The processor they are announcing at the Microprocesssor Forum in October does have a vector unit and is 64 bits. This could be the confirmation that this vector unit is an Altivec unit (per IBM we already knew that it had the same number of instructions as Altivec - 160+)
Don't say that, you'll damage then thread's credibility
I suspect that G4 will be moved to the iBook, possibly stay on the iMac and eMac with a 166MHz FSB. All depends how quickly the next best hope of PPC arrives...
Far too much G4-slamming happens around here. The G4 is not a bad chip. Quite the contrary, actually. Moving the entire lineup to this new IBM monster PowerPC would be stupid because it would seriously compromise power consumption, heat, and cost. Apple's biggest problem right now is that it has a broad lineup of machines but only two processors to choose between -- one of those lacks AltiVec support, and the other isn't up to the task of taking the performance crown.
Having a selection of PowerPCs to choose from gives Apple engineering options, and we know they like to have options.
<strong>Far too much G4-slamming happens around here. The G4 is not a bad chip. Quite the contrary, actually.</strong><hr></blockquote>
Granted.
[quote]<strong>Moving the entire lineup to this new IBM monster PowerPC would be stupid because it would seriously compromise power consumption, heat, and cost. </strong><hr></blockquote>
How do you know? IBM will not be releasing technical details until Oct 15th. What's your source? For all you know, it's a kitten, not a monster.
[quote]<strong>Apple's biggest problem right now is that it has a broad lineup of machines but only two processors to choose between -- one of those lacks AltiVec support, and the other isn't up to the task of taking the performance crown.</strong><hr></blockquote>
No, actually, Apple's biggest problem is the economy. Apple's second biggest problem is staying competitive in the PC market.
[ 09-02-2002: Message edited by: Tomb of the Unknown ]</p>
Programmer, do you know something specific we don't, or has it just been a bad weekend? Outright slams aren't normally your style.
I do agree putting the entire lineup wouldn't be smart, but for different reasons. There needs to be a sound capability / cost continuum from entry level to high end and having the high-end family across the line blurs that distinction too far. Much as we have had G3 -> G4 lineups for the last several years.</strong><hr></blockquote>
Look at the description of the thing: 8-way superscalar based on the POWER4, supporting SMP, SIMD, high speed bus, 64-bit ISA. IBM doesn't have some magical ability to make each transistor do more -- to build a processor that is this capable they need to use a large number of transistors. This is going to be at least as complex as the current Pentium4, and probably more so. Their stated target is "desktops and low-end servers", and we aren't going to see it until at least 2003 (probably late in the year). If portables was an intended target they would have said so. And this is IBM so "low-end server" is still going to represent some serious iron. Processor design is a slow and expensive project so they will choose a manufacturing process that will be available (and productive) at the time they finish the design, and they will aim to put as many transistors on that process as they can and still have an acceptable desktop power & heat budget. The stated 6.4 GB/sec system interface is going to have some interesting board-level implications. All of this means that putting this chip into a portable machine is highly unlikely (at least not in a portable with the level of refinement that Apple delivers). Successive generations may go there, but its going to be a long time before the iBook goes this route.
The 7455 is an excellent chip and delivers terrific performance... using about 30 million transistors, IIRC. Hopefully Motorola moves it to a smaller process and adopts RapidIO in the G4 line so that Apple can continue using it in machines where a 60+ million transistor powerhouse just isn't appropriate (which includes the iMac with its carefully designed small enclosure).
BTW: when I said "Apple's biggest problem" I was only referring specifically to their ongoing performance problems.
I agree that there's a no chance of this now legendary IBM processor making it into the whole Mac lienup but if it is put into the PowerMac, then lower end machines (iMac, eMac and portables) can be moved to faster G4s with 166MHz+ or RIO FSB, etc... New high end chip => current high end can be moved to other machines.
[quote]Far too much G4-slamming happens around here. The G4 is not a bad chip. Quite the contrary, actually.<hr></blockquote>
The G4 may be cool, provide good performance per clock cycle for its transistor count and well designed but the MHz myth can't be stretched indefinitely. I guess folks are just (overly) eagerly awaiting a CPU that everyone agrees sounds high end.
<strong>The G4 may be cool, provide good performance per clock cycle for its transistor count and well designed but the MHz myth can't be stretched indefinitely. I guess folks are just (overly) eagerly awaiting a CPU that everyone agrees sounds high end.</strong><hr></blockquote>
My point is that it doesn't need to be. Not every machine has to have the highest performance. I'm not saying that the iBook, iMac, and PowerBook won't move up the performance "food chain" -- just that they won't suddenly leap to the top of it.
The "MHz myth" is going to get even more significant with this new PowerPC chip. Intel is going to be up to 3 or 4 GHz and it sounds like IBM is aiming for about 2 GHz. The efficiency per clock cycle will be much better, however, which means that the importance of the "MHz myth" will be even greater. Get used to it. The concept is valid and correct, and MHz are only going to become more meaningless in determining processor performance as time goes on with the coming of multi-core, multi-threaded, SIMD, and clockless designs. Benchmarks need to catch up with reality at some point.
If Motorola really can come up with a 90nm process by the time they say and they build a G4 on it, then you will see a 4GHz G4 chip in the not too distant future.
It would seem mad for them to skip a step (.13micron), but depending on the process changes between .18µm and .09µm shrinking it by 2 may amount to no more work than shrinking it by sqrt(2).
Jumping to a .09µ micron process would cut the power consumption to 1/4 of what it is. So a 4GHz G4 on a .09µ process would consume as much as a 1GHz .18µ G4 does today.
[quote]I'm not saying that the iBook, iMac, and PowerBook won't move up the performance "food chain" -- just that they won't suddenly leap to the top of it.<hr></blockquote>
We agree then .
MHz myth: I meant that Apple couldn't keep stretching the MHz myth for the same processor.
High Hz numbers sell processors, even if this is not the whole story w.r.t. real world performance . Hopefully a better understanding of CPU performance will arise in the computer buying public in future.
[quote]So a 4GHz G4 on a .09µ process would consume as much as a 1GHz .18µ G4 does today.<hr></blockquote>
Does power consumption scales linearly with transistor size and is heat is the only factor in scaling the G4? A 0.09u G4 would certainly be interesting.
Does power consumption scales linearly with transistor size and is heat is the only factor in scaling the G4? A 0.09u G4 would certainly be interesting.</strong><hr></blockquote>
Power consumption scales with the square of the process size, this is why processes get shrunk by a factor of sqrt(2) with each generation (cutting the power consumption by 1/2). Power consumption also scales with frequency, but linearly (double the frequency = double the power consumption).
There is a lot more than just power involved, but what I described is good rule of thumb. Process rules are very complicated, they have to do with what material can be within what distance of another material (has to do with the precision of micro photo lithography, and the layering of the chip).
<strong>IBM doesn't have some magical ability to make each transistor do more</strong><hr></blockquote>
Uhmmm, actually they do. IBM's research labs have announced several advances that are capable of making each transistor "do more" -- or more accurately, do the same with less power.
[quote]<strong>to build a processor that is this capable they need to use a large number of transistors. This is going to be at least as complex as the current Pentium4, and probably more so.</strong><hr></blockquote>
Nonsense. The ISA is why the P4 is as complex as it is, not the capabiliy of the chip. Intel engineers made several tradeoffs for clock speed, power consumption being just one of them. There is no reason to suppose that IBM had to make the same compromises in their design. For all you know, this chip runs cooler or as cool as the MPC7455. (OK, probably not, but the point is that we don't know and won't know until IBM releases it's thermal profile.)
[quote]<strong>If portables was an intended target they would have said so.</strong><hr></blockquote>
Oh, I see, you just want a faster iBook. Well, you'll just have to wait for .09 G4 chips then. And what's wrong with that?
[quote]<strong>Processor design is a slow and expensive project so they will choose a manufacturing process that will be available (and productive) at the time they finish the design, and they will aim to put as many transistors on that process as they can and still have an acceptable desktop power & heat budget.</strong><hr></blockquote>
Again, so the initial silicon will not be for mobile solutions. If you'll recall, neither was the first G4.
[quote]<strong>The stated 6.4 GB/sec system interface is going to have some interesting board-level implications. All of this means that putting this chip into a portable machine is highly unlikely (at least not in a portable with the level of refinement that Apple delivers). Successive generations may go there, but its going to be a long time before the iBook goes this route.</strong><hr></blockquote>
I'd say that's a reasonable assumption, but we won't really know until we see the specs on this chip. It's not impossible that it's thermal characteristics would allow it to be used in a laptop that employs some kind of speedstepping, for instance. And yes, if Apple thought there was money to be made there you can bet they're capable of delivering. I'm just not sure that current demand -- which currently consists a market of one as far as I can tell -- is sufficient to make the effort economically viable.
[quote]<strong>The 7455 is an excellent chip and delivers terrific performance... using about 30 million transistors, IIRC. Hopefully Motorola moves it to a smaller process and adopts RapidIO in the G4 line so that Apple can continue using it in machines where a 60+ million transistor powerhouse just isn't appropriate (which includes the iMac with its carefully designed small enclosure).</strong><hr></blockquote>
Stranger things have been known to happen. But it's pointless to speculate about transistor densities in this new chip when we don't even know what process size it will use or the dimensions of the chip itself. Right now, we should just chill. October will be here soon enough.
[ 09-03-2002: Message edited by: Tomb of the Unknown ]</p>
<strong>I suspect that G4 will be moved to the iBook, possibly stay on the iMac and eMac with a 166MHz FSB. All depends how quickly the next best hope of PPC arrives...</strong><hr></blockquote>
That would make sense, keep both companies involved and interested in R&D like they are currently (but reversing roles between the two). I'd agree the Powerbook would probably take some time to switch over (the chip and the 'book will both need to adapt, most likely. Apple likes to have Options....
Comments
<strong>RS/6000 machines don't have AltiVec.</strong><hr></blockquote>
Hawkeye (can I call you Hawkeye)
Very good call !
Source? A guy who knows a guys who knows a guy at IBM. Kinda thin, yeah, but, well, this is AI.
edit: almost forgot: ***CONFIRMED!!!***
[ 09-02-2002: Message edited by: spotbug ]</p>
I suspect that G4 will be moved to the iBook, possibly stay on the iMac and eMac with a 166MHz FSB. All depends how quickly the next best hope of PPC arrives...
Having a selection of PowerPCs to choose from gives Apple engineering options, and we know they like to have options.
<strong>Far too much G4-slamming happens around here. The G4 is not a bad chip. Quite the contrary, actually.</strong><hr></blockquote>
Granted.
[quote]<strong>Moving the entire lineup to this new IBM monster PowerPC would be stupid because it would seriously compromise power consumption, heat, and cost. </strong><hr></blockquote>
How do you know? IBM will not be releasing technical details until Oct 15th. What's your source? For all you know, it's a kitten, not a monster.
[quote]<strong>Apple's biggest problem right now is that it has a broad lineup of machines but only two processors to choose between -- one of those lacks AltiVec support, and the other isn't up to the task of taking the performance crown.</strong><hr></blockquote>
No, actually, Apple's biggest problem is the economy. Apple's second biggest problem is staying competitive in the PC market.
[ 09-02-2002: Message edited by: Tomb of the Unknown ]</p>
<strong>
Programmer, do you know something specific we don't, or has it just been a bad weekend? Outright slams aren't normally your style.
I do agree putting the entire lineup wouldn't be smart, but for different reasons. There needs to be a sound capability / cost continuum from entry level to high end and having the high-end family across the line blurs that distinction too far. Much as we have had G3 -> G4 lineups for the last several years.</strong><hr></blockquote>
Look at the description of the thing: 8-way superscalar based on the POWER4, supporting SMP, SIMD, high speed bus, 64-bit ISA. IBM doesn't have some magical ability to make each transistor do more -- to build a processor that is this capable they need to use a large number of transistors. This is going to be at least as complex as the current Pentium4, and probably more so. Their stated target is "desktops and low-end servers", and we aren't going to see it until at least 2003 (probably late in the year). If portables was an intended target they would have said so. And this is IBM so "low-end server" is still going to represent some serious iron. Processor design is a slow and expensive project so they will choose a manufacturing process that will be available (and productive) at the time they finish the design, and they will aim to put as many transistors on that process as they can and still have an acceptable desktop power & heat budget. The stated 6.4 GB/sec system interface is going to have some interesting board-level implications. All of this means that putting this chip into a portable machine is highly unlikely (at least not in a portable with the level of refinement that Apple delivers). Successive generations may go there, but its going to be a long time before the iBook goes this route.
The 7455 is an excellent chip and delivers terrific performance... using about 30 million transistors, IIRC. Hopefully Motorola moves it to a smaller process and adopts RapidIO in the G4 line so that Apple can continue using it in machines where a 60+ million transistor powerhouse just isn't appropriate (which includes the iMac with its carefully designed small enclosure).
BTW: when I said "Apple's biggest problem" I was only referring specifically to their ongoing performance problems.
[quote]Far too much G4-slamming happens around here. The G4 is not a bad chip. Quite the contrary, actually.<hr></blockquote>
The G4 may be cool, provide good performance per clock cycle for its transistor count and well designed but the MHz myth can't be stretched indefinitely. I guess folks are just (overly) eagerly awaiting a CPU that everyone agrees sounds high end.
<strong>The G4 may be cool, provide good performance per clock cycle for its transistor count and well designed but the MHz myth can't be stretched indefinitely. I guess folks are just (overly) eagerly awaiting a CPU that everyone agrees sounds high end.</strong><hr></blockquote>
My point is that it doesn't need to be. Not every machine has to have the highest performance. I'm not saying that the iBook, iMac, and PowerBook won't move up the performance "food chain" -- just that they won't suddenly leap to the top of it.
The "MHz myth" is going to get even more significant with this new PowerPC chip. Intel is going to be up to 3 or 4 GHz and it sounds like IBM is aiming for about 2 GHz. The efficiency per clock cycle will be much better, however, which means that the importance of the "MHz myth" will be even greater. Get used to it. The concept is valid and correct, and MHz are only going to become more meaningless in determining processor performance as time goes on with the coming of multi-core, multi-threaded, SIMD, and clockless designs. Benchmarks need to catch up with reality at some point.
It would seem mad for them to skip a step (.13micron), but depending on the process changes between .18µm and .09µm shrinking it by 2 may amount to no more work than shrinking it by sqrt(2).
Jumping to a .09µ micron process would cut the power consumption to 1/4 of what it is. So a 4GHz G4 on a .09µ process would consume as much as a 1GHz .18µ G4 does today.
So the G4 may have a life after all.
We agree then
MHz myth: I meant that Apple couldn't keep stretching the MHz myth for the same processor.
High Hz numbers sell processors, even if this is not the whole story w.r.t. real world performance . Hopefully a better understanding of CPU performance will arise in the computer buying public in future.
[quote]So a 4GHz G4 on a .09µ process would consume as much as a 1GHz .18µ G4 does today.<hr></blockquote>
Does power consumption scales linearly with transistor size and is heat is the only factor in scaling the G4? A 0.09u G4 would certainly be interesting.
/waits for G4 iBook (Expo Paris unlikely
<strong>
Does power consumption scales linearly with transistor size and is heat is the only factor in scaling the G4? A 0.09u G4 would certainly be interesting.</strong><hr></blockquote>
Power consumption scales with the square of the process size, this is why processes get shrunk by a factor of sqrt(2) with each generation (cutting the power consumption by 1/2). Power consumption also scales with frequency, but linearly (double the frequency = double the power consumption).
There is a lot more than just power involved, but what I described is good rule of thumb. Process rules are very complicated, they have to do with what material can be within what distance of another material (has to do with the precision of micro photo lithography, and the layering of the chip).
<strong>IBM doesn't have some magical ability to make each transistor do more</strong><hr></blockquote>
Uhmmm, actually they do. IBM's research labs have announced several advances that are capable of making each transistor "do more" -- or more accurately, do the same with less power.
[quote]<strong>to build a processor that is this capable they need to use a large number of transistors. This is going to be at least as complex as the current Pentium4, and probably more so.</strong><hr></blockquote>
Nonsense. The ISA is why the P4 is as complex as it is, not the capabiliy of the chip. Intel engineers made several tradeoffs for clock speed, power consumption being just one of them. There is no reason to suppose that IBM had to make the same compromises in their design. For all you know, this chip runs cooler or as cool as the MPC7455. (OK, probably not, but the point is that we don't know and won't know until IBM releases it's thermal profile.)
[quote]<strong>If portables was an intended target they would have said so.</strong><hr></blockquote>
Oh, I see, you just want a faster iBook. Well, you'll just have to wait for .09 G4 chips then. And what's wrong with that?
[quote]<strong>Processor design is a slow and expensive project so they will choose a manufacturing process that will be available (and productive) at the time they finish the design, and they will aim to put as many transistors on that process as they can and still have an acceptable desktop power & heat budget.</strong><hr></blockquote>
Again, so the initial silicon will not be for mobile solutions. If you'll recall, neither was the first G4.
[quote]<strong>The stated 6.4 GB/sec system interface is going to have some interesting board-level implications. All of this means that putting this chip into a portable machine is highly unlikely (at least not in a portable with the level of refinement that Apple delivers). Successive generations may go there, but its going to be a long time before the iBook goes this route.</strong><hr></blockquote>
I'd say that's a reasonable assumption, but we won't really know until we see the specs on this chip. It's not impossible that it's thermal characteristics would allow it to be used in a laptop that employs some kind of speedstepping, for instance. And yes, if Apple thought there was money to be made there you can bet they're capable of delivering. I'm just not sure that current demand -- which currently consists a market of one as far as I can tell -- is sufficient to make the effort economically viable.
[quote]<strong>The 7455 is an excellent chip and delivers terrific performance... using about 30 million transistors, IIRC. Hopefully Motorola moves it to a smaller process and adopts RapidIO in the G4 line so that Apple can continue using it in machines where a 60+ million transistor powerhouse just isn't appropriate (which includes the iMac with its carefully designed small enclosure).</strong><hr></blockquote>
Stranger things have been known to happen. But it's pointless to speculate about transistor densities in this new chip when we don't even know what process size it will use or the dimensions of the chip itself. Right now, we should just chill. October will be here soon enough.
[ 09-03-2002: Message edited by: Tomb of the Unknown ]</p>
<strong>I suspect that G4 will be moved to the iBook, possibly stay on the iMac and eMac with a 166MHz FSB. All depends how quickly the next best hope of PPC arrives...</strong><hr></blockquote>
That would make sense, keep both companies involved and interested in R&D like they are currently (but reversing roles between the two). I'd agree the Powerbook would probably take some time to switch over (the chip and the 'book will both need to adapt, most likely. Apple likes to have Options....