suggesting that the GUI gives the impression that the machines are much slower than they actually are. The actuall processing power of the Pm's is huge, it's not as fast as the latest Intel offerings, but it is still hugh.
Good God, don't you remember SJ demo'ing DVD software rendering on a 1:1 time basis in software. We all thoght that was a breakthrough.
The GUI makes these machines look as though they are much slower than they actually are.</strong><hr></blockquote>
Couldn't agree more on that! As I said, Altivec performance is not easy to beat, and DVD encoding is a good example.
I remember reading some time ago on Ars that the OS X GUI was totally based on vectors, so I thought that on a G4 it'd have felt snappy as OS 9, but sadly that is not the case. I got 1 GB of Ram and a 667 MHz G4, I got a QE-supported machine, but still the GUI is so slow it actually reminds me of Win98 under VPC... OS X is not essential for me, but I want to look only at the future of things: this is why I don't dare to install OS 9 on my PB, because I might never want to come back to X!
<digression>I remember the 2nd half of 1996, and all the talk about Apple going to buy Be Inc. It certainly was a less mature OS than OpenSTEP, but goddammit it was FAST! On a ridiculously slow PowerBase 200 (PPC 603 with 32 MB Ram) it could play a QT movie and an OpenGL teapot render without ever dropping a frame. Never seen anything like that ever since! OS X needs to be THAT astonishing: instead we have a bloated GUI full of effects nobody wants, a pitiful file system and an even more pitiful suffix-based file naming scheme... Is this innovation??? <img src="graemlins/oyvey.gif" border="0" alt="[No]" /> </digression>
ZoSo
<strong> because my feeling is that the processor known as G5 should come a little sooner. Say mid-February.</strong><hr></blockquote>
If your talking about an IBM cpu, who knows. But if your talking about a G5 from Motorola, that is based on the MPC 8540/8560 architecture(re: Rapid I/O, Ocean switched web, etc.), I'm inclined to believe the end of 2003 date.
As yet, no indication that even samples of the MPC 8540/8560 have been made for distribution to customer's testing. Supposedly, samples were to be made available in the 2nd half of 2002, they weren't and aren't still.
Also, the MPC8540/8560 are NOT intended for desktops and would be less complicated to design/manufacture than a desktop cpu, therefore, my belief is these processors will appear before any desktop version using the above mentioned tech.
Unfortunately, in my opinion, any cpu's for Apple from Motorola, will be modifications of the G4 using the MPX bus which will not migrate to DDR. Granted a G4 on a 0.13µ or even a 0.09µ process will be significantly faster I don't expect bus modifications. Hope I'm very wrong or IBM actually is going to intro a cpu for Apple soon.
This would comply with their commitment to keep MPX SDR (and Apple will have a dedicated bus for PCI, AGP and peripherals) that doesn't have to share the bus with memory calls (but keep the option open for legacy embedded applications) and yet an integraded memory controller will not only be quite useful for desktop purposes but especially for embedded uses. It all falls in with the SOC concept they are pursuing.
ZoSo - u r so very right! where oh where is an OS like that? Be was built from th ground up to be modern (as was OS X???) & it shone & it loved 2 cpus & it ate multimedia stuff & it ran tonsa stuff fast &this is 5 yrs ago & yeah howcum OS X isn't that yet?
Motorola has said that they will be fabbing at 90nm by the end of the year (4 months ahwad of Intel). Remains to know what they will be fabbing.</strong>
Well, how much do you want to bet that it will be microprocessors?
<strong>As for performance enhancement, speed normally doubles for each sqrt(2) process shrinkage, so if they were to jump from 180nm to 90nm (2 steps - I know it sounds crazy), you could expect 4-5HGz G4s.</strong>
It sounds crazy to me. 5 GHz is little out there, but I can buy into 4 GHz for a G4 at 0.09u. If they do that, it wound mean HiP 7 has been a catastrophic failure. At least HiP 5 was reasonably useful, but so far HiP 7 has been a bust for Moto processors.
That is interesting. I understand the MPX bus is very efficient, but in the "Altivec Programing Anyone" thread @ Arstechnica some very explicit posts showed that the bus was choked if the data could not be held in L2/3 cache?? correct??</strong>
Yes.
But there is a mythology developing in the Mac community, and the computing masses at large, that if the G4 had a DDR processor bus, it's performance problems would be solved. All of the sudden, that 100% improvement in bandwidth would translate to 100% improvement in performance. It's a boogeyman for the G4's performance problems because it's only true for a few cases. 95% of the time, and for 95% of Apple's users, they'll maybe see 5% improvement, maybe. Nearly everyone on this board won't see anything more than 5%.
The G4 needs a few things to really outshine x86. It needs at least 2 out of the 4 (preferably 3) integer units to be able to do pipelined multiplies, divides, and multiply-add ops, a second FPU that can at least do multiplies, divides, and multiply-add ops, a better bus, and maybe more pipeline stages. The bus is but one component of it's problems. The memory controller can be left to Apple. All this talk of backward compatibility for MPX bus seems bogus to me. Moto is in the process of moving its customers to on-die memory controllers and RapidIO. Leaving its highest performance processor with MPX only bifurcates its market.
Dual AthlonXP 1566 256MB L/W ver7.5 Rad. Ref. Things time:77
Single Intel P4 2400 512MB L/W ver7.5 Rad. Ref. Things time:60 (there was no dual benchmark)
Dual PowerPC G4 1000 256MB L/W ver7.5 Rad. Ref. Things time:56
looks to me like the Mac seems to do pretty well here. haven't had a chance to check out the other benchmarks, but i think this is a decent start. will check others and get back to you.
so lightwave is twice as fast on which system?
G