<strong>Wouldn't you consider it a design failure if it didn't out-produce a 604e?</strong><hr></blockquote>
"failure"? No -- it met its design goals to be a high performance 64-bit PowerPC. The 604 met its goal to be a high performance 32-bit PowerPC. Almost nobody at the time wanted a 64-bit PowerPC, so the 604/604e was a better solution (simpler, cheaper, faster). Remember, running 32-bit code is faster than running 64-bit code.
[quote]Originally posted by Transcendental Octothorpe:
<strong>
Kurt and Brendon
No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
</strong><hr></blockquote>
If you are right (and I hope you are) I would guess that Pixar is swimming in them. I'm guessing that Pixar either is or soon will be the test bed for new Apple hardware. Steve would have much better control of leaks that way. I must have just mis-read your post. Also I believe that the Apple system chip is the main chip to be redesigned, but the memory system may require major rework to keep latency down, don't know much hear.
[quote]Originally posted by Transcendental Octothorpe:
<strong>No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
But, alas, I probably just jinxed it
(BTW, what the heck does that [ ] smilie mean?)</strong><hr></blockquote>
If you're right (and I sure hope you are), I would expect the actual systems to be announced at MWNY, not earlier. It isn't only hardware here. These are not moto chips, as well as being 64 bit, and it's reasonable to expect a few nasty surprises when Apple starts testing them with a wide range of third party software. I would be delighted to see them actually ship by Septmber. I have rarely seen the optimistic predictions come true with the towers - the pessimists most always turn out to be right. I'm really looking forward to the hype for MWNY since it was almost nonexistent for MWSF.
BTW, I always thought that signified embarassment.
I'd be willing to bet that Apple has been testing with 970 samples for a while already. IBM's announcement of "sample Q1 03" is probably just the public availability of samples, and it doesn't actually say anything in particular about when Apple started getting them.
More than likely Apple will wait for full production, I believe, and I'm guessing now that Apple is selling around 250,000 to 300,000 PowerMacs per quarter. I could see that number tripling easy, so around 1,000,000 per quarter. </strong><hr></blockquote>
176,000 powermacs last quarter, this quarter's results aren't out. That's up 10,000 over the previous quarter - which is a _very_ small jump due to the speed bumps in August. If Apple doubled that I think there'd be dancing in the streets. Just saying that a million powermacs a quarter is too much to aim for - they can't stock up THAT much inventory unless the computer is amazingly compelling. If Apple is prepared for double the current rate (say 400,000), it is pretty reasonable to turn around and say 'jeez, we sold out!'.
BTW does anybody know exactly what is wrong with the Motorola fab? I would think that these yield problems would extend to their entire line of processors including the embedded ones and be very bad for business.
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.
[quote]Originally posted by Transcendental Octothorpe:
<strong>
.
Kurt and Brendon
No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
</strong><hr></blockquote>
If you have a source with real information than that is great. I certainly would like to see the 970 as soon as possible. My thoughts were pure guesses with the limited information I have seen on the web. I know enough about chip design (which isn' much) to know that designing the support chips and boards is no small task. Especially when you think that the bus architecture and speeds are changing significantly. This will certainly be no little speed bump and a whole new skill set for Apple engineers.
Just curious to anyone who knows chip production. If they are planning to release the chips in the second half, does that mean the chips are finalized? Or are there last minute revisions possible?
I used to think that the change in memory types in the last Powermac revision meant that the 970 wouldn't be ready for some time. There was not a huge improvement in the overall speed of the machines and it must have been a pretty big investment if the 970's were close. But seeing it in the new 17" Powerbook seems to indicate that that architecture is probably due to arrive in the next generation of iMacs and generally work its way down the line. That would make sense as to why they developed it.
You said the design was frozen. Any hints as to what it might look like? I.E. memory type, single/dual processors, new case, BENCHMARKS!!
<strong>BTW does anybody know exactly what is wrong with the Motorola fab? I would think that these yield problems would extend to their entire line of processors including the embedded ones and be very bad for business.
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.</strong><hr></blockquote>
Just curious to anyone who knows chip production. If they are planning to release the chips in the second half, does that mean the chips are finalized? Or are there last minute revisions possible?
</strong><hr></blockquote>
Very minor revisions are common (call "steppings" by some companies, I think). These are generally just bug fixes, which (in addition to outright incorrect function) include correcting problems that limit clock rate and yields.
Apple's work on a DDR chipset doesn't just get "thrown away". It will migrate to the low end, plus these things are usually designed in a modular fashion so the design & layout work can be probably be reused in a RapidIO or HT based chipset. Internally they could even design the thing in 2 halves which internally communicate using HyperTransport, so replacing one half or the other isn't too big a deal.
Apple's chipset includes several pieces of functionality:
MPX bus interface
AGP bus interface
Memory controller
I/O system interface
I split AGP out because it has higher bandwidth requirements that the rest of the I/O system so it might be appropriate to attach it directly to the memory controller. These pieces of functionality will likely need to be divided between a couple of chips in a 970 system, and some mechanism like HT or RIO introduced to communicate between them. This division may have already been done internally to the existing chipset.
I would think IBM would have a pretty easy time to get these out early for Apple since clock speed doesn't really matter at this point since moto is at 1.2 and these new chips could be used in quads if apple wanted too. Quad 1.2-1.4 ghz 970's sound nice to me
Also i read a while back on news.com way before talk of the 970 about IBM's megaherz goals...and I beleive I remember them mentioning being above 4 ghz around 2004 so I believe these puppies will scale well eventually unlike some other powerpc see makers
176,000 powermacs last quarter, this quarter's results aren't out. That's up 10,000 over the previous quarter - which is a _very_ small jump due to the speed bumps in August. If Apple doubled that I think there'd be dancing in the streets. Just saying that a million powermacs a quarter is too much to aim for - they can't stock up THAT much inventory unless the computer is amazingly compelling. If Apple is prepared for double the current rate (say 400,000), it is pretty reasonable to turn around and say 'jeez, we sold out!'.
Thanks for the correction, I should not rely on memory. 400,000 powermace would be a nice change.
So I think I'm hearing that IBM will be shipping 970's in March, Apple would be wise to stock-up during at least part of the ramp-up, which would put possible release dates around April or May??
Also IBM is ramping up for Apple and themselves..I mean it's not like Intel ramping up needing millions of these processors to fill supply...Apple just needs enough at first to supply the Powermac line and the Xserve which would only be in the 10s of thousands per month...and if Apple first introduced the chip in the Xserve then it would only need a few thousand chips to start off with...
Does anyone remember the schematics regarding the G4 in the Motorola pdf, where they showed that the max speed of the 7450 was 1ghz? According to this, and other peoples' claims, the 1.25ghz G4 was just an overclocked 1ghz. If this is true, how does this affect the likelihood of seeing a non updated (7450) G4 reaching 1.4ghz+?
<strong>BTW does anybody know exactly what is wrong with the Motorola fab? I would think that these yield problems would extend to their entire line of processors including the embedded ones and be very bad for business.
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.</strong><hr></blockquote>
Poor management -> lack of money -> the need to cut corners, such as shutting off fans in their fabs to cut down on (mammoth) electricity bills. Mot shut down all their older fabs (cost again) and farmed out production to contractors such as TSMC, so the products they make that don't have to be fabbed on the latest and greatest don't suffer from the state of Mot's own fabs. When Mot moves to .09 micron, they'll be able to use a fab built by STM in France, courtesy of a recent partnership.
Six Sigma had been abandoned, and was only recently reinstated, so I don't think that's the problem - except to the extent that in the current state of Mot's fabs it's all but impossible to meet that standard of quality. But then, Mot aren't using their own fabs for much of their own product line; before long, they won't even be using them for cutting-edge fabrication.
<strong>Does anyone remember the schematics regarding the G4 in the Motorola pdf, where they showed that the max speed of the 7450 was 1ghz?</strong><hr></blockquote>
If you mean the 745x family (Mot currently makes the 7455) they still advertize a top speed of 1GHz.
[quote]<strong>According to this, and other peoples' claims, the 1.25ghz G4 was just an overclocked 1ghz. If this is true, how does this affect the likelihood of seeing a non updated (7450) G4 reaching 1.4ghz+?</strong><hr></blockquote>
They're not overclocked AFAIK, they're just parts that beat the spec, available in enough numbers to supply Apple for their top of the line machine.
I can't imagine the 7455 reaching 1.4GHz. 25% over the design maximum is already unusual (at least, when it's in saleable quantities). There's a die-shrunk 7455 in the works (apparently called the 7457), and that will account for the next G4 speed increase.
They're not overclocked AFAIK, they're just parts that beat the spec, available in enough numbers to supply Apple for their top of the line machine.
</strong><hr></blockquote>
Motorola uses the prefixes "MPC" and "XPC" (among others) to denote a PPC's qualification status. MPC means the chip is fully qualified while XPC means that it is not. There was a pic on the web of the 1.25 taken after someone removed the heatsink. The prefix on the chip was XPC. IMO, this supports Amorph's statement that the current 1.25's are "freak" chips. I just wonder how many more Apple has stockpiled- they can't continue to sell them for the next 6 months with a finite quantity and they won't be able to move them down the product ladder for the same reason.
Maybe the 970 will be here by March and we will never see the '57s.
Comments
<strong>Wouldn't you consider it a design failure if it didn't out-produce a 604e?</strong><hr></blockquote>
"failure"? No -- it met its design goals to be a high performance 64-bit PowerPC. The 604 met its goal to be a high performance 32-bit PowerPC. Almost nobody at the time wanted a 64-bit PowerPC, so the 604/604e was a better solution (simpler, cheaper, faster). Remember, running 32-bit code is faster than running 64-bit code.
<strong>
Kurt and Brendon
No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
</strong><hr></blockquote>
If you are right (and I hope you are) I would guess that Pixar is swimming in them. I'm guessing that Pixar either is or soon will be the test bed for new Apple hardware. Steve would have much better control of leaks that way. I must have just mis-read your post. Also I believe that the Apple system chip is the main chip to be redesigned, but the memory system may require major rework to keep latency down, don't know much hear.
<strong>No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
But, alas, I probably just jinxed it
(BTW, what the heck does that [ ] smilie mean?)</strong><hr></blockquote>
If you're right (and I sure hope you are), I would expect the actual systems to be announced at MWNY, not earlier. It isn't only hardware here. These are not moto chips, as well as being 64 bit, and it's reasonable to expect a few nasty surprises when Apple starts testing them with a wide range of third party software. I would be delighted to see them actually ship by Septmber. I have rarely seen the optimistic predictions come true with the towers - the pessimists most always turn out to be right. I'm really looking forward to the hype for MWNY since it was almost nonexistent for MWSF.
BTW, I always thought that signified embarassment.
<strong>
More than likely Apple will wait for full production, I believe, and I'm guessing now that Apple is selling around 250,000 to 300,000 PowerMacs per quarter. I could see that number tripling easy, so around 1,000,000 per quarter. </strong><hr></blockquote>
176,000 powermacs last quarter, this quarter's results aren't out. That's up 10,000 over the previous quarter - which is a _very_ small jump due to the speed bumps in August. If Apple doubled that I think there'd be dancing in the streets. Just saying that a million powermacs a quarter is too much to aim for - they can't stock up THAT much inventory unless the computer is amazingly compelling. If Apple is prepared for double the current rate (say 400,000), it is pretty reasonable to turn around and say 'jeez, we sold out!'.
<a href="http://www.apple.com/pr/library/2002/oct/16earnings.html" target="_blank">Apple 4Q Report</a>
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.
<strong>
.
Kurt and Brendon
No, I didn't mean pre-production chips. I just believe that Apple has had enough samples to do the bulk of the design work, and the design in frozen now (acording to my sources) and Apple would definately get samples from the final design versions, so they may have them already. If the 970 really does enter production in March, then they could concievably have a few chips to slap 'n' ship by the last day of April. ('course there can always be problems in production forcing an additional design change. Hope not).
All I'm saying: I hear production starts in March, and I believe that Steve could possibly annouce and (only technically) ship some boxes before April is out.
But I could be wrong.
</strong><hr></blockquote>
If you have a source with real information than that is great. I certainly would like to see the 970 as soon as possible. My thoughts were pure guesses with the limited information I have seen on the web. I know enough about chip design (which isn' much) to know that designing the support chips and boards is no small task. Especially when you think that the bus architecture and speeds are changing significantly. This will certainly be no little speed bump and a whole new skill set for Apple engineers.
Just curious to anyone who knows chip production. If they are planning to release the chips in the second half, does that mean the chips are finalized? Or are there last minute revisions possible?
I used to think that the change in memory types in the last Powermac revision meant that the 970 wouldn't be ready for some time. There was not a huge improvement in the overall speed of the machines and it must have been a pretty big investment if the 970's were close. But seeing it in the new 17" Powerbook seems to indicate that that architecture is probably due to arrive in the next generation of iMacs and generally work its way down the line. That would make sense as to why they developed it.
You said the design was frozen. Any hints as to what it might look like? I.E. memory type, single/dual processors, new case, BENCHMARKS!!
Thanks.
<strong>BTW does anybody know exactly what is wrong with the Motorola fab? I would think that these yield problems would extend to their entire line of processors including the embedded ones and be very bad for business.
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.</strong><hr></blockquote>
Six sigma.
<strong>
Just curious to anyone who knows chip production. If they are planning to release the chips in the second half, does that mean the chips are finalized? Or are there last minute revisions possible?
</strong><hr></blockquote>
Very minor revisions are common (call "steppings" by some companies, I think). These are generally just bug fixes, which (in addition to outright incorrect function) include correcting problems that limit clock rate and yields.
Apple's work on a DDR chipset doesn't just get "thrown away". It will migrate to the low end, plus these things are usually designed in a modular fashion so the design & layout work can be probably be reused in a RapidIO or HT based chipset. Internally they could even design the thing in 2 halves which internally communicate using HyperTransport, so replacing one half or the other isn't too big a deal.
Apple's chipset includes several pieces of functionality:
- MPX bus interface
- AGP bus interface
- Memory controller
- I/O system interface
I split AGP out because it has higher bandwidth requirements that the rest of the I/O system so it might be appropriate to attach it directly to the memory controller. These pieces of functionality will likely need to be divided between a couple of chips in a 970 system, and some mechanism like HT or RIO introduced to communicate between them. This division may have already been done internally to the existing chipset.Also i read a while back on news.com way before talk of the 970 about IBM's megaherz goals...and I beleive I remember them mentioning being above 4 ghz around 2004 so I believe these puppies will scale well eventually unlike some other powerpc see makers
[ 01-11-2003: Message edited by: Producer ]</p>
<strong>
176,000 powermacs last quarter, this quarter's results aren't out. That's up 10,000 over the previous quarter - which is a _very_ small jump due to the speed bumps in August. If Apple doubled that I think there'd be dancing in the streets. Just saying that a million powermacs a quarter is too much to aim for - they can't stock up THAT much inventory unless the computer is amazingly compelling. If Apple is prepared for double the current rate (say 400,000), it is pretty reasonable to turn around and say 'jeez, we sold out!'.
<a href="http://www.apple.com/pr/library/2002/oct/16earnings.html" target="_blank">Apple 4Q Report</a></strong><hr></blockquote>
Thanks for the correction, I should not rely on memory. 400,000 powermace would be a nice change.
So I think I'm hearing that IBM will be shipping 970's in March, Apple would be wise to stock-up during at least part of the ramp-up, which would put possible release dates around April or May??
<strong>
Six sigma.</strong><hr></blockquote>
Ha! That's pretty funny. Motorola popularized this back in the 1980s.
I wonder if the dual gigger will be the bottom end?
Ati 9700 option? Geforce Fx?
I'd like to see deep price cuts. Couple of hundred for each model.
I can't see them selling the 'pre-970' model if they don't.
Lemon Bon Bon
<strong>BTW does anybody know exactly what is wrong with the Motorola fab? I would think that these yield problems would extend to their entire line of processors including the embedded ones and be very bad for business.
Is it lack of good talent? Is it management? Is it equipment? Money? I mean AMD (partly because of Mot) and Nvidia have their yield problems but this is crazy.</strong><hr></blockquote>
Poor management -> lack of money -> the need to cut corners, such as shutting off fans in their fabs to cut down on (mammoth) electricity bills. Mot shut down all their older fabs (cost again) and farmed out production to contractors such as TSMC, so the products they make that don't have to be fabbed on the latest and greatest don't suffer from the state of Mot's own fabs. When Mot moves to .09 micron, they'll be able to use a fab built by STM in France, courtesy of a recent partnership.
Six Sigma had been abandoned, and was only recently reinstated, so I don't think that's the problem - except to the extent that in the current state of Mot's fabs it's all but impossible to meet that standard of quality. But then, Mot aren't using their own fabs for much of their own product line; before long, they won't even be using them for cutting-edge fabrication.
[ 01-14-2003: Message edited by: Amorph ]</p>
<strong>Does anyone remember the schematics regarding the G4 in the Motorola pdf, where they showed that the max speed of the 7450 was 1ghz?</strong><hr></blockquote>
If you mean the 745x family (Mot currently makes the 7455) they still advertize a top speed of 1GHz.
[quote]<strong>According to this, and other peoples' claims, the 1.25ghz G4 was just an overclocked 1ghz. If this is true, how does this affect the likelihood of seeing a non updated (7450) G4 reaching 1.4ghz+?</strong><hr></blockquote>
They're not overclocked AFAIK, they're just parts that beat the spec, available in enough numbers to supply Apple for their top of the line machine.
I can't imagine the 7455 reaching 1.4GHz. 25% over the design maximum is already unusual (at least, when it's in saleable quantities). There's a die-shrunk 7455 in the works (apparently called the 7457), and that will account for the next G4 speed increase.
I hope it's targeted for the iMac revision.
Duals for the 'power'Macs.
Not that it matters to me...
Lemon Bon Bon
<strong>
They're not overclocked AFAIK, they're just parts that beat the spec, available in enough numbers to supply Apple for their top of the line machine.
</strong><hr></blockquote>
Motorola uses the prefixes "MPC" and "XPC" (among others) to denote a PPC's qualification status. MPC means the chip is fully qualified while XPC means that it is not. There was a pic on the web of the 1.25 taken after someone removed the heatsink. The prefix on the chip was XPC. IMO, this supports Amorph's statement that the current 1.25's are "freak" chips. I just wonder how many more Apple has stockpiled- they can't continue to sell them for the next 6 months with a finite quantity and they won't be able to move them down the product ladder for the same reason.
Maybe the 970 will be here by March and we will never see the '57s.