[quote]So kudos to Intel! It might be heretic for a Mac user to hail Intel but it is not Intels fault that Motorola has trouble.
<hr></blockquote>
Indeed, with the P4, Intel's R+D department is way ahead of everyone else. (Itanic is a different matter.) However, don't come down too hard on Motorola. I can't think of *any* company that has produced anything even close to the P4 at the price range except AMD. And AMD is bleeding money at a fiercesome rate...
Slow, steady speed increments is how its always been until very recently (and with x86 chips being the only exception). I remember even Intel being suprised by the rate that the speed scaled on their processor. (The P3 scaled much faster than original roadmaps.)
Of course, financially, Intel would probably have been better off just giving much of the R+D money
back to the shareholders, for all the extra cash that these fast processors have generated. I seriously doubt that if they'd stuck with 2 GHz processors and incremented slowly from there, and not spent several billion on R+D that they'd have lost several billion in sales. I think Intel has realized that, though and is slowing down the rate at which it throws money at the P4.
Anyway, as long as Dell sells only Intel, Intel's product doesn't have to be the fastest for 97.5% of users...
Off topic now, but wasn't the MPC7400 announced @ the Micropocessor Forum, in advance of Apple announcing they were going to use it??</strong>
Yes.
Motorola had a presentation on the MPC 7400 about 10 months before Apple shipped the 1st Power Mac G4, and if one recalls, Motorola wasn't exactly ready to ship them at the time. They also had a MPC 7450 presentation at uPF the following year and Apple didn't ship a G4+ machine until 15 months later. Lastly, they also had some sort of weird SOI G4+ (Apollo) presentation at uPF prior to Apple shipping machines with them.
If one notices, Motorola hasn't had a microprocessor presentation since then. The uPF01 presentation on the Apollo G4 was there last real microprocessor presentation. I think it's been about 2 years, maybe 27 months.
Well, I for one like this idea. If Eleven really existed and if it performed as well as claimed and if it was killed by Moto's refusal to incorporate ApplePI, then it would have been silly to just let it die there.
So instead, Apple licenses the core design, implements the features they want, and has IBM fab it. This could all be done in total secrecy. Neither Moto nor IBM would have an interest in pre-announcing it.
Would it be too much of a stretch to wonder if this is what Moki was talking about regarding IBM's "sandbagging"?
[Edit:]
Having re-read the thread title, I suppose the stance I'm taking here is a little at odds with Arkangel's. Since the 970's relationship to the POWER4 is a matter of public record, I doubt that the 970 is derived from Eleven in any significant way.
More likely that if Eleven ever sees the light of day, it will come about in something like the manner I described above.
Odds that the 970 is close kin to the (rumored) Eleven: Zilch.
Odds that the 970 uses Mot's AltiVec implementation, licensed via the AIM agreement: Good. Sure, IBM could roll their own, but Mot did a bang-up job designing theirs, IBM can probably license it on favorable terms, and compatibility with the G4 is not an issue. So why not?
I can believe that Mot had the Eleven project going, and I can believe that they had a really hot performer. Their designers are good. I can also believe that they couldn't produce it with any reliability, and decided to shelve it rather than spend more money they didn't have trying to get it going.
I think it's in mothballs (Motballs?) until Mot SPS is back in the black, whenever that is. Good thing the 970 is coming.
"There has been disappointment in the DSP world that the current proposed architecture for G5 (the MPC8500 series) is not AltiVec-enabled, leading many to continue banking on further improvements in the G4 line until release of the G6 -- perhaps by the end of 2003. The G5 architecture does, however, call for a built-in RapidIO interface, which also is something many in the industry are anticipating; the G6 is expected to have AltiVec and RapidIO, although performance parameters are defined by the telecommunications and computer industries, not the military."
So, Motorola's next chip with altivec will be the G6?
I believe that "Eleven" is the fabled pinacle to the Apollo series. IIRC, Motoman was the first to mention it's existence, and he defined it as a HiP7 64-bit part with RIO.
But, above all, remember the link between Apollo and the number "11".
Honestly, I never really connected the G5 aka Eleven with the 970. It's just that since this last MacWorld, which contained several announcements that caught most if not all of us by complete surprise, I began to wonder if maybe the Mac Community is missing something.
Clearly the 970 and Apple make great sense for IBM because as Apple and the open source community develop closer ties, I could see how easy it might become for the unix/linux community to look at the 970 as with or without OS X, but developing an ever greater appreciation for the 970 platform. Good for Apple to be considered to be using a quality platform, lesser barriers to the business market. Obviously this is good for IBM because they are twice as likely to get the sale.
What I can't account for is that while IBM has every right to yell to the heavens the existence of the 970, Apple continues to enjoy shocking the market with the advent of a new product. As we saw last week, they're getting ever better at it. I read someone's blog earlier that noted as Steve made clear that Apple did in fact have a browser, everyone just knew it had to be based on Mozilla. Thus one annoucement, 2 points for catching the masses unawares.
We'll all know soon enough, but I just really have to wonder what would prevent Apple from getting IBM to fab the G5 for them it Moto allowed it. I know AMD probably could fab as well, but for whatever reasons, I'm under the impression they are having enough problems getting their chips out the door.
Oh well, good night all and may tomorrow be a cool one for the Australian Open. Night Anna K. :cool:
"There has been disappointment in the DSP world that the current proposed architecture for G5 (the MPC8500 series) is not AltiVec-enabled, leading many to continue banking on further improvements in the G4 line until release of the G6 -- perhaps by the end of 2003. The G5 architecture does, however, call for a built-in RapidIO interface, which also is something many in the industry are anticipating; the G6 is expected to have AltiVec and RapidIO, although performance parameters are defined by the telecommunications and computer industries, not the military."
So, Motorola's next chip with altivec will be the G6?</strong><hr></blockquote>
I wouldnt count on it even then. IBM has a vested interest in the success of the Power PC in the desktop/server market place becouse they use them in their own products. Motorolla hanst used them in a retail computer since the clones were killed.
IBM has made moves to free themselves of Microsoft in their investments in Linex, this hints at a long term goal of limiting or droping Microsoft/Intel products from their lin-up. They may not achieve this goal, but I would be suprised if it is not an internal goal. They could also start a "clone" buisness in Linex boxes with a competative CPU, sort of like they were working on with the CHRP when AIM was "new". And as a plus, they have a hungry client, Apple, for these new chips while they are building further markets for their homegrown chips and computers.
Apple has 2 other alternatives, move to another processor (AMD), or design their own and contract out the fab of the chip. The first of these might work in a year or 2, once the move to OS X is truely complete. The second would be a long shot, that would have had to start a few years ago if it were to save Apple today. It has its merits in the flexability that it could afford Apple, but a lot of pitfalls in cost.
All in all the only solution for Apple for the next 12 months is IBM. I would like to see the IBM have a G3+++ (770MXX?) to replace the G4 so Apple could drop Moto altogether (good design, substandard production) but I dont see that happening either. The G4 will be around for the next 24-36 months in Apples consumer lines.
<strong>What I can't account for is that while IBM has every right to yell to the heavens the existence of the 970, Apple continues to enjoy shocking the market with the advent of a new product. </strong><hr></blockquote>
And this is why people think IBM is 'sandbagging'. There's a LOT of different areas from the official announcement where the wording was not definative, and can be interpreted in a more positive light. Plus, IBM is extremely unlikely to engage in hype of any sort - they are under anti-trust injunctions from shenanigans in the 80's. -> IBM Vaporware leads to big fines.
Key things: (paraphrased, sorry)
The IBM PPC 970 will be widely available 2H03.
The IBM PPC 970, first in a new series of processors.
Estimated specmarks ... (somewhere around 5x a G4 in FP)
The IMP PPC 970's bus aggregate bandwith = 6.4 Gbps.
Ok. If any of those end up being materially false, -> stiff fines.
So nothing says 'widely available' doesn't mean Apple gets them before that. Nothing says that other chips in this family might be dual core and ship RSN. Nothing in there says that the estimate is the most conservative they could do to hedge against fines.
I'll be perfectly content if the ppc970 ships meeting precisely that timetable & numbers (look real FP!).
And I think that if Apple shipped a dual ppc970 precisely on that timetable (with a month or two to accumulate stock) -> one heck of a bake off. Dual CPUs each with dual floating point units + two altivec units -> one heck of a lot of floating point if you are hand coding your plugin. Or whatever. Say Shake.
I think Steve Jobs with a dual ppc 970, a year to prepare after buying something like Shake, and some axes to grind might be just a TAD dangerous.
Color me optimistic if you like, and I don't necessarily see this as being _cheap_, but as the new $5000 flagship.... I think it would still be shocking enough for SJ to amplifiy the RDF a little
Processor design is simply too expensive for Apple to get into it... they might struggle along for a year or two, but their effort would be pretty much doomed to failure (i.e. they'd run out of money or be unable to keep up with Intel).</strong>
Are you really sure about that Programmer? Processor design is eminently affordable for Apple. Processor manufacturing is not.
As a real world example, just look at Transmeta, a 300 employee $150 million/year company just produced their second generation 8-issue wide VLIW processor. Or SGI with their MIPS chips and proprietary core logic chipsets. Or even AMD itself, which is about the same size or smaller than Apple. And they have fabs to maintain.
Apple can get into processor design if they wanted to. They would have to if they want to keep their market share or improve upon it. Whether their chip designs will be able to compete is more a measure of their design talent than the amount of money funding available.
As for Motorola, I'd just give up on them now. I gave up them a long long time ago. They are turning themselves into a communications company, and they don't have the funding to move to next generation fabs and must rely on others. Their semiconductor business will be in a continuous state of trouble for the foreseeable future because of this.
[/QB]<hr></blockquote>
Transmeta, MIPS, and Sparc aren't competing with particuarly well with the high end Intel chips. AMD is, but AMD is also almost completely focused on that (and their flash chips, which sell well) ... and they aren't in the best shape right now it sounds like. Apple flirted with their own PowerPC designs back in '95 or so when they bought Exponential, but that never saw the light of day. They were also involved in Somerset from the beginning, but I was told they didn't really have people doing the design work -- it was more of a consulting role. They may have rights to some of the designs, I don't know, but at this point those designs are pretty old and probably don't carry much value in designing the latest, greatest PowerPC chip. IBM has the advantage of having the POWER4/5/6 work to draw on, with the added bonus of being intimately involved with their own latest fab technologies.
You're probably right, they could start a processor design team... I was concerned more whether this would be a sensible thing to do. It certainly isn't to be taken on lightly.
BTW: My guess is that IBM did their own VMX design. There may be common elements from the original joint development of it, but given the couple of IBM research papers on the subject and the 970's much longer pipelines I highly doubt they used the Motorola designs verbatim.
Transmeta, MIPS, and Sparc aren't competing with particuarly well with the high end Intel chips.</strong>
Granted. But I don't think it means that Apple can't. SGI designs it's MIPS RXX000 processors, not MIPS (who just does IP now), though they were one and the same way back. They lost their direction with their Pentium and Itanium infatuation, and haven't recovered. Transmeta isn't competing on the high end, they are competing on the MIPS/watt front, and Transmeta will have good potential if Tablets takeoff. Sun is trudging along fine with their SPARC chips, not top of the line, but within respectability.
The high end competitors were or are PA-RISC, Power4 and the Alpha. They all competed fine with Intel and they were all fabbed on less advanced fabs to boot. It's about talent, not necessarilly money. Isn't it a bit coincidental that Intel buys out Alpha developers, and voila, Pentium 4 has multithreading?
<strong>Apple flirted with their own PowerPC designs back in '95 or so when they bought Exponential, but that never saw the light of day.</strong>
Apple invested in Exponential, sort of like they invested in Somerset or Microsoft investing in some company. They never bought them out. Exponential tried to keep themselves alive through litigating Apple, since they thought Apple was going to use their chips. It never saw the light of day because their 0.35u bi-CMOS chip was a lame-brain dead-end idea to base a processor on, and their processor didn't compete well against IBM/Moto's 604e/750 which were fabbed on more advanced CMOS tech.
<strong>IBM has the advantage of having the POWER4/5/6 work to draw on, with the added bonus of being intimately involved with their own latest fab technologies.</strong>
No problems with that. Apple was just late in recognizing Moto's fabs were going to be bad and lagging. This time around, they have to be intimately involved with IBM.
<strong>You're probably right, they could start a processor design team... I was concerned more whether this would be a sensible thing to do. It certainly isn't to be taken on lightly.</strong>
It's only their future as a PC hardware company...
<strong>BTW: My guess is that IBM did their own VMX design. There may be common elements from the original joint development of it, but given the couple of IBM research papers on the subject and the 970's much longer pipelines I highly doubt they used the Motorola designs verbatim.</strong>
No doubt. The AltiVec implementation can be totally different from the AltiVec ISA. Just like the Athlon and P4 are x86 processors, yet they are totally different architecturally.
It's possible that Apple had designed the Altivec portion of the 970 since they were Motorolas partner in getting this technology on the G4 in the first place. Their engineers may have been ahead WRT Altivec, more so than an IBM engineer.
Comments
<hr></blockquote>
Indeed, with the P4, Intel's R+D department is way ahead of everyone else. (Itanic is a different matter.) However, don't come down too hard on Motorola. I can't think of *any* company that has produced anything even close to the P4 at the price range except AMD. And AMD is bleeding money at a fiercesome rate...
Slow, steady speed increments is how its always been until very recently (and with x86 chips being the only exception). I remember even Intel being suprised by the rate that the speed scaled on their processor. (The P3 scaled much faster than original roadmaps.)
Of course, financially, Intel would probably have been better off just giving much of the R+D money
back to the shareholders, for all the extra cash that these fast processors have generated. I seriously doubt that if they'd stuck with 2 GHz processors and incremented slowly from there, and not spent several billion on R+D that they'd have lost several billion in sales. I think Intel has realized that, though and is slowing down the rate at which it throws money at the P4.
Anyway, as long as Dell sells only Intel, Intel's product doesn't have to be the fastest for 97.5% of users...
[ 01-14-2003: Message edited by: Tom West ]</p>
Off topic now, but wasn't the MPC7400 announced @ the Micropocessor Forum, in advance of Apple announcing they were going to use it??</strong>
Yes.
Motorola had a presentation on the MPC 7400 about 10 months before Apple shipped the 1st Power Mac G4, and if one recalls, Motorola wasn't exactly ready to ship them at the time. They also had a MPC 7450 presentation at uPF the following year and Apple didn't ship a G4+ machine until 15 months later. Lastly, they also had some sort of weird SOI G4+ (Apollo) presentation at uPF prior to Apple shipping machines with them.
If one notices, Motorola hasn't had a microprocessor presentation since then. The uPF01 presentation on the Apollo G4 was there last real microprocessor presentation. I think it's been about 2 years, maybe 27 months.
I'd say.
THT kind nailed the hammer on the Motorola head for me.
Lemon Bon Bon
So instead, Apple licenses the core design, implements the features they want, and has IBM fab it. This could all be done in total secrecy. Neither Moto nor IBM would have an interest in pre-announcing it.
Would it be too much of a stretch to wonder if this is what Moki was talking about regarding IBM's "sandbagging"?
[Edit:]
Having re-read the thread title, I suppose the stance I'm taking here is a little at odds with Arkangel's. Since the 970's relationship to the POWER4 is a matter of public record, I doubt that the 970 is derived from Eleven in any significant way.
More likely that if Eleven ever sees the light of day, it will come about in something like the manner I described above.
[ 01-14-2003: Message edited by: boy_analog ]</p>
Odds that the 970 uses Mot's AltiVec implementation, licensed via the AIM agreement: Good. Sure, IBM could roll their own, but Mot did a bang-up job designing theirs, IBM can probably license it on favorable terms, and compatibility with the G4 is not an issue. So why not?
I can believe that Mot had the Eleven project going, and I can believe that they had a really hot performer. Their designers are good. I can also believe that they couldn't produce it with any reliability, and decided to shelve it rather than spend more money they didn't have trying to get it going.
I think it's in mothballs (Motballs?) until Mot SPS is back in the black, whenever that is. Good thing the 970 is coming.
<a href="http://www.infinibandta.org/parchives/clippings/msg00072.html"; target="_blank">http://www.infinibandta.org/parchives/clippings/msg00072.html</a>;
"There has been disappointment in the DSP world that the current proposed architecture for G5 (the MPC8500 series) is not AltiVec-enabled, leading many to continue banking on further improvements in the G4 line until release of the G6 -- perhaps by the end of 2003. The G5 architecture does, however, call for a built-in RapidIO interface, which also is something many in the industry are anticipating; the G6 is expected to have AltiVec and RapidIO, although performance parameters are defined by the telecommunications and computer industries, not the military."
So, Motorola's next chip with altivec will be the G6?
<strong>Can someone explain what this "Eleven" is that people are mentioning?</strong><hr></blockquote>
Architosh had an article a few months ago that described what "eleven" is.
But, above all, remember the link between Apollo and the number "11".
Clearly the 970 and Apple make great sense for IBM because as Apple and the open source community develop closer ties, I could see how easy it might become for the unix/linux community to look at the 970 as with or without OS X, but developing an ever greater appreciation for the 970 platform. Good for Apple to be considered to be using a quality platform, lesser barriers to the business market. Obviously this is good for IBM because they are twice as likely to get the sale.
What I can't account for is that while IBM has every right to yell to the heavens the existence of the 970, Apple continues to enjoy shocking the market with the advent of a new product. As we saw last week, they're getting ever better at it. I read someone's blog earlier that noted as Steve made clear that Apple did in fact have a browser, everyone just knew it had to be based on Mozilla. Thus one annoucement, 2 points for catching the masses unawares.
We'll all know soon enough, but I just really have to wonder what would prevent Apple from getting IBM to fab the G5 for them it Moto allowed it. I know AMD probably could fab as well, but for whatever reasons, I'm under the impression they are having enough problems getting their chips out the door.
Oh well, good night all and may tomorrow be a cool one for the Australian Open. Night Anna K. :cool:
<strong>From this link...
<a href="http://www.infinibandta.org/parchives/clippings/msg00072.html"; target="_blank">http://www.infinibandta.org/parchives/clippings/msg00072.html</a>;
"There has been disappointment in the DSP world that the current proposed architecture for G5 (the MPC8500 series) is not AltiVec-enabled, leading many to continue banking on further improvements in the G4 line until release of the G6 -- perhaps by the end of 2003. The G5 architecture does, however, call for a built-in RapidIO interface, which also is something many in the industry are anticipating; the G6 is expected to have AltiVec and RapidIO, although performance parameters are defined by the telecommunications and computer industries, not the military."
So, Motorola's next chip with altivec will be the G6?</strong><hr></blockquote>
I wouldnt count on it even then. IBM has a vested interest in the success of the Power PC in the desktop/server market place becouse they use them in their own products. Motorolla hanst used them in a retail computer since the clones were killed.
IBM has made moves to free themselves of Microsoft in their investments in Linex, this hints at a long term goal of limiting or droping Microsoft/Intel products from their lin-up. They may not achieve this goal, but I would be suprised if it is not an internal goal. They could also start a "clone" buisness in Linex boxes with a competative CPU, sort of like they were working on with the CHRP when AIM was "new". And as a plus, they have a hungry client, Apple, for these new chips while they are building further markets for their homegrown chips and computers.
Apple has 2 other alternatives, move to another processor (AMD), or design their own and contract out the fab of the chip. The first of these might work in a year or 2, once the move to OS X is truely complete. The second would be a long shot, that would have had to start a few years ago if it were to save Apple today. It has its merits in the flexability that it could afford Apple, but a lot of pitfalls in cost.
All in all the only solution for Apple for the next 12 months is IBM. I would like to see the IBM have a G3+++ (770MXX?) to replace the G4 so Apple could drop Moto altogether (good design, substandard production) but I dont see that happening either. The G4 will be around for the next 24-36 months in Apples consumer lines.
<strong>What I can't account for is that while IBM has every right to yell to the heavens the existence of the 970, Apple continues to enjoy shocking the market with the advent of a new product. </strong><hr></blockquote>
And this is why people think IBM is 'sandbagging'. There's a LOT of different areas from the official announcement where the wording was not definative, and can be interpreted in a more positive light. Plus, IBM is extremely unlikely to engage in hype of any sort - they are under anti-trust injunctions from shenanigans in the 80's. -> IBM Vaporware leads to big fines.
Key things: (paraphrased, sorry)
The IBM PPC 970 will be widely available 2H03.
The IBM PPC 970, first in a new series of processors.
Estimated specmarks ... (somewhere around 5x a G4 in FP)
The IMP PPC 970's bus aggregate bandwith = 6.4 Gbps.
Ok. If any of those end up being materially false, -> stiff fines.
So nothing says 'widely available' doesn't mean Apple gets them before that. Nothing says that other chips in this family might be dual core and ship RSN. Nothing in there says that the estimate is the most conservative they could do to hedge against fines.
I'll be perfectly content if the ppc970 ships meeting precisely that timetable & numbers (look real FP!).
And I think that if Apple shipped a dual ppc970 precisely on that timetable (with a month or two to accumulate stock) -> one heck of a bake off. Dual CPUs each with dual floating point units + two altivec units -> one heck of a lot of floating point if you are hand coding your plugin. Or whatever. Say Shake.
I think Steve Jobs with a dual ppc 970, a year to prepare after buying something like Shake, and some axes to grind might be just a TAD dangerous.
Color me optimistic if you like, and I don't necessarily see this as being _cheap_, but as the new $5000 flagship.... I think it would still be shocking enough for SJ to amplifiy the RDF a little
[ 01-15-2003: Message edited by: Nevyn ]</p>
<strong>[qb]Originally posted by Programmer:
Processor design is simply too expensive for Apple to get into it... they might struggle along for a year or two, but their effort would be pretty much doomed to failure (i.e. they'd run out of money or be unable to keep up with Intel).</strong>
Are you really sure about that Programmer? Processor design is eminently affordable for Apple. Processor manufacturing is not.
As a real world example, just look at Transmeta, a 300 employee $150 million/year company just produced their second generation 8-issue wide VLIW processor. Or SGI with their MIPS chips and proprietary core logic chipsets. Or even AMD itself, which is about the same size or smaller than Apple. And they have fabs to maintain.
Apple can get into processor design if they wanted to. They would have to if they want to keep their market share or improve upon it. Whether their chip designs will be able to compete is more a measure of their design talent than the amount of money funding available.
As for Motorola, I'd just give up on them now. I gave up them a long long time ago. They are turning themselves into a communications company, and they don't have the funding to move to next generation fabs and must rely on others. Their semiconductor business will be in a continuous state of trouble for the foreseeable future because of this.
[/QB]<hr></blockquote>
Transmeta, MIPS, and Sparc aren't competing with particuarly well with the high end Intel chips. AMD is, but AMD is also almost completely focused on that (and their flash chips, which sell well) ... and they aren't in the best shape right now it sounds like. Apple flirted with their own PowerPC designs back in '95 or so when they bought Exponential, but that never saw the light of day. They were also involved in Somerset from the beginning, but I was told they didn't really have people doing the design work -- it was more of a consulting role. They may have rights to some of the designs, I don't know, but at this point those designs are pretty old and probably don't carry much value in designing the latest, greatest PowerPC chip. IBM has the advantage of having the POWER4/5/6 work to draw on, with the added bonus of being intimately involved with their own latest fab technologies.
You're probably right, they could start a processor design team... I was concerned more whether this would be a sensible thing to do. It certainly isn't to be taken on lightly.
BTW: My guess is that IBM did their own VMX design. There may be common elements from the original joint development of it, but given the couple of IBM research papers on the subject and the 970's much longer pipelines I highly doubt they used the Motorola designs verbatim.
[ 01-15-2003: Message edited by: Programmer ]</p>
Transmeta, MIPS, and Sparc aren't competing with particuarly well with the high end Intel chips.</strong>
Granted. But I don't think it means that Apple can't. SGI designs it's MIPS RXX000 processors, not MIPS (who just does IP now), though they were one and the same way back. They lost their direction with their Pentium and Itanium infatuation, and haven't recovered. Transmeta isn't competing on the high end, they are competing on the MIPS/watt front, and Transmeta will have good potential if Tablets takeoff. Sun is trudging along fine with their SPARC chips, not top of the line, but within respectability.
The high end competitors were or are PA-RISC, Power4 and the Alpha. They all competed fine with Intel and they were all fabbed on less advanced fabs to boot. It's about talent, not necessarilly money. Isn't it a bit coincidental that Intel buys out Alpha developers, and voila, Pentium 4 has multithreading?
<strong>Apple flirted with their own PowerPC designs back in '95 or so when they bought Exponential, but that never saw the light of day.</strong>
Apple invested in Exponential, sort of like they invested in Somerset or Microsoft investing in some company. They never bought them out. Exponential tried to keep themselves alive through litigating Apple, since they thought Apple was going to use their chips. It never saw the light of day because their 0.35u bi-CMOS chip was a lame-brain dead-end idea to base a processor on, and their processor didn't compete well against IBM/Moto's 604e/750 which were fabbed on more advanced CMOS tech.
<strong>IBM has the advantage of having the POWER4/5/6 work to draw on, with the added bonus of being intimately involved with their own latest fab technologies.</strong>
No problems with that. Apple was just late in recognizing Moto's fabs were going to be bad and lagging. This time around, they have to be intimately involved with IBM.
<strong>You're probably right, they could start a processor design team... I was concerned more whether this would be a sensible thing to do. It certainly isn't to be taken on lightly.</strong>
It's only their future as a PC hardware company...
<strong>BTW: My guess is that IBM did their own VMX design. There may be common elements from the original joint development of it, but given the couple of IBM research papers on the subject and the 970's much longer pipelines I highly doubt they used the Motorola designs verbatim.</strong>
No doubt. The AltiVec implementation can be totally different from the AltiVec ISA. Just like the Athlon and P4 are x86 processors, yet they are totally different architecturally.