actually i remember way back when the G4 was introduced, i believe the head of the microprocossor report who I think had some involvement in atlivec (Peter Glaskosky?) said at the time that although Atlivec was well designed the G4 would fall further and further behind the competition because of its inability to scale as well... He was pretty vocal about it at the time...if only Apple/moto had listened back then...He stated this upon the chips introduction before all the problems that followed... how could he know this an others not? Oh well....well as far as current chips he seems to like the 970 and feels it should be competative...so hopefully the future will bode better than the past....
actually i remember way back when the G4 was introduced, i believe the head of the microprocossor report who I think had some involvement in atlivec (Peter Glaskosky?) said at the time that although Atlivec was well designed the G4 would fall further and further behind the competition because of its inability to scale as well... He was pretty vocal about it at the time...if only Apple/moto had listened back then...He stated this upon the chips introduction before all the problems that followed... how could he know this an others not?</strong>
Feh. It isn't about scaling that much. It's about microprocessor design and fab techonology. Anyone with basic knowledge of the semiconductor industry and microprocessor design had that figured out a long long time ago. G4 processors had 4 or 7 stage pipelines. x86 processors had 10 to 20 stage pipelines. Motorola fabs were following the pack. Intel's fabs were leading the pack. That's all one needed to know.
then the question is..Why didn't they know? And if they knew then why start their advertising campaign with "Supercomputer" for the desktop...
just to bring back some more memories....I remember Andy Grove being interview on tv after the powerpc was first introduced and was much hyped...and he was asked about what he was going to do and how would he be able to keep up? Everyone at the time was like cisc is over with and hitting a wall...and powerpc is is risc and blah blah..yeh I know it never had a good implementation at first...and then bad fabs and then blah blah...
another thought...as some know Sun and Apple came close to merging twice... now I wonder what would have happened processor wise...would we have had a Solaris based OS on Sparc?...
more thoughts...if Apple had gone with BEOS would they have gone straight to x86?....hopefully in a year from now the whole processor worries will be over with....and so will motorolla...
<strong>He was pretty vocal about it at the time...if only Apple/moto had listened back then...He stated this upon the chips introduction before all the problems that followed... how could he know this an others not?</strong><hr></blockquote>
He didn't know anything, he took a bet and he's now remembered, in hindsight, as being spot on. His actual reasoning for this problem could be entirely wrong.
If he'd got it wrong would we mention his name either way?
Could we know his reasoning? Did he give his reasons WHY he thought the PPC would fall behind?
IIRC, even Jobs, when he was hocking Next cubes, was pretty sure nobody would get too far ahead of Intel for long.
But back to my original questions?
Did he have valid reasons? The pipeline vs fab tech projections seem valid enough.
Why didn't anyone listen? (Assuming it was not unfounded at the time) Did they think Moto/IBM would add pipeline stages as needed to crank up the Hz? Did they expect the core to be re-worked more substantially and more often? And if they did expect that, did they run into technical problems that immediately made more substantial redesigns too expensive/impractical? DOes anyone think the drawing rooms of Moto and IBM are littered with one or two significant false starts?
Is it possible that knowing what they knew, they didn't expect Intel to go so far so fast?
<strong>Did they expect the core to be re-worked more substantially and more often? And if they did expect that, did they run into technical problems that immediately made more substantial redesigns too expensive/impractical?</strong><hr></blockquote>
They did expect to go through more frequent core revisions than the Pentium line does. That was part of the plan.
A number of things happened, like IBM balking at AltiVec, and Mot squandering billions of dollars on satellites and frittering away their legacy under the direction of a poor CEO.
[quote]<strong>Is it possible that knowing what they knew, they didn't expect Intel to go so far so fast?</strong><hr></blockquote>
Intel didn't expect Intel to go so far so fast. They had settled into a nice, comfortable routine of designing cores to scale for years and scaling them up at nice, leisurely and highly profitable pace while erstwhile competitors fought with each other for the bottom end. Then AMD decided to get Intel in a race. Intel's architecture at the time, the P3, ran out of steam pretty quickly. Now AMD's has. The P4 is designed to be a better fighter in this new territory, and presumably AMD's next offering will be as well.
And, of course, this happened at just about the time when Mot and IBM were both frantically trying to get around the L1 cache timing bug that capped their processors at 500MHz. Mot got over the hump a year ago, but they're so cash-starved (for reasons unrelated to their semiconductor division) that the expensive nature of high-end processor development looks unattractive to them (that, or they're keeping one hell of a secret). IBM was a bit ahead of them, but they didn't have a high-end PPC until the 970.
So the answer is, basically: They had good intentions and a reasonable game plan, but bad luck, bad timing, and a speed war on the other side that nobody foresaw.
***CONFIRMED**** The newest G4 processor is slower than the newest Pentium 4 processor....(Sorry, just had to do that. Never posted a "CONFIRMED" yet, and this thread deserved something like this....) <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
Comments
<strong>Apparently, MOT chips run at less than half the clock speed. They say it's been like that for 3 years.
wow.</strong><hr></blockquote>
Please could you elaborate ?
Less than half the clock speed of what ? :
- the P4
- the fastest sample of the G4
- the G4 at 1,25 ghz clock only at 600 mhz in reality ?
Some precisions are needed.
IBL
smilie breakdown.
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> :eek: <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
bye
<img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
actually i remember way back when the G4 was introduced, i believe the head of the microprocossor report who I think had some involvement in atlivec (Peter Glaskosky?) said at the time that although Atlivec was well designed the G4 would fall further and further behind the competition because of its inability to scale as well... He was pretty vocal about it at the time...if only Apple/moto had listened back then...He stated this upon the chips introduction before all the problems that followed... how could he know this an others not?</strong>
Feh. It isn't about scaling that much. It's about microprocessor design and fab techonology. Anyone with basic knowledge of the semiconductor industry and microprocessor design had that figured out a long long time ago. G4 processors had 4 or 7 stage pipelines. x86 processors had 10 to 20 stage pipelines. Motorola fabs were following the pack. Intel's fabs were leading the pack. That's all one needed to know.
just to bring back some more memories....I remember Andy Grove being interview on tv after the powerpc was first introduced and was much hyped...and he was asked about what he was going to do and how would he be able to keep up? Everyone at the time was like cisc is over with and hitting a wall...and powerpc is is risc and blah blah..yeh I know it never had a good implementation at first...and then bad fabs and then blah blah...
another thought...as some know Sun and Apple came close to merging twice... now I wonder what would have happened processor wise...would we have had a Solaris based OS on Sparc?...
more thoughts...if Apple had gone with BEOS would they have gone straight to x86?....hopefully in a year from now the whole processor worries will be over with....and so will motorolla...
hello moto
<strong>He was pretty vocal about it at the time...if only Apple/moto had listened back then...He stated this upon the chips introduction before all the problems that followed... how could he know this an others not?</strong><hr></blockquote>
He didn't know anything, he took a bet and he's now remembered, in hindsight, as being spot on. His actual reasoning for this problem could be entirely wrong.
If he'd got it wrong would we mention his name either way?
Could we know his reasoning? Did he give his reasons WHY he thought the PPC would fall behind?
IIRC, even Jobs, when he was hocking Next cubes, was pretty sure nobody would get too far ahead of Intel for long.
But back to my original questions?
Did he have valid reasons? The pipeline vs fab tech projections seem valid enough.
Why didn't anyone listen? (Assuming it was not unfounded at the time) Did they think Moto/IBM would add pipeline stages as needed to crank up the Hz? Did they expect the core to be re-worked more substantially and more often? And if they did expect that, did they run into technical problems that immediately made more substantial redesigns too expensive/impractical? DOes anyone think the drawing rooms of Moto and IBM are littered with one or two significant false starts?
Is it possible that knowing what they knew, they didn't expect Intel to go so far so fast?
In short, WTF happened?
<strong>Did they expect the core to be re-worked more substantially and more often? And if they did expect that, did they run into technical problems that immediately made more substantial redesigns too expensive/impractical?</strong><hr></blockquote>
They did expect to go through more frequent core revisions than the Pentium line does. That was part of the plan.
A number of things happened, like IBM balking at AltiVec, and Mot squandering billions of dollars on satellites and frittering away their legacy under the direction of a poor CEO.
[quote]<strong>Is it possible that knowing what they knew, they didn't expect Intel to go so far so fast?</strong><hr></blockquote>
Intel didn't expect Intel to go so far so fast. They had settled into a nice, comfortable routine of designing cores to scale for years and scaling them up at nice, leisurely and highly profitable pace while erstwhile competitors fought with each other for the bottom end. Then AMD decided to get Intel in a race. Intel's architecture at the time, the P3, ran out of steam pretty quickly. Now AMD's has. The P4 is designed to be a better fighter in this new territory, and presumably AMD's next offering will be as well.
And, of course, this happened at just about the time when Mot and IBM were both frantically trying to get around the L1 cache timing bug that capped their processors at 500MHz. Mot got over the hump a year ago, but they're so cash-starved (for reasons unrelated to their semiconductor division) that the expensive nature of high-end processor development looks unattractive to them (that, or they're keeping one hell of a secret). IBM was a bit ahead of them, but they didn't have a high-end PPC until the 970.
So the answer is, basically: They had good intentions and a reasonable game plan, but bad luck, bad timing, and a speed war on the other side that nobody foresaw.
[ 12-13-2002: Message edited by: Amorph ]</p>
IBL.
Lame thread. Kill it.