Will Apple move to the POWER 5 instead of PPC?

13567

Comments

  • Reply 41 of 121
    G6 - More POWER on the way





    OSnews seems to agree with Brendon's speculation about Apple getting a POWER 5 "lite" from IBM



    Quote:

    All in all, the POWER5 is a very impressive processor, especially given that it is implemented in a previous-generation (130nm) silicon process and IBM use a deliberately conservative design process.



    If rumours are to believed, IBM are working on a desktop version of this CPU. The question is can IBM translate these sorts of performance gains to the desktop? We can't expect performance like POWER5 but what can we expect?





    ... The CPU will most likely be made using a 90nm process in a less conservative manner than the POWER5 so it will run faster and cooler. Quite how fast it will be able to run is open to question and I think it's fairly likely IBM will go the same route as AMD in their dual core plans and not clock the processor as high as possible to keep power consumption within reasonable limits. This will not be easy as POWER5 consumes 160 Watts at 1.8GHz (power consumption isn't much of an issue at the high end).



    One possibility is to use the same technique the POWER5 already uses which is to constantly adjust the clock frequency to keep heat output down. This technique is becoming popular with Transmeta and Intel doing or planning to do the same.



    Another possibility would be to use a technique Intel plan to use for the next Itanium "Montecito," which includes two peltiers in the heat sink. Peltiers actually consume quite a bit of power themselves but reducing the CPU temperature reduces transistor leakage, this lowers the power consumed by the CPU itself allowing boosts in clock frequency which might not otherwise be possible.



    Montecito is expected to consume 100 Watts but its heat sink requires a further 75 watts. The end effect is overall power consumption does not change (it may even go up) as part if moved to the heat sink but the CPU itself does not get so hot when working. AMD have filed a patent on an on-chip peltier so they're evidently considering similar technology.



    I don't know if the 9x0 will be so hot as to require such aggressive cooling but things are heading that way. "Power density" is becoming a problem and will seemingly only get worse in the future. Power density is the heat generated in a specific area; as CPUs get ever smaller the heat is generated in a smaller area and thus the unit becomes progressively more difficult to cool. The 970FX used in Apple's PowerMacs actually uses less power than the previous 970 but liquid cooling was added because of the higher power density.



  • Reply 42 of 121
    onlookeronlooker Posts: 5,252member
    pffffft
  • Reply 43 of 121
    BUMP!



    Interesting article:



    http://www.macworld.co.uk/news/index...S&NewsID=10439



    Maybe it will happen.
  • Reply 44 of 121
    Quote:

    Originally posted by onlooker

    pffffft



    If you must fart in public, please excuse yourself.
  • Reply 45 of 121
    Quote:

    Originally posted by cubist

    And how much market is there for $10K+ Macs anyway? Weren't they "the computer for the rest of us"?



    Believe it or not I thought that this would be why IBM would be interested in the idea, except from the other way around. I have no idea what the size of the POWER5 market is except I wold guess that the Mac market is many times larger. It appears that the cost of the POWER5 chips is mostly in R&D costs, because the manufacturing process is not cutting edge, and I would assume chosen because it is well understood and therefore more rubust. By selling these to Apple the R&D costs per unit go down quite allot. If IBM is selling these POWER5s in the 100,000s per year and they could sell them to Apple in the 1,000,000s pe year, the expanded market and the significantly lower R&D costs per unit would make good economic sence for IBM. Read IBM could sell them to Apple and to themselves for a much lower cost. As far as Altivec goes, I don't have an answer for that. Or maybe we are just talking servers. Or, what if there were to be a processor that because of its other technologies pretty much nullified Altivec, sorry Altivec fans, but then Apple would then not have to promote using a special processing unit, which may not be done just for the economics of time to port. Maybe this is what core media is all about, just let Apple write the libraries and other companies don't have to worry about wheather Altiec is being used or not. At some point Altivec because it cannot evolve at a fast pace will fall behind and Apple should be prepared, maybe we are seeing the beginning of that time. Maybe the transition is at foot.
  • Reply 46 of 121
    Quote:

    Originally posted by Brendon

    Or, what if there were to be a processor that because of its other technologies pretty much nullified Altivec, sorry Altivec fans, but then Apple would then not have to promote using a special processing unit, which may not be done just for the economics of time to port. Maybe this is what core media is all about, just let Apple write the libraries and other companies don't have to worry about wheather Altiec is being used or not. At some point Altivec because it cannot evolve at a fast pace will fall behind and Apple should be prepared, maybe we are seeing the beginning of that time. Maybe the transition is at foot.





    You are spewing nonsense here. The POWER5 currently shipping in IBM product was designed for big iron uses, not for Apple. Apple simply could not use the POWER5 as it is and the reason has nothing to do with AltiVec. There is no "move to POWER5 instead of PowerPC". They will move to POWER5 when it is a PowerPC.



    When they bring the POWER5 technology out in a "lite" version they will add AltiVec, just like they did for the 970. There is nothing about it not evolving at a fast pace -- you've only seen one processor (well, two if you count the 970/970FX) with it from IBM and you are making predictions about how they can't evolve it? Take it from me, there is no problem adding it to the POWER5 and making improvements to VMX is no different than making improvements to the other parts of the processor if the improvements are requested by their customer(s).



    The Core media initiatives are created by Apple as a way to provide services written by Apple to application developers that take direct advantage of any and all hardware advances, without application developers having to be aware of it. AltiVec can be used for many things beyond the scope of the Core services, however, and so it retains substantial value all on its own.
  • Reply 47 of 121
    ompusompus Posts: 163member
    The OSnews article acknowledges "there is pretty much zero public data available on the project..." In other words, its fantasy.



    As for Altivec 2, show me where the clamor is for an improved Altivec implementation? There is none and the dream of Altivec 2 is fueled by the same lust that drives folks to buy amps that "go to 11."



    So here's an extrordinarily bold GUARANTEE... IBM will make a successor to the 970. The successor will draw upon designs from other IBM chips. The bus and clock speed will be higher. It will more thermally efficient. It will be fabbed at a smaller size. This chip will appear sometime in the future.
  • Reply 48 of 121
    tinktink Posts: 395member
    Since were talking about if Apple will move to the POWER 5......</sarcasm>

    I'm hoping that the upcoming xbox, Playstation, Cube, etc. all use some implementation of Altvec/VMX/Velocity Engine, in their PPC cpu's. This increased programing knowledge base would no doubt trickle over to Apple software over time.
  • Reply 49 of 121
    Quote:

    Originally posted by Programmer

    You are spewing nonsense here. The POWER5 currently shipping in IBM product was designed for big iron uses, not for Apple. Apple simply could not use the POWER5 as it is and the reason has nothing to do with AltiVec. There is no "move to POWER5 instead of PowerPC". They will move to POWER5 when it is a PowerPC.



    When they bring the POWER5 technology out in a "lite" version they will add AltiVec, just like they did for the 970. There is nothing about it not evolving at a fast pace -- you've only seen one processor (well, two if you count the 970/970FX) with it from IBM and you are making predictions about how they can't evolve it? Take it from me, there is no problem adding it to the POWER5 and making improvements to VMX is no different than making improvements to the other parts of the processor if the improvements are requested by their customer(s).



    The Core media initiatives are created by Apple as a way to provide services written by Apple to application developers that take direct advantage of any and all hardware advances, without application developers having to be aware of it. AltiVec can be used for many things beyond the scope of the Core services, however, and so it retains substantial value all on its own.




    Spewing? A little graphic. I spew therefore I am. So I am spewing my opinions and speculations on a rumors board. It is a rumors forum. The future hardware forum is steeped in fact, I can't look at a single thread and say wow that is pure speculation. I have said that Altivec is difficult to evolve due to the fact argued by you and others that big changes to Altivec hamper its adoption, and could make life difficult for those that utilize it. By having to support Altivec new and Altivec old. So following that logic I would assume that Apple along with IBM would like to make Altivec like instructions part of the processor core. This appears t be what IBM is doing, in that they add special capabilties to parts of the execution units instead of having special unit like Altivec. Take for example the discussion about the design of the POWER6 processor, they are adding more stuff to be implemented in hardware, and not in a special execution unit like Altivec/VMX, instead of software. Will IBM continue to grow the PPC line yes, but that has little to do with wheather Apple chooses to use them. Again let's look at the adoption rate of Altivec, at some point it may be easier to move away from those issues. How well does the compilier work with making code work well with Altivec. How many people have the time to learn about when and how to use Altivec. As I see it Altivec bought Apple lead time at being able to encode DVDs cheaply. There are uses for it in products like Motion, FCP, Photograph and the like, but at some point just adding some extra capabilities to the FPU along with the fact that processors are so much faster than they were when we needed the 733MHz G4 to burn DVDs, when will those advantages not be so great. The point is everything evolves. Altivec goes back a long time with little evolution, great for adoption, but is that adoption rate, the rate that customers can see benefit enough to use the tools and to continue to use the tool, enough for Apple to continue to back it. Altivec is great for some things, but the rest of the processor is getting faster still, and IBM seems to like to add capabilities to the core and not rely on special processing units like VMX. IBM said from the beginning that they had no plans to use Altivec, it is there because Apple wants it there. I think that IBM would rather add special capabilities to the usual processing units that are easier to incorporate into the compilier and easier for the programmers to use. Just a different way of slicing the Apple, if you will.



    PS I just got an amp that does go to 11, love it, really. Much better than the 10 version.
  • Reply 50 of 121
    wizard69wizard69 Posts: 13,377member
    Quote:

    Originally posted by Ompus

    As for Altivec 2, show me where the clamor is for an improved Altivec implementation? There is none and the dream of Altivec 2 is fueled by the same lust that drives folks to buy amps that "go to 11."



    At least in my case, it is not a clamor for an improved Altvec implementation, rather it is rejection of the idea that AltVec can't be improved significantly. Some of the reasoning I see presented in this forum surprises me to no end as I expect that some of the people are technically literate here. The number one thing that people have to realize is that hardware that doesn't evolve dies. It is like TI saying that their last DSP chip is the best that ever will be and dropping all R&D for anything new. The fact that IBM, Motorola/Freescale & Apple did a good job with AltVec doesn't elminate the fact that technology moves forward.

    Quote:



    So here's an extrordinarily bold GUARANTEE... IBM will make a successor to the 970. The successor will draw upon designs from other IBM chips. The bus and clock speed will be higher. It will more thermally efficient. It will be fabbed at a smaller size. This chip will appear sometime in the future.



    See you do understand!!!!

    Others see ways obstructing the march forward, you do seem to realize that there is much yet that can come to market.



    Thanks

    Dave
  • Reply 51 of 121
    Quote:

    Originally posted by Brendon

    I have said that Altivec is difficult to evolve due to the fact argued by you and others that big changes to Altivec hamper its adoption, and could make life difficult for those that utilize it.



    This subject certainly is frequently misunderstood. What I argue against is changing the instruction set architecture unless there is huge benefit to be had. This applies to the basic PowerPC ISA just as much as it applies to the VMX extension to it. VMX is no more difficult to evolve than the rest of the ISA. There are all sorts of opportunities to improve on the current implementation in all areas, including VMX. And VMX is part of the core, it was a change made to the PowerPC ISA back when the G4 was introduced. It is just as much part of the core as the FPU is.



    Quote:

    Again let's look at the adoption rate of Altivec, at some point it may be easier to move away from those issues. How well does the compilier work with making code work well with Altivec. How many people have the time to learn about when and how to use Altivec.



    Come back again in 2 years and you will see much much wider adoption of VMX. You'll also see better tools for taking advantage of it more easily. I'm also hopeful that we'll see better implementations, much like the G4's implementation improved from the 7400 to the 7450. This is actually more likely than improvements to, say, the integer multiplier since IBM has evidently not improved that unit in the POWER4->5 transition.



    It will be interesting to see what IBM does in the POWER6. That will apparently be an all-new microarchitecture. I would not be at all surprised to see IBM build VMX into that, even for their servers, simply because the marginal cost of adding it is relatively slight and the demand for it is growing.



    Quote:

    Altivec is great for some things, but the rest of the processor is getting faster still, and IBM seems to like to add capabilities to the core and not rely on special processing units like VMX. IBM said from the beginning that they had no plans to use Altivec, it is there because Apple wants it there. I think that IBM would rather add special capabilities to the usual processing units that are easier to incorporate into the compilier and easier for the programmers to use. [/B]



    That was IBM's original stance, but things change. IBM will deliver what its customers want, and VMX is on that list. VMX isn't mutually exclusive with those other capabilities you're talking about -- we will likely see more specialized units in system-on-chip designs, but they usually will not take the form of modifications to the ISA. And most of them will be exposed via an operating system API as opposed to having direct hardware access.
  • Reply 52 of 121
    Quote:

    Originally posted by wizard69

    At least in my case, it is not a clamor for an improved Altvec implementation, rather it is rejection of the idea that AltVec can't be improved significantly. Some of the reasoning I see presented in this forum surprises me to no end as I expect that some of the people are technically literate here. The number one thing that people have to realize is that hardware that doesn't evolve dies. It is like TI saying that their last DSP chip is the best that ever will be and dropping all R&D for anything new. The fact that IBM, Motorola/Freescale & Apple did a good job with AltVec doesn't elminate the fact that technology moves forward.



    And sometimes hardware that does evolve dies -- if the direction is wrong then it wasn't a good evolution. This has been demonstrated repeatedly over the years, including a painful example or two by IBM.



    There is plenty of opportunity for improvement in AltiVec (and PowerPC in general) without changing the ISA. IBM said exactly this when they did a study to determine whether to build the POWER4, and the result was a compatible architecture in POWER4, POWER5 and (according to their plans) POWER6 & POWER7. It is pretty much the whole point of http://power.org/.





    This is one case where an automotive analogy works really well: to build faster cars, more fuel efficient cars, more capacious cars, etc. there is no need to change the basic driver interface: steering wheel, gas, brake, shift lever, and (maybe) a clutch. Having to deal with the difference between auto and manual causes enough trouble, imagine if every vehicle added a new kind of pedal or lever? And remember that you're more like a valet than a normal driver -- you need to be able to hop into any car and drive immediately. Now go and look at the diversity of vehicles available with this same basic interface. The advantages of maintaining the interface are tremendous, and there is still plenty of potential to be exploited.
  • Reply 53 of 121
    wizard69wizard69 Posts: 13,377member
    Quote:

    Originally posted by Programmer

    This is one case where an automotive analogy works really well: to build faster cars, more fuel efficient cars, more capacious cars, etc. there is no need to change the basic driver interface: steering wheel, gas, brake, shift lever, and (maybe) a clutch. Having to deal with the difference between auto and manual causes enough trouble, imagine if every vehicle added a new kind of pedal or lever? And remember that you're more like a valet than a normal driver -- you need to be able to hop into any car and drive immediately. Now go and look at the diversity of vehicles available with this same basic interface. The advantages of maintaining the interface are tremendous, and there is still plenty of potential to be exploited.



    This is the whole point there is much room to improve AltVec without changing what the existing code base sees. You can add a tubo charge to a car an many people will never know its true potential, but the user that knows it is there can exploit it to the maxium. The addition does not change the interface at all for the normal user.



    Likewise Atlvec can be extended to improve its performance. Like a turbo some of those improvements might give a user a boost that they are not aware of. For others they can milk (program) even more performance out of the unit.



    I think the bigger question is will AltVec even be supported on newer hardware as we know it today. Apple does have options as it moves forward, maybe we will see vector operations handled in a seperate execution unit.



    Dave
  • Reply 54 of 121
    Quote:

    Originally posted by Programmer

    And sometimes hardware that does evolve dies -- if the direction is wrong then it wasn't a good evolution. This has been demonstrated repeatedly over the years, including a painful example or two by IBM.



    There is plenty of opportunity for improvement in AltiVec (and PowerPC in general) without changing the ISA. IBM said exactly this when they did a study to determine whether to build the POWER4, and the result was a compatible architecture in POWER4, POWER5 and (according to their plans) POWER6 & POWER7. It is pretty much the whole point of http://power.org/.





    This is one case where an automotive analogy works really well: to build faster cars, more fuel efficient cars, more capacious cars, etc. there is no need to change the basic driver interface: steering wheel, gas, brake, shift lever, and (maybe) a clutch. Having to deal with the difference between auto and manual causes enough trouble, imagine if every vehicle added a new kind of pedal or lever? And remember that you're more like a valet than a normal driver -- you need to be able to hop into any car and drive immediately. Now go and look at the diversity of vehicles available with this same basic interface. The advantages of maintaining the interface are tremendous, and there is still plenty of potential to be exploited.




    OK the point for discussion purposes only, is Altivec or no. Let's assume no. The assumption could be based on assumptions that maybe 15% of the programs for the Mac use Altivec, and out of those that do use it only 25% fully exploit most of the capabilities of Altivec. If that is anywhere close to the case then Apple maybe better served by asking IBM to alter the FPU, and save the transistors and the development costs. Basically take the few instructions that are used the most and keep those and lose the rest. Now it appears to me that the compilier is easier to write so that it takes advantage of the extra capabilities of the FPU and or Int.U and more people will be able to benefit from those capabilities, on a more straight forward way. No more special Altivec code and then the rest of the code. To me it appears that Apple could be better served by having more people to work on the whole compilier rather than supporting and improving tools for a unit that sees very little use in the market as a whole. I know that the adoption rate will grow but will that rate keep pace with the growth of developers coming to the Mac or will Altivec be marginalized further? When the economics of coding choice meet the economics of tools and support of those tools. At that time maybe the POWER5 may look very good in that, the manufacturing process is very robust, they probabily get high yields. All of the redundentcy may not be needed making the processors made for Apple to be indenticle to the POWER5 except for the redundency, making those easier to make and cheaper. Yields may also play a key role as well in that if Wall Street is to be believed and Apple comes out with a flash memory player and the amount of people caught in the halo effect grows or continues to grow, Apple could easily double the amount of shipping Macs, in one year, '05. Will IBM be able to keep up with that growth? Maybe they could if the processors used were spread out over a larger product line, one that includes the POWER5. I don't know anything about the yields of the POWER line but that along with the fact that the POWER line is kicking some serious assumptions aside, may be another reason for Apple to put these in the desktop line thus freeing up more PPC for the consumer lines.
  • Reply 55 of 121
    Quote:

    Originally posted by wizard69

    This is the whole point there is much room to improve AltVec without changing what the existing code base sees. You can add a tubo charge to a car an many people will never know its true potential, but the user that knows it is there can exploit it to the maxium. The addition does not change the interface at all for the normal user.



    What's this? We agree? Do I hear hell freezing over?



    Quote:

    I think the bigger question is will AltVec even be supported on newer hardware as we know it today. Apple does have options as it moves forward, maybe we will see vector operations handled in a seperate execution unit.



    First of all, what do you mean by "a seperate execution unit"? Right now AltiVec is a pair of seperate execution units -- seperate from the load/store, integer, and floating point execution units. I suspect you mean as a seperate core, like Cell is rumoured to be. In this case I agree but they don't have to be mutually exclusive: the presence of specialized vector cores doesn't obviate the need for vector execution units in the general purpose core any more than specialized floating point cores would obviate the need in the general purpose core.



    VMX will be supported, mark my words.
  • Reply 56 of 121
    Quote:

    Originally posted by Brendon

    OK the point for discussion purposes only, is Altivec or no. Let's assume no. The assumption could be based on assumptions that maybe 15% of the programs for the Mac use Altivec, and out of those that do use it only 25% fully exploit most of the capabilities of Altivec.



    Only that many use it directly, but many use libraries and OS functionality that leverages AltiVec. Some of this functionality would make sense to move onto new kinds of hardware but some doesn't.



    Quote:

    If that is anywhere close to the case then Apple maybe better served by asking IBM to alter the FPU, and save the transistors and the development costs.



    I think Apple has proven that AltiVec is more than valuable enough to retain. The same number of transistors added to the FP or integer capabilities of the basic PowerPC core would have much less return on investment.



    Quote:

    Basically take the few instructions that are used the most and keep those and lose the rest.



    Either you lose 'em all or you keep 'em all. The marginal savings by removing selected instructions is virtually zero, and the cost is breaking compatibility.



    Quote:

    Now it appears to me that the compilier is easier to write so that it takes advantage of the extra capabilities of the FPU and or Int.U and more people will be able to benefit from those capabilities, on a more straight forward way. No more special Altivec code and then the rest of the code. To me it appears that Apple could be better served by having more people to work on the whole compilier rather than supporting and improving tools for a unit that sees very little use in the market as a whole.



    Vectorization support is coming in the next GCC, and Apple has optimized a great deal of its support code to use AltiVec already. Going back now would be pointless, and in the future this kind of SIMD hardware is going to be more common, not less. That will further improve the tools and increase the amount of coding using it.



    There is no problem adding VMX to the POWER5 (and future) designs. They did it with POWER4 without causing trouble, and at the very least POWER5 can adopt the VMX unit from the POWER4 (but will likely improve on it). There are much bigger factors that prevent Apple from using POWER5 as it is, so much so that the lack of VMX is pretty much inconsequential. In the process of addressing those issues, adding in VMX isn't an issue.
  • Reply 57 of 121
    Quote:

    Originally posted by Programmer

    Only that many use it directly, but many use libraries and OS functionality that leverages AltiVec. Some of this functionality would make sense to move onto new kinds of hardware but some doesn't.



    I think Apple has proven that AltiVec is more than valuable enough to retain. The same number of transistors added to the FP or integer capabilities of the basic PowerPC core would have much less return on investment.



    Either you lose 'em all or you keep 'em all. The marginal savings by removing selected instructions is virtually zero, and the cost is breaking compatibility.



    Vectorization support is coming in the next GCC, and Apple has optimized a great deal of its support code to use AltiVec already. Going back now would be pointless, and in the future this kind of SIMD hardware is going to be more common, not less. That will further improve the tools and increase the amount of coding using it.



    There is no problem adding VMX to the POWER5 (and future) designs. They did it with POWER4 without causing trouble, and at the very least POWER5 can adopt the VMX unit from the POWER4 (but will likely improve on it). There are much bigger factors that prevent Apple from using POWER5 as it is, so much so that the lack of VMX is pretty much inconsequential. In the process of addressing those issues, adding in VMX isn't an issue.




    Thanks for helping me with my adoption numbers, I was sweating all of those assumptions.



    Are you sure about "Either you lose 'em all or you keep 'em all." statement, that seems strange to me. But if you are correct your point of all or nothing would mean a lock for VMX.



    I didn't know vectorization was coming to the GCC, I wonder if this will boost adoption or splinter the tools support, I really have no idea.



    Could you please elaborate on the whys of why Apple would have problems and factors with using the POWER5 with and without VMX. Assume that I know nothing. Would heat be one of these issues? Price? IBM being a hording bastion? Apple waiting for POWER11?
  • Reply 58 of 121
    Quote:

    Originally posted by Programmer



    There is no problem adding VMX to the POWER5 (and future) designs. They did it with POWER4 without causing trouble, and at the very least POWER5 can adopt the VMX unit from the POWER4 (but will likely improve on it). There are much bigger factors that prevent Apple from using POWER5 as it is, so much so that the lack of VMX is pretty much inconsequential. In the process of addressing those issues, adding in VMX isn't an issue.




    VMX or not the POWER5 real world numbers are right up there with some theoretical numbers. With and without VMX, if you know what I mean.
  • Reply 59 of 121
    Quote:

    Originally posted by Brendon

    Thanks for helping me with my adoption numbers, I was sweating all of those assumptions.



    I don't have any hard numbers, but those sound like a reasonable guess. The fallacy of that kind of metric, however, is that most of the 75% that don't use AltiVec don't need AltiVec.



    Quote:

    Are you sure about "Either you lose 'em all or you keep 'em all." statement, that seems strange to me. But if you are correct your point of all or nothing would mean a lock for VMX.



    I argue for leaving the VMX instruction set as-is so that existing software that uses it will continue to operate correctly. If you remove instructions, some software will break. If you add some instructions and developers use those then you've fragmented your installed base. The major cost of VMX is that there are execution units & pipelines for it, and the 32 128-bit registers. As soon as you have 1 VMX instruction you need those things, so you might as well go the whole 9 yards.



    If legacy software wasn't an issue (i.e. you're not Apple) then you might make changes to the ISA, depending on your anticipated audience. This is the most likely reason for VMX2 to show up -- if IBM decides to add it to POWER6 (or some other member of power.org decides they want it in another processor) they have no legacy and thus are free to introduce additional instructions and changes, regardless of Apple's legacy requirements. In the case of IBM doing it, this might then trickle down to a PPC990 or something then Apple machines would end up with it. So far I have seen no sign of IBM making changes to the VMX ISA, however.



    Quote:

    I didn't know vectorization was coming to the GCC, I wonder if this will boost adoption or splinter the tools support, I really have no idea.



    Even more telling is that IBM is adding it to their POWER compiler. That ought to be a clue of things to come in the future.



    Quote:

    Could you please elaborate on the whys of why Apple would have problems and factors with using the POWER5 with and without VMX. Assume that I know nothing. Would heat be one of these issues? Price? IBM being a hording bastion? Apple waiting for POWER11?



    The cache architecture, the MCM physical organization, the size of the cache, the memory architecture, that the chip is designed for reliability over density/yield and the costs that result from that, etc. There are a lot of things that go into chip design that don't show up from an ISA point of view. These things just aren't condusive to the kinds of machines and the price point that Apple targets.
  • Reply 60 of 121
    powerdocpowerdoc Posts: 8,123member
    This is a recurrent thread, many posters like The programmer have emphasized this : the power line is much more than a dual core PPC line. It's a server chip, with features especially made for such tasks.



    You will never see a power chip in an Apple desktop. It cost way too much, it will not improve the performances.

    The PPC 970 chip and derived are the future of Apple.

    There is many improvement to do with this chip :

    - going dual core and multicore

    - improving the core without changing the ISA. For example Altivec can be optimised, like Mot did with the 7450 compared to the 7400. The core can be also more pipelined ...

    - greater L1 and L2 cache
Sign In or Register to comment.