Two things can explain Mr. Joswiak's comments about the G4 from Motorola. First, Apple does not want potential buyer to think the Motorola G4 is dead and hold off buying a Mac because of fears that it will soon be obsolete. It will take a while for the transition to the G5 and during that time Apple wants to sell more than just G5 Power Macs. Second, Apple still needs Motorola to be a good, reliable partner for another year or so.
Just a comment. I think the G4 will be with us for many years, but IBM will likely be the supplier. It will have a better FPU and bus than the present G4. It will be for lower end, consumer Macs, and would not need to have SMP, but may have it anyway. Mojave?
Joswiak has pretty much informed us that Motorola isn't going to make any significant gains by stating only IBM can take Apple to the next level. This means to me that there won't be any FSB replacement for the G4, at least from Motorola. Certainly the PB will have to go to the G5, but that leaves at least the iMac in G4 land. Perhaps the IBM's G3 train can be upgraded with Altivec and a modern bus, thus creating a G4.5. I hope Apple realizes that customers won't tolerate crippled consumer machines for too long a period of time.
Joswiak has pretty much informed us that Motorola isn't going to make any significant gains by stating only IBM can take Apple to the next level. This means to me that there won't be any FSB replacement for the G4, at least from Motorola. Certainly the PB will have to go to the G5, but that leaves at least the iMac in G4 land. Perhaps the IBM's G3 train can be upgraded with Altivec and a modern bus, thus creating a G4.5. I hope Apple realizes that customers won't tolerate crippled consumer machines for too long a period of time.
Yes, quite correct. Moto isn't going anywhere. They will continue to ship the G4 for the next ten years. They will never update their FSB. This company is DEFINITELY not going anywhere.
To get the G4 to compete with the Pentium 4 and the IBM 970 the G4 has to get past 3 GHz and have an improved buss design. How likely is that to happen
A die shrink might boost the clock speed by 25-40%. And CPUs in the 1.5 to 2 GHz range is a usefull as a stop gap measure until the 90nm IBM 970 can be put in all macintoshes.
The writing is on the wall, the IBM 970 is about to take over and the G4 is on the way out fast
To get the G4 to compete with the Pentium 4 and the IBM 970 the G4 has to get past 3 GHz and have an improved buss design. How likely is that to happen
The G4 would not compete with the IBM 970. For consumer Macs it need only compete with run of the mill P4s that go into consumer Windows PCs. Surely it could do that at 2 GHz with a fast bus and reasonable FPU. The Mojave may be just such a chip.
My prediction would be a G5 for the iMac and PowerBooks, and Mojave for the iBook and one or more new Macs to meet the needs of general office use, and simple home computing needs.
Just a note, since I mention a possible new Mojave Mac, there may be a new G5 Mac, like a new cube or the Mini G5 being discussed elsewhere.
Expect the IBM750VX Q1 next year, scaling up to 2GHz with the 100nm process, a 450MHz FSB, 1024KB of L2 cache, SMP support, and Altivec.
Looks better than what Moto' is offering anytime soon.
With .09 970s putting Apple really back in the game come 04 then I don't see why G3/G4s are really needed.
Die Shrink 970s should be stuffed into everything.
When Apple are giving Moto' plaudits like 'huge' then that is double talk...'Yer on yer way out...and not until we sue your ass off.' The G4 is so overwhelmingly behind...it's on borrowed time. Even a 2 gig G3+ with 450 bus. Dunno. Cooler lower-end clocked 0.09 970s all the way?
I think it's going to be well past MWSF before we see a 970 based PB.
Next up is the Xserve and an additional dual in the PowerMac line. Remember, the G5 is "optimized for dual processor applications." Then possibly a "small" tower/cube-redux to replace the PMG4 at the 1299 price point.
Then a 17" PB
followed by a 15" sometime later, unless both come together, but later
You're looking at 12 months, and mebbe 16-18. There's going to be bitching at MWNY, MWSF, and MWNY'04 (if there's one at all, if not then there'll be some whining at whatever replaces it, or following it, you know what I mean.)
. . . With .09 970s putting Apple really back in the game come 04 then I don't see why G3/G4s are really needed.
Die Shrink 970s should be stuffed into everything. . .
I must disagree. It is inconceivable that Apple would use the same processor for their top workstation and their lowest priced consumer Macs. There needs to be at least two processors. The high performance G5 will always cost more and run hotter than what the low end market needs. Sure, the 970+ will be a die shrink, but do you suppose IBM will leave the design unchanged? This is the top end processor, and IBM will figure out a way to boost performance, and make the die larger. Many are betting that the L2 cache goes up to at least 1 MB, maybe 2 MB.
I have to think that even Apple would prefer a G4/5 split in the consumer/pro machines, for marketing reasons, safety net reasons (you NEVER know) and cost reasons. IBM wants to make killer speed demons, and it looks like they're good at it. Moto wants to make efficient econo-chips, and if left to it, they're good at that. So you get two supply streams that match your needs (consumer pro) and your suppliers' capabilities (performance vs embedded)
I think Moto will surprise a lot of people with a pretty quick move to .09u.
I think Moto will surprise a lot of people with a pretty quick move to .09u.
I think that was the plan all along. Actually, I remember reading that they wanted to jump from 180 to 90, skipping 150 & 130 altogether.
The G4/G5 split will really help Apple in the long run, assuming IBM doesn't muck things up (and I have no reason to believe they will.) The consumer and pro lines will finally be able to grow at their own paces, depending on the market and margins, rather than supply.
. . . I think Moto will surprise a lot of people with a pretty quick move to .09u.
But will .09u give us a fast bus and a better FPU? The consumer Macs must compete with Intel too.
I believe IBM will proceed very well with the 970 family. IBM has a vested interest in these wickedly powerful chips for lower end servers. Dell servers are a threat to IBM's core business, but this family of processors lets them build servers that beat Dell in the performance game. Otherwise, using x86 processors IBM could only have 'me too' products that perform no better than Dells. Price would continue to erode IBM's server market.
The question is whether IBM will provide a G4 replacement with a fast bus and better FPU? Will Mojave be the future consumer PPC processor? I have absolutely no faith in Motorola suppling a G4 that compete well against Intel consumer processors.
Say that "run of the mill" is 2/3 of these scores.
A 2 GHz P4 very much run of the mill.
A 1.2 GHz IBM 970, a newcomer but hardly leading edge...
A Motorola G4 in the 2 to 4 GHz range, "run of the mill" Riight! Then what is the current 1.42
My hope is that we during the following 12 months migrate to
Servers dual 970
Tower all dual 970
imac/eMac single 970
PB single 970
iBook 2 GHz G3
What ever G4 there is or will be aviable can be used as a drop in speed bump on the current G4 computers on their way to 970.
This is unlike all other CPU transitions
G4(2x1.42)->IBM970(2x2.0) offers way better performance right of the bat
G3(450)->G4(450) offered no performance benefits outside AV and SMP that anyhow was not supportet by either OS or apps
604E(350)->G3(266) was a marginal speed boost
601(120)->604(150) was a marginal speed boost
68040 (40)->601(80) was in practice a marginal speed boost due to all the emulation.
Apple have been behind since late 1999, they have just now a vastly better CPU aviable why would they waste time with the G4, a CPU were the current top of the line offer one third the integer performance of a current P4 and even worse fpu performance.
I have to think that even Apple would prefer a G4/5 split in the consumer/pro machines, for marketing reasons, safety net reasons (you NEVER know) and cost reasons. IBM wants to make killer speed demons, and it looks like they're good at it. Moto wants to make efficient econo-chips, and if left to it, they're good at that. So you get two supply streams that match your needs (consumer pro) and your suppliers' capabilities (performance vs embedded)
I think Moto will surprise a lot of people with a pretty quick move to .09u.
The problem that I see with a G4/G5 split, at least in the long run is that Moto was planning on Rapid IO for the G4 if I remember and the G5 (aka 970) uses Hyper transport. That would require Apple to design and maintain 2 chip sets. Right now the G4 is using the same basic (or at least an one that evolved over time, spreading out development cost) motherboard/companion chip/etc that the G4 (G3?) was introduced with. This has helped Apple keep the development of their consumer computers down since they just "adopted" the high end specs as the PM's grew. Moving to a unique archetecture (RM) for the G4 in the consumer line would break this "hand me down" technology advantage that they currently enjoy, and increase development cost.
A few months ago, many were saying the G5 is dead. The 970 PowerMac is the future. The G5 name was associated with Motorola, and some folks wanted to get rid of it. Yet here we are today with a G5 Power Mac that uses the IBM 970, and no one appears to be complaining.
So why does there seem to be this same aversion to the G4 name? There is a lot of misunderstanding going on, because for many the G4 name means Motorola, a slow bus and a history of performance problems. The G3 name on the other hand still seems to have good vibrations.
Maybe, only when we get a G4 iBook that uses a 2 GHz 750 VX Mojave processor that the cloud will lift from the G4 name. It will be a G4 because it will have the Velocity Engine and run applications that require a G4.
A point .13u G4 would stretch rom 1-1.8Ghz, it's a fairly safe bet that a .09u chip would run from 1.5-2.5Ghz with not too much rouble and more than equal run of the mill X86.
I can hardly believe I'm sying this, but... How much performance does an i/eMac/iBook need? I think a far more critical issue is to get the prices of these things DOWN, way DOWN, to a level where they nearly sell themselves. Numbers I've heard about the potential power-drain of a .09u G4 make it sound like a PERFECT laptop/AIO chip. Perfect, low power, cheap, and fast. It took a long time to get here, but once it does, there's no real reason NOt to use it. Will Winte;lon have a 64bit "consumer" chip? Not likely, nor for laptops. Lets assume that it uses a RapidI/O bus (there's the real surprise) while it still won't be as fast as PC970, it will give at least DDR266 level throughput, stretching to DDR400. I doubt you'll see laptops with that kind of performance, real laptops, not mobile desktops.
Doe Apple need a cheaper tower? yes, and it should be G5 based, kind of a pro's iMac, if you follow, small and unobtrusive, but with the same display flexibility as a tower and similar Gfx/CPU upgradability. A low end tower. This space is nearly decently covered now by the PMG4. I expect it to move to a G5 when supplies and costs allow, and shrink to sit between an iMac and a PMG5.
The Consumer machines just have to get cheaper, Apple needs to take advantage of plummeting Drive/LCD/RAM prices to make iMacs as cheap as possible, and not add the new expense of a new architecture just when the possibility to make a real impact has FINALLY presented itself. With G4 based drop in replacements at .13u and .09u getting cheaper, cooler, and faster, now is the time to strike at the heart of the issue, PRICE, and not chase some vain geek dream of supercomputer performance in an AIO/consumer notebook.
About the chipsets, I believe they are both point to point, and make it easier to overcome difficulties with compatibility once a controller chip is in place. Everywhere the would interface with standard I/O's USB, DDR, AGP, FW, Ethernet. It would most likely be transparent to the OS.
In one sense of the I agree with your statement about iMac/eMac performance. However right now it is lagging way behind, and needs to catch up. I also think that there is going to be a trend for more applications running concurrently as the digital hub model takes off.
My understanding of the archetecture is not that of a hardware engineer's, but what I do know of it seams to suggest that MP systems excell at the more parellel demands on the computer. I think that once the mhz gets up to a more respectable level, and the system bus inadequicies are brought up to speed, that the logical step would to move to MP systems. However the processors need to be fast enough so that the user "percieves" it to be on par or better than the offerings in the window world to sell these systems to switchers. I dont think that we are there yet with the G4's current speeds. Maybe the 57' will get us there with the added benefit of lowering the cost of the systems overall. Moto's Performance over that last 4 years has me skeptical on this though.
I think Moto will surprise a lot of people with a pretty quick move to .09u.
It doesn't matter how quick they get the G4 going because without a new FSB, it is still data starved. This becomes even worse as the G4 gets faster. A 0.9nm G4 with MaxBus is really pathetic. Moto needs to also come out with a FSB for the G4.
Comments
Seems that IBM-970 SOI helps Moto to switch faster:
http://www.theregister.co.uk/content/39/31424.html
Rooster
Just a comment. I think the G4 will be with us for many years, but IBM will likely be the supplier. It will have a better FPU and bus than the present G4. It will be for lower end, consumer Macs, and would not need to have SMP, but may have it anyway. Mojave?
Originally posted by Big Mac
Joswiak has pretty much informed us that Motorola isn't going to make any significant gains by stating only IBM can take Apple to the next level. This means to me that there won't be any FSB replacement for the G4, at least from Motorola. Certainly the PB will have to go to the G5, but that leaves at least the iMac in G4 land. Perhaps the IBM's G3 train can be upgraded with Altivec and a modern bus, thus creating a G4.5. I hope Apple realizes that customers won't tolerate crippled consumer machines for too long a period of time.
http://www.theregister.co.uk/content/39/30640.html
Expect the IBM750VX Q1 next year, scaling up to 2GHz with the 100nm process, a 450MHz FSB, 1024KB of L2 cache, SMP support, and Altivec.
A die shrink might boost the clock speed by 25-40%. And CPUs in the 1.5 to 2 GHz range is a usefull as a stop gap measure until the 90nm IBM 970 can be put in all macintoshes.
The writing is on the wall, the IBM 970 is about to take over and the G4 is on the way out fast
Originally posted by DrBoar
To get the G4 to compete with the Pentium 4 and the IBM 970 the G4 has to get past 3 GHz and have an improved buss design. How likely is that to happen
The G4 would not compete with the IBM 970. For consumer Macs it need only compete with run of the mill P4s that go into consumer Windows PCs. Surely it could do that at 2 GHz with a fast bus and reasonable FPU. The Mojave may be just such a chip.
My prediction would be a G5 for the iMac and PowerBooks, and Mojave for the iBook and one or more new Macs to meet the needs of general office use, and simple home computing needs.
Just a note, since I mention a possible new Mojave Mac, there may be a new G5 Mac, like a new cube or the Mini G5 being discussed elsewhere.
Expect the IBM750VX Q1 next year, scaling up to 2GHz with the 100nm process, a 450MHz FSB, 1024KB of L2 cache, SMP support, and Altivec.
Looks better than what Moto' is offering anytime soon.
With .09 970s putting Apple really back in the game come 04 then I don't see why G3/G4s are really needed.
Die Shrink 970s should be stuffed into everything.
When Apple are giving Moto' plaudits like 'huge' then that is double talk...'Yer on yer way out...and not until we sue your ass off.' The G4 is so overwhelmingly behind...it's on borrowed time. Even a 2 gig G3+ with 450 bus. Dunno. Cooler lower-end clocked 0.09 970s all the way?
I guess we'll know by the end of the year...
Lemon Bon Bon
Next up is the Xserve and an additional dual in the PowerMac line. Remember, the G5 is "optimized for dual processor applications." Then possibly a "small" tower/cube-redux to replace the PMG4 at the 1299 price point.
Then a 17" PB
followed by a 15" sometime later, unless both come together, but later
You're looking at 12 months, and mebbe 16-18. There's going to be bitching at MWNY, MWSF, and MWNY'04 (if there's one at all, if not then there'll be some whining at whatever replaces it, or following it, you know what I mean.)
12 months, at least, deal with it.
Originally posted by Lemon Bon Bon
. . . With .09 970s putting Apple really back in the game come 04 then I don't see why G3/G4s are really needed.
Die Shrink 970s should be stuffed into everything. . .
I must disagree. It is inconceivable that Apple would use the same processor for their top workstation and their lowest priced consumer Macs. There needs to be at least two processors. The high performance G5 will always cost more and run hotter than what the low end market needs. Sure, the 970+ will be a die shrink, but do you suppose IBM will leave the design unchanged? This is the top end processor, and IBM will figure out a way to boost performance, and make the die larger. Many are betting that the L2 cache goes up to at least 1 MB, maybe 2 MB.
I think Moto will surprise a lot of people with a pretty quick move to .09u.
Originally posted by Matsu
I think Moto will surprise a lot of people with a pretty quick move to .09u.
I think that was the plan all along. Actually, I remember reading that they wanted to jump from 180 to 90, skipping 150 & 130 altogether.
The G4/G5 split will really help Apple in the long run, assuming IBM doesn't muck things up (and I have no reason to believe they will.) The consumer and pro lines will finally be able to grow at their own paces, depending on the market and margins, rather than supply.
Originally posted by Matsu
. . . I think Moto will surprise a lot of people with a pretty quick move to .09u.
But will .09u give us a fast bus and a better FPU? The consumer Macs must compete with Intel too.
I believe IBM will proceed very well with the 970 family. IBM has a vested interest in these wickedly powerful chips for lower end servers. Dell servers are a threat to IBM's core business, but this family of processors lets them build servers that beat Dell in the performance game. Otherwise, using x86 processors IBM could only have 'me too' products that perform no better than Dells. Price would continue to erode IBM's server market.
The question is whether IBM will provide a G4 replacement with a fast bus and better FPU? Will Mojave be the future consumer PPC processor? I have absolutely no faith in Motorola suppling a G4 that compete well against Intel consumer processors.
The G4 would not compete with the IBM 970. For consumer Macs it need only compete with run of the mill P4s that go into consumer Windows PCs.
Pentium IV @ 3.06 GHz:
SPECint2000 1032 @ 3.06 [email protected]=Motorola G4@ 3.5GHz
SPECfp2000 1092 @ 3.06 [email protected]=Motorola G4@ 6.3GHz
Say that "run of the mill" is 2/3 of these scores.
A 2 GHz P4 very much run of the mill.
A 1.2 GHz IBM 970, a newcomer but hardly leading edge...
A Motorola G4 in the 2 to 4 GHz range, "run of the mill" Riight! Then what is the current 1.42
My hope is that we during the following 12 months migrate to
Servers dual 970
Tower all dual 970
imac/eMac single 970
PB single 970
iBook 2 GHz G3
What ever G4 there is or will be aviable can be used as a drop in speed bump on the current G4 computers on their way to 970.
This is unlike all other CPU transitions
G4(2x1.42)->IBM970(2x2.0) offers way better performance right of the bat
G3(450)->G4(450) offered no performance benefits outside AV and SMP that anyhow was not supportet by either OS or apps
604E(350)->G3(266) was a marginal speed boost
601(120)->604(150) was a marginal speed boost
68040 (40)->601(80) was in practice a marginal speed boost due to all the emulation.
Apple have been behind since late 1999, they have just now a vastly better CPU aviable why would they waste time with the G4, a CPU were the current top of the line offer one third the integer performance of a current P4 and even worse fpu performance.
Nothing to see, time to move on
Originally posted by Matsu
I have to think that even Apple would prefer a G4/5 split in the consumer/pro machines, for marketing reasons, safety net reasons (you NEVER know) and cost reasons. IBM wants to make killer speed demons, and it looks like they're good at it. Moto wants to make efficient econo-chips, and if left to it, they're good at that. So you get two supply streams that match your needs (consumer pro) and your suppliers' capabilities (performance vs embedded)
I think Moto will surprise a lot of people with a pretty quick move to .09u.
The problem that I see with a G4/G5 split, at least in the long run is that Moto was planning on Rapid IO for the G4 if I remember and the G5 (aka 970) uses Hyper transport. That would require Apple to design and maintain 2 chip sets. Right now the G4 is using the same basic (or at least an one that evolved over time, spreading out development cost) motherboard/companion chip/etc that the G4 (G3?) was introduced with. This has helped Apple keep the development of their consumer computers down since they just "adopted" the high end specs as the PM's grew. Moving to a unique archetecture (RM) for the G4 in the consumer line would break this "hand me down" technology advantage that they currently enjoy, and increase development cost.
So why does there seem to be this same aversion to the G4 name? There is a lot of misunderstanding going on, because for many the G4 name means Motorola, a slow bus and a history of performance problems. The G3 name on the other hand still seems to have good vibrations.
Maybe, only when we get a G4 iBook that uses a 2 GHz 750 VX Mojave processor that the cloud will lift from the G4 name. It will be a G4 because it will have the Velocity Engine and run applications that require a G4.
I can hardly believe I'm sying this, but... How much performance does an i/eMac/iBook need? I think a far more critical issue is to get the prices of these things DOWN, way DOWN, to a level where they nearly sell themselves. Numbers I've heard about the potential power-drain of a .09u G4 make it sound like a PERFECT laptop/AIO chip. Perfect, low power, cheap, and fast. It took a long time to get here, but once it does, there's no real reason NOt to use it. Will Winte;lon have a 64bit "consumer" chip? Not likely, nor for laptops. Lets assume that it uses a RapidI/O bus (there's the real surprise) while it still won't be as fast as PC970, it will give at least DDR266 level throughput, stretching to DDR400. I doubt you'll see laptops with that kind of performance, real laptops, not mobile desktops.
Doe Apple need a cheaper tower? yes, and it should be G5 based, kind of a pro's iMac, if you follow, small and unobtrusive, but with the same display flexibility as a tower and similar Gfx/CPU upgradability. A low end tower. This space is nearly decently covered now by the PMG4. I expect it to move to a G5 when supplies and costs allow, and shrink to sit between an iMac and a PMG5.
The Consumer machines just have to get cheaper, Apple needs to take advantage of plummeting Drive/LCD/RAM prices to make iMacs as cheap as possible, and not add the new expense of a new architecture just when the possibility to make a real impact has FINALLY presented itself. With G4 based drop in replacements at .13u and .09u getting cheaper, cooler, and faster, now is the time to strike at the heart of the issue, PRICE, and not chase some vain geek dream of supercomputer performance in an AIO/consumer notebook.
In one sense of the I agree with your statement about iMac/eMac performance. However right now it is lagging way behind, and needs to catch up. I also think that there is going to be a trend for more applications running concurrently as the digital hub model takes off.
My understanding of the archetecture is not that of a hardware engineer's, but what I do know of it seams to suggest that MP systems excell at the more parellel demands on the computer. I think that once the mhz gets up to a more respectable level, and the system bus inadequicies are brought up to speed, that the logical step would to move to MP systems. However the processors need to be fast enough so that the user "percieves" it to be on par or better than the offerings in the window world to sell these systems to switchers. I dont think that we are there yet with the G4's current speeds. Maybe the 57' will get us there with the added benefit of lowering the cost of the systems overall. Moto's Performance over that last 4 years has me skeptical on this though.
Originally posted by Matsu
I think Moto will surprise a lot of people with a pretty quick move to .09u.
It doesn't matter how quick they get the G4 going because without a new FSB, it is still data starved. This becomes even worse as the G4 gets faster. A 0.9nm G4 with MaxBus is really pathetic. Moto needs to also come out with a FSB for the G4.