more rampant speculation on future of Power Mac
I know it's yet more rampant speculation and this board is probably already saturated, but just thought i'd throw in an alternate hypothesis. NOTICE, most of this is probably wrong or unfeasable as well. Also it all depends on something like the M.I.A. PPC7500 eventually being produced by somebody. anyway, something thats been bothering me about all the speculation around here is that most people seem to assume that a next generation processor is imminent, as in sometime in 2003. i would tend to go with a more pessimistic outlook that actually wouldn't be too bad.
there is no G5, at least not in the next couple of years. instead, what Apple does is draw out the life of the G4 via multi processing. the major hurdle to that right now is the FSB. it seems pretty certain that the FSB issue will be taken care of very soon (either late 02 or early 03). I would tend to think that at MWNY02 we'll see PowerMacs with 166mhz MPX bus and 266mhz or 333mhz DDR RAM, which will soon trickle down to the entire lineup. Then sometime in 03 we will see the PPC7500 emerge on a 0.13 process with support for Rapid IO.
Even if Moto isn't actually developing this chip for themselves (they have taken it off their PPC roadmap), Apple has the funds to pay for all the development costs and then some. It's not too terribly complicated to move the same basic G4 design over the 0.13 process and provide support for RIO as well as other smaller optimizations they could do if they wanted. Maybe the chip could be more like a 7480 or 90. The big one would be RIO support tho, 0.13 process isn't really necessary for what i had in mind, tho it would be nice. I don't think Moto would be averse to this because it would extend the life of the G4 for them as well, if Apple allowed them to use the resulting chip for their own purposes (that would be added incentive for moto to give it a shot).
anyways, what we would see sometime in the beginning of 03 would be a PPC7500 (with RIO) powered PowerMac with a fast FSB and 333mhz DDR RAM at least. I forget exactly what the maximum throughput of RIO is but it should be plenty fine for a long time. Anyways, this would allow Apple to offer systems with 4 and more processors immediately because 1)the FSB can handle it and 2)the 7500 on 0.13 process would be smaller, be less power hungry and create less heat than an earlier 0.18 7450. Also, as the G4 keeps being produced the unit costs will continue to decline so basically Apple can put as many procs as the FSB and OS can handle without losing their margins too much or raising prices.
the disadvantage to lots of less powerful procs over 1 or 2 really powerful procs is that software needs to be designed specifically to fully take advantage of the multiple procs. this would make it less attractive, but Apple seems to be solving some of this problem by bringing lot's of important apps in house (Shake, Logic etc.) Therefore lots of the more powerhungry apps for markets that apple is obviously after will run great on quad and octo or whatever systems that apple is selling because apple will make sure that the programs are optimized to run well on these systems (phew). And apple could concievably send engineers out as consultants to other software companies to make sure that all software they make will take advantage of the MPs (like they did when Altivec was released) as well as releasing guidelines for programming for MP to the general public.
I would guess that the 7500 would eventually max out at around 2 ghz or so which isn't that impressive. But an 8 way 2ghz system running at the limits of RIO and DDRAM capabilities should last into 2004/05 I would think, and would be competitive with the best of the consumer/low-pro wintel offerings. I'm pretty sure that apple has the mobo technology to make this happen, and it seems to make sense. in that way the edge that the intel world has over apple as far as individual processor tech and power can be overcome at least for a time. It all relies on whether Apple can somehow get a RIO capable G4 into the marketplace. even the 0.13 process isn't really necessary for a system like that, but the RIO is.
if this or something like it does actually take place then i would guess the next gen processor would in fact be some derivative of the IBM Power5. The Power4 in it's current form is not practical for a desktop machine, and I doubt that IBM would want to make the significant changes that it would take to make this possible because the Power4 is already fully designed and in production, and the potential marketplace is small. apple should have realized when the G4 stagnated that Moto couldn't be relied upon to produce high powered desktop chips for the future and were more interested in the embedded market. if then they got involved with the develpment of the Power5 at an early stage, it would be considerably easier to convince IBM to design the chip in such a way to allow for easily-designed low-power desktop/laptop derivatives of that chip that could be used in future apple hardware at all levels. i don't know the exact timeframe for the Power5's release but I would imagine it to be a fair ways off.
sorry for the long and ill informed post.
there is no G5, at least not in the next couple of years. instead, what Apple does is draw out the life of the G4 via multi processing. the major hurdle to that right now is the FSB. it seems pretty certain that the FSB issue will be taken care of very soon (either late 02 or early 03). I would tend to think that at MWNY02 we'll see PowerMacs with 166mhz MPX bus and 266mhz or 333mhz DDR RAM, which will soon trickle down to the entire lineup. Then sometime in 03 we will see the PPC7500 emerge on a 0.13 process with support for Rapid IO.
Even if Moto isn't actually developing this chip for themselves (they have taken it off their PPC roadmap), Apple has the funds to pay for all the development costs and then some. It's not too terribly complicated to move the same basic G4 design over the 0.13 process and provide support for RIO as well as other smaller optimizations they could do if they wanted. Maybe the chip could be more like a 7480 or 90. The big one would be RIO support tho, 0.13 process isn't really necessary for what i had in mind, tho it would be nice. I don't think Moto would be averse to this because it would extend the life of the G4 for them as well, if Apple allowed them to use the resulting chip for their own purposes (that would be added incentive for moto to give it a shot).
anyways, what we would see sometime in the beginning of 03 would be a PPC7500 (with RIO) powered PowerMac with a fast FSB and 333mhz DDR RAM at least. I forget exactly what the maximum throughput of RIO is but it should be plenty fine for a long time. Anyways, this would allow Apple to offer systems with 4 and more processors immediately because 1)the FSB can handle it and 2)the 7500 on 0.13 process would be smaller, be less power hungry and create less heat than an earlier 0.18 7450. Also, as the G4 keeps being produced the unit costs will continue to decline so basically Apple can put as many procs as the FSB and OS can handle without losing their margins too much or raising prices.
the disadvantage to lots of less powerful procs over 1 or 2 really powerful procs is that software needs to be designed specifically to fully take advantage of the multiple procs. this would make it less attractive, but Apple seems to be solving some of this problem by bringing lot's of important apps in house (Shake, Logic etc.) Therefore lots of the more powerhungry apps for markets that apple is obviously after will run great on quad and octo or whatever systems that apple is selling because apple will make sure that the programs are optimized to run well on these systems (phew). And apple could concievably send engineers out as consultants to other software companies to make sure that all software they make will take advantage of the MPs (like they did when Altivec was released) as well as releasing guidelines for programming for MP to the general public.
I would guess that the 7500 would eventually max out at around 2 ghz or so which isn't that impressive. But an 8 way 2ghz system running at the limits of RIO and DDRAM capabilities should last into 2004/05 I would think, and would be competitive with the best of the consumer/low-pro wintel offerings. I'm pretty sure that apple has the mobo technology to make this happen, and it seems to make sense. in that way the edge that the intel world has over apple as far as individual processor tech and power can be overcome at least for a time. It all relies on whether Apple can somehow get a RIO capable G4 into the marketplace. even the 0.13 process isn't really necessary for a system like that, but the RIO is.
if this or something like it does actually take place then i would guess the next gen processor would in fact be some derivative of the IBM Power5. The Power4 in it's current form is not practical for a desktop machine, and I doubt that IBM would want to make the significant changes that it would take to make this possible because the Power4 is already fully designed and in production, and the potential marketplace is small. apple should have realized when the G4 stagnated that Moto couldn't be relied upon to produce high powered desktop chips for the future and were more interested in the embedded market. if then they got involved with the develpment of the Power5 at an early stage, it would be considerably easier to convince IBM to design the chip in such a way to allow for easily-designed low-power desktop/laptop derivatives of that chip that could be used in future apple hardware at all levels. i don't know the exact timeframe for the Power5's release but I would imagine it to be a fair ways off.
sorry for the long and ill informed post.
Comments