What is the real reason then? Contracts with Freescale and IBM? Negative perception of Mac OS X? Deal not closed yet with Intel? hmmm intriguing
We may never know the full details, but just consider all the advantages of going with Intel that are just now coming to light: lower engineering and other development costs, better support for graphics chips and more graphics options, faster introduction of new technology, and just better commitment to the needs of personal computers rather than the embedded chip market and other specialties. I'm sure such items were a thorn in Apple's side for a long time. I think Intel's commitment to low power and the laptop platform in general was the clincher in this deal, however. Also the timing was right. I don't believe the move would have been as easy if it came earlier.
On the subject of keeping doors open, it would not surprise me if Apple keeps the door open for some PPC models of the Mac in the future. That should be easy to do since all code will now be written for two CPUs.
On the subject of keeping doors open, it would not surprise me if Apple keeps the door open for some PPC models of the Mac in the future. That should be easy to do since all code will now be written for two CPUs.
Right, I was just thinking that we may eventually see both CPUs in the same line of products in the beginning, for example: in the powermac line: a low-end dual (core) G5, a high-end dual-dual-core G5 and a Intel PowerMac @ +3GHz (for the good cause). We may eventually see another form factor for the Intel-based powermacs because of the cooler CPUs, 19" rack or small towers. PPC and Intel may co-exist for a while... at least until the end of 2007.
. . . PPC and Intel may co-exist for a while... at least until the end of 2007.
It's a fact that PPC and Intel will coexist into 2007, when the transition will finally be complete. I'm thinking it will be a total transition to Intel initially, so developers in all market segments completely convert their code and have applications running on both CPUs. Afterward, there should be no technical reason preventing Apple from reintroducing PPC models of the Mac. I could see where there might be good reasons for doing so, but I believe the majority of Apple's computer products will stay with Intel from now on.
It's a fact that PPC and Intel will coexist into 2007, when the transition will finally be complete. I'm thinking it will be a total transition to Intel initially, so developers in all market segments completely convert their code and have applications running on both CPUs. Afterward, there should be no technical reason preventing Apple from reintroducing PPC models of the Mac. I could see where there might be good reasons for doing so, but I believe the majority of Apple's computer products will stay with Intel from now on.
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
Good question. In addition to the many advantages of Apple going with Intel, the decision was at least partly based on performance. But even Steve Jobs doesn't have a perfect Crystal Ball. IBM is going to continue their pursuit of better CPU design, and IBM is determined to beat Intel in the server game. Servers are IBM's bread and butter. The semiconductor devision was even merged into the server devision at one point. It is also clear that IBM will continue to look for areas where their technology can harvest more profit. Hence IBMs pursuit of the relatively large game console CPU market. Apple was beginning to be a burden to IBM it seems, and IBM didn't have the time and engineering resources to invest in keeping Apple satisfied. But all this should begin to change once the current rush of server and game console development is over.
Sooner of later IBM will have chips that Apple may be very interested in. Who knows what is coming after the first Cell and Trinity chips? IBM will not be standing still. If there is a juicy PPC processor that would work better than an Intel chip for some particular model of Mac, like a server, Apple would be foolish to not include it in their product line. It may not happen for maybe four years, but it's a good bet that it will happen.
I'm puzzled by your last comment about Apple developing PPC software for such a Mac. The software will be done. In a couple years, all applications will work on either a PPC or Intel Mac. There is no reason to suddenly drop the PPC versions of Mac OS and applications once the transition is complete.
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
hi mel,
i think one of the good things that apple is encouraging is basic decent computer science. there is no reason to write low-level hardware-specific code unless you have bloody good reason to.
mac os X as a platform has a lot going for it, and i think what is happening is that apple is trying to say to everyone, well, steve did say, "this is the soul of the Mac". it doesn't matter if it wears a cross, a rosary in hand, or a turban, or a skullcap, whatever, it's still a Mac.
once all major and new developers get used to writing good hardware-abstracted universal binaries, and for example start using accelerate.framework more than specific code for altivec, or this or that spe or sse (not sure what the difference is, i'm typing this on something with MMX )
...i think that then apple will get to pick and choose the hardware, which we know, hardware is becoming a real commodity game.
personally though i suspect that IBM has to come up with some hardcore serious shit to entice apple back into the PPC fold post 2007.
although, for that matter, if tomorrow some big corporation maybe out of china or india invented a laptop/dekstop platform that had some "smart" nano-material in it that could convert 50% of the waste heat of the CPU and components back into electrical energy into the battery, stevie J would be all over them like a bad rash.
he would encourage them to use Intels in that system though or sell the tech lock stock and barrel to Intel/Apple
i think one of the good things that apple is encouraging is basic decent computer science. there is no reason to write low-level hardware-specific code unless you have bloody good reason to.
mac os X as a platform has a lot going for it, and i think what is happening is that apple is trying to say to everyone, well, steve did say, "this is the soul of the Mac". it doesn't matter if it wears a cross, a rosary in hand, or a turban, or a skullcap, whatever, it's still a Mac.
once all major and new developers get used to writing good hardware-abstracted universal binaries, and for example start using accelerate.framework more than specific code for altivec, or this or that spe or sse (not sure what the difference is, i'm typing this on something with MMX )
...i think that then apple will get to pick and choose the hardware, which we know, hardware is becoming a real commodity game.
personally though i suspect that IBM has to come up with some hardcore serious shit to entice apple back into the PPC fold post 2007.
although, for that matter, if tomorrow some big corporation maybe out of china or india invented a laptop/dekstop platform that had some "smart" nano-material in it that could convert 50% of the waste heat of the CPU and components back into electrical energy into the battery, stevie J would be all over them like a bad rash.
he would encourage them to use Intels in that system though or sell the tech lock stock and barrel to Intel/Apple
...............
I was going to abbreviate your name, but all I came up with was SM, so I decided not to.
Well, I don't think that it will happen.
Firstly, if that were possible Jobs didn't have to make such a break. He didn't have to say that they would transition over to x86, but rather that they would be adding x86 machines to the line where they were needed. Remember the power/watt thing? The Powermacs and servers could have stayed with PPC and the rest made the move. But he didn't say that. ALL will be transitioned over.
There is also credibility. If he moves, even in part, back to PPC, Apple will really be looked upon with distaste by the developers.
Don't forget that this universal binary thingy is not easy. If a program is coded for the G5 and Altivec it will be difficult to re-encode. This is in the documents. It's a matter of maintaining two code bases. Who will want to do that? Apple will lose some developers as it is. They will lose more if they believe that they will have to continue two seperate development lines. Even with Cocoa and Xcode.
What would happen if they are told two or three years after the transition, when they are getting ready to slow developement down for PPC that they have to bring it back up? It won't be good.
Apple will also have to maintain two seperate hardware development lines. That's expensive.
I can believe that they will maintain the PPC code base for OS X the way they did with x86, but only in case disaster strikes.
. . . Don't forget that this universal binary thingy is not easy. If a program is coded for the G5 and Altivec it will be difficult to re-encode. This is in the documents. It's a matter of maintaining two code bases.
What you say is true if the application was not developed with Apple tools. In anticipation of the move to Intel, Apple has been preparing its tools to make the transition easy. The only exception I know of is accelerate framework, which provides only the most common functions for AltiVec code. If a developer makes heavy use of AltiVec for operations outside of accelerate framework, he or she will have some work to do. Even so, it is no more work than the developer would do today if both Mac and Windows versions of the product are made.
I think the whole point of WWDC was to show developers how easy their task can be, using proper tools. I got a call from some developers just before Steve came on stage. They didn't believe the rumors and had a lot of time and energy invested in the PPC. After the keynote, their spirits were very low, but by the end of the conference most were even enthusiastic about the move to Intel. I can check back, but the last I heard it was not a big deal to maintain two Mac versions. My real concern at the time was that developers would drop PPC support too quickly, making the PPC Macs almost worthless.
Quote:
Apple will also have to maintain two seperate hardware development lines. That's expensive.
IBM maintains a line of PPC and Intel servers, giving customers a choice of what they want. Using two different CPUs in hardware is no big deal. Each product is individually designed. The majority of Macs using an Intel CPU will be faster and easier to design because Intel provides complete chip sets. Apple may even use a few Intel motherboards on some products. Yet this will not make the design of a PPC Mac any more difficult than it is today. If a PPC chip is good enough to warrant inclusion in the Mac lineup, Apple could easily do it. And don't forget, this would be years away. With IBM's slow but sure push of the PPC, IBM may begin to make some supporting chips for systems designers.
What you say is true if the application was not developed with Apple tools. In anticipation of the move to Intel, Apple has been preparing its tools to make the transition easy. The only exception I know of is accelerate framework, which provides only the most common functions for AltiVec code. If a developer makes heavy use of AltiVec for operations outside of accelerate framework, he or she will have some work to do. Even so, it is no more work than the developer would do today if both Mac and Windows versions of the product are made.
I think the whole point of WWDC was to show developers how easy their task can be, using proper tools. I got a call from some developers just before Steve came on stage. They didn't believe the rumors and had a lot of time and energy invested in the PPC. After the keynote, their spirits were very low, but by the end of the conference most were even enthusiastic about the move to Intel. I can check back, but the last I heard it was not a big deal to maintain two Mac versions. My real concern at the time was that developers would drop PPC support too quickly, making the PPC Macs almost worthless.
IBM maintains a line of PPC and Intel servers, giving customers a choice of what they want. Using two different CPUs in hardware is no big deal. Each product is individually designed. The majority of Macs using an Intel CPU will be faster and easier to design because Intel provides complete chip sets. Apple may even use a few Intel motherboards on some products. Yet this will not make the design of a PPC Mac any more difficult than it is today. If a PPC chip is good enough to warrant inclusion in the Mac lineup, Apple could easily do it. And don't forget, this would be years away. With IBM's slow but sure push of the PPC, IBM may begin to make some supporting chips for systems designers.
What I'm seeing is that it's difficult to transition Altivec, only some of it has an equivalent command. Some of the rest has to be painfully redone.
In many cases, code written for the G5 has to be hand tuned afterwards.
Insofar as maintaining a code line for Windows and OS X, you are now asking developers to also maintain a code line for PPC. That's three. If OS X had a much larger marketshare, fine. But it doesn't. It can't cost more to maintain code for OS X than it for for the much larger market that Windows represents.
As far as hardware lines go, IBM is a much larger company than Apple is, almost $100 billion vs 13 billion. Don't forget that IBM is also about equal with MS as far as software goes. They are much LARGER than MS is in every category except for OS and Office. All of these lines of computers they make come with hefty maintenance contracts as well, not just HW maintenance, but software to run a business as well.
Apple has nothing to compete, and has no interest in doing anything to make a business easer to run using their machines.
Apple's software sales will be a bit over $1 billion this year vs IBM's $50 billion. As software is much more profitable than HW, IBM can afford to cater to various customers desires in HW in order to sell their their business services. They are a service company. Apple is not.
What I'm seeing is that it's difficult to transition Altivec, only some of it has an equivalent command. Some of the rest has to be painfully redone. . .
This is true. The developer who call just before the keynote has a lot of AltiVec code that must be converted because there is nothing in Accelerate framework to do the job. However, when it is complete he will have that part of a Windows version of his application finished. So it is not "extra" work, but something he would have to do anyway. I'm not saying that keeping a PPC Mac version will be free, but it should be relatively easy. In the worst case scenario, some developer will simply not optimize the PPC code after a while and PPC performance will suffer. However, if Apple introduce another PPC model several years from now, I'm sure those doing mission critical code will put in the extra effort.
Quote:
As far as hardware lines go, IBM is a much larger company than Apple is, almost $100 billion vs 13 billion.
My main point was that using two different CPUs in hardware is not a big deal. Each product is individually designed. A PPC will cost more to develop because IBM does not now make chip sets to go with it. That is why Apple would stay with Intel for the majority of Macs. However, if the PPC performance warrents it, and the market can bear the little extra cost burden of a PPC Mac, why not do it?
This is true. The developer who call just before the keynote has a lot of AltiVec code that must be converted because there is nothing in Accelerate framework to do the job. However, when it is complete he will have that part of a Windows version of his application finished. So it is not "extra" work, but something he would have to do anyway. I'm not saying that keeping a PPC Mac version will be free, but it should be relatively easy. In the worst case scenario, some developer will simply not optimize the PPC code after a while and PPC performance will suffer. However, if Apple introduce another PPC model several years from now, I'm sure those doing mission critical code will put in the extra effort.
My main point was that using two different CPUs in hardware is not a big deal. Each product is individually designed. A PPC will cost more to develop because IBM does not now make chip sets to go with it. That is why Apple would stay with Intel for the majority of Macs. However, if the PPC performance warrents it, and the market can bear the little extra cost burden of a PPC Mac, why not do it?
Because it's not as easy as you're making it out to be.
If a developer makes improvements to a program, adds features, or rewrites the code from the groung up as they have to do after a while, it's a major job. As people migrate over to the new machines it will become less attractive to maintain an increasingly improved and complex program for both platforms (Mac platforms, that is). After a while the PPC software will become a problem child, as OS 9 programs have begun to be.
If after several years Apple suddenly announces a new PPC machine, that announcement will land with a thud. Developers will be angry and frustrated. If they have Windows products they may leave altogether. This has happened before. Developers are still leaving.
No, I hope that Apple never feels as though it has to do that.
As fas as hardware goes, it costs a fortune to develop support chips. If Apple is only going to use them on one line of machines they won't be able to make enough to get the cost down. Two series of mobo's with different architectures, etc. I don't think so. If IBM built controllers into the cpu as AMD does, it might be different. But IBM doesn't have the interest in the G5 for Apple's use. Apple bought about 75% of that chip's production. What will IBM do with it now> Apple will still be using it for another 2 years or so. After that - poof!
What THT is saying is most general user apps have several threads, but not all of them are actively running at the same time. At any particular time most of the work is being done by one thread and that will run faster on a faster single processor than on a slower multi-processor box.
Yes, most of the web browsers, office apps and Internet apps, which consumers spend 99% of their time in, seem to have only 1 compute-bound thread while the others are waiting for input.
Quote:
But if you try to do many things simultaneously, you would finish the collective tasks faster on a multi-CPU box, than a faster singler CPU box.
Well, it would depend on how much of a clock rate advantage the single CPU box has. In addition, how many apps are going to have 4 compute-bound threads to fully take advantage of a quad box?
If Apple ever comes out with a quad box, I would think it would have to be a dual 2.5 GHz 970mp at the top end rather than 2 GHz. With the 2 GHz, I'm not too sure it would really provide enough of a performance improvement over the dual 2.7 970fx PowerMac G5.
Yes, most of the web browsers, office apps and Internet apps, which consumers spend 99% of their time in, seem to have only 1 compute-bound thread while the others are waiting for input.
Well, it would depend on how much of a clock rate advantage the single CPU box has. In addition, how many apps are going to have 4 compute-bound threads to fully take advantage of a quad box?
If Apple ever comes out with a quad box, I would think it would have to be a dual 2.5 GHz 970mp at the top end rather than 2 GHz. With the 2 GHz, I'm not too sure it would really provide enough of a performance improvement over the dual 2.7 970fx PowerMac G5.
. . . As fas as hardware goes, it costs a fortune to develop support chips. If Apple is only going to use them on one line of machines they won't be able to make enough to get the cost down. . . .
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
It would cost significantly more to build chips for perhaps one quarter the number of machines. how much more would people pay for these machines?
When my old company (I was a partner in a Pro/hi end audio manufacturing concern) decided to come out with a new product, we had to determine what the purpose of this product was. How many we would sell, and what we could charge, based on whatever competing products were already out there, as well as possible new ones. All of these factors are interrelated.
Apple has the same problem.If other products are decreasing in cost, then Apple must also decrease the costs of its products.
I can't go into all of the complexities of the design and manufacturing process here, but every item adds cost in different ways. If the item is off the shelf, then the cost of the item might be doubled in the final product pricing.
If R&D is required for a part, then the cost is going to be much higher. The overall R&D costs have to be recovered. If it costs $50 million for the R&D, the resulting price has to be determined by simple math. How many products are to be assumed as being sold within a specific time to cover those costs? It also has to cover start up costs involved in production, etc.
If total costs are $100 million, and it's expected to sell 500 thousand units during that time, then it will add $200 to the final product's cost.
If other competetitors don't have those costs, then you are at a disadvantage.
Apple's costs are much higher because they already have costs that others don't. The R&D for the OS, for example. That cost is pretty much fixed. Whether Apple sells 3 million machines or 10 million, the cost to develop it remains the same. MS sells about 200 million copies a year, so even if their costs are much higher, it's less per copy.
To burden Apple with two different lines makes it more difficult for them to recover those costs. You can't just say that they can do it. You have to cost it out. Have you done that?
As far as IBM goes, you don't know what they will do. The 970 is a derivative of the Power, but it is not the Power. The Power chips are a chip set, and they are far more expensive than the G5's. Will they have a controllor on board? Maybe. But most likely not. Could IBM build such a chip with the thermal and pricing characteristics that Apple would need? Why should they?
If IBM finds that they don't need such a chip, they won't built it. Remember the R&D costs. How many of those chips would they have to sell? To what market would they be selling it? And what would be the guarantee that they would have any more success than they are having now in meeting Apple's needs?
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
if IBM offers Apple a realistic and deliverable 9xx Power-6 derivative that meets it needs, Apple would give it some thought. But performance/watt has gotta be high, yields have got to be promising (how it is promised i have no idea ), and IBM has until the end of 2006 to put it on the table. By the start of 2007, if such a Power-6 derivative does not look promising enough to Apple, that's bye bye to IBM... maybe they'll be back next decade with some killer chip for Mac OS XI, but who knows i might be abducted by aliens or something by then.
[B]Because it's not as easy as you're making it out to be.
If a developer makes improvements to a program, adds features, or rewrites the code from the groung up as they have to do after a while, it's a major job. As people migrate over to the new machines it will become less attractive to maintain an increasingly improved and complex program for both platforms (Mac platforms, that is). After a while the PPC software will become a problem child, as OS 9 programs have begun to be.
If after several years Apple suddenly announces a new PPC machine, that announcement will land with a thud. Developers will be angry and frustrated. If they have Windows products they may leave altogether. This has happened before. Developers are still leaving.[...]/B]
I agree that it's highly unlikely Apple will switch back, or offer its own line of PPC hardware as an alternative once Intel-based versions of all its hardware lines are on the market a couple of years down the road. However, I believe software support for PPC Macs will continue for years to come. Correct me if I'm wrong, but I don't believe Xcode in its current form even allows developers to create an Intel-only binary for Mac OS X; just a PPC binary or a PPC+Intel "fat" binary. So, in Xcode at least, maintaining support for legacy PPC hardware should be automatic.
Comments
Originally posted by sunilraman
What is the real reason then? Contracts with Freescale and IBM? Negative perception of Mac OS X? Deal not closed yet with Intel? hmmm intriguing
We may never know the full details, but just consider all the advantages of going with Intel that are just now coming to light: lower engineering and other development costs, better support for graphics chips and more graphics options, faster introduction of new technology, and just better commitment to the needs of personal computers rather than the embedded chip market and other specialties. I'm sure such items were a thorn in Apple's side for a long time. I think Intel's commitment to low power and the laptop platform in general was the clincher in this deal, however. Also the timing was right. I don't believe the move would have been as easy if it came earlier.
On the subject of keeping doors open, it would not surprise me if Apple keeps the door open for some PPC models of the Mac in the future. That should be easy to do since all code will now be written for two CPUs.
Originally posted by snoopy
On the subject of keeping doors open, it would not surprise me if Apple keeps the door open for some PPC models of the Mac in the future. That should be easy to do since all code will now be written for two CPUs.
Right, I was just thinking that we may eventually see both CPUs in the same line of products in the beginning, for example: in the powermac line: a low-end dual (core) G5, a high-end dual-dual-core G5 and a Intel PowerMac @ +3GHz (for the good cause). We may eventually see another form factor for the Intel-based powermacs because of the cooler CPUs, 19" rack or small towers. PPC and Intel may co-exist for a while... at least until the end of 2007.
Originally posted by mjteix
. . . PPC and Intel may co-exist for a while... at least until the end of 2007.
It's a fact that PPC and Intel will coexist into 2007, when the transition will finally be complete. I'm thinking it will be a total transition to Intel initially, so developers in all market segments completely convert their code and have applications running on both CPUs. Afterward, there should be no technical reason preventing Apple from reintroducing PPC models of the Mac. I could see where there might be good reasons for doing so, but I believe the majority of Apple's computer products will stay with Intel from now on.
Originally posted by snoopy
It's a fact that PPC and Intel will coexist into 2007, when the transition will finally be complete. I'm thinking it will be a total transition to Intel initially, so developers in all market segments completely convert their code and have applications running on both CPUs. Afterward, there should be no technical reason preventing Apple from reintroducing PPC models of the Mac. I could see where there might be good reasons for doing so, but I believe the majority of Apple's computer products will stay with Intel from now on.
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
Originally posted by melgross
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
Good question. In addition to the many advantages of Apple going with Intel, the decision was at least partly based on performance. But even Steve Jobs doesn't have a perfect Crystal Ball. IBM is going to continue their pursuit of better CPU design, and IBM is determined to beat Intel in the server game. Servers are IBM's bread and butter. The semiconductor devision was even merged into the server devision at one point. It is also clear that IBM will continue to look for areas where their technology can harvest more profit. Hence IBMs pursuit of the relatively large game console CPU market. Apple was beginning to be a burden to IBM it seems, and IBM didn't have the time and engineering resources to invest in keeping Apple satisfied. But all this should begin to change once the current rush of server and game console development is over.
Sooner of later IBM will have chips that Apple may be very interested in. Who knows what is coming after the first Cell and Trinity chips? IBM will not be standing still. If there is a juicy PPC processor that would work better than an Intel chip for some particular model of Mac, like a server, Apple would be foolish to not include it in their product line. It may not happen for maybe four years, but it's a good bet that it will happen.
I'm puzzled by your last comment about Apple developing PPC software for such a Mac. The software will be done. In a couple years, all applications will work on either a PPC or Intel Mac. There is no reason to suddenly drop the PPC versions of Mac OS and applications once the transition is complete.
Originally posted by melgross
What benefit would Apple get from spend all of that money to continue to develop PPC machines ans software? The market for it would be very small.
hi mel,
i think one of the good things that apple is encouraging is basic decent computer science. there is no reason to write low-level hardware-specific code unless you have bloody good reason to.
mac os X as a platform has a lot going for it, and i think what is happening is that apple is trying to say to everyone, well, steve did say, "this is the soul of the Mac". it doesn't matter if it wears a cross, a rosary in hand, or a turban, or a skullcap, whatever, it's still a Mac.
once all major and new developers get used to writing good hardware-abstracted universal binaries, and for example start using accelerate.framework more than specific code for altivec, or this or that spe or sse (not sure what the difference is, i'm typing this on something with MMX
...i think that then apple will get to pick and choose the hardware, which we know, hardware is becoming a real commodity game.
personally though i suspect that IBM has to come up with some hardcore serious shit to entice apple back into the PPC fold post 2007.
although, for that matter, if tomorrow some big corporation maybe out of china or india invented a laptop/dekstop platform that had some "smart" nano-material in it that could convert 50% of the waste heat of the CPU and components back into electrical energy into the battery, stevie J would be all over them like a bad rash.
he would encourage them to use Intels in that system though or sell the tech lock stock and barrel to Intel/Apple
...............
Originally posted by sunilraman
hi mel,
i think one of the good things that apple is encouraging is basic decent computer science. there is no reason to write low-level hardware-specific code unless you have bloody good reason to.
mac os X as a platform has a lot going for it, and i think what is happening is that apple is trying to say to everyone, well, steve did say, "this is the soul of the Mac". it doesn't matter if it wears a cross, a rosary in hand, or a turban, or a skullcap, whatever, it's still a Mac.
once all major and new developers get used to writing good hardware-abstracted universal binaries, and for example start using accelerate.framework more than specific code for altivec, or this or that spe or sse (not sure what the difference is, i'm typing this on something with MMX
...i think that then apple will get to pick and choose the hardware, which we know, hardware is becoming a real commodity game.
personally though i suspect that IBM has to come up with some hardcore serious shit to entice apple back into the PPC fold post 2007.
although, for that matter, if tomorrow some big corporation maybe out of china or india invented a laptop/dekstop platform that had some "smart" nano-material in it that could convert 50% of the waste heat of the CPU and components back into electrical energy into the battery, stevie J would be all over them like a bad rash.
he would encourage them to use Intels in that system though or sell the tech lock stock and barrel to Intel/Apple
...............
I was going to abbreviate your name, but all I came up with was SM, so I decided not to.
Well, I don't think that it will happen.
Firstly, if that were possible Jobs didn't have to make such a break. He didn't have to say that they would transition over to x86, but rather that they would be adding x86 machines to the line where they were needed. Remember the power/watt thing? The Powermacs and servers could have stayed with PPC and the rest made the move. But he didn't say that. ALL will be transitioned over.
There is also credibility. If he moves, even in part, back to PPC, Apple will really be looked upon with distaste by the developers.
Don't forget that this universal binary thingy is not easy. If a program is coded for the G5 and Altivec it will be difficult to re-encode. This is in the documents. It's a matter of maintaining two code bases. Who will want to do that? Apple will lose some developers as it is. They will lose more if they believe that they will have to continue two seperate development lines. Even with Cocoa and Xcode.
What would happen if they are told two or three years after the transition, when they are getting ready to slow developement down for PPC that they have to bring it back up? It won't be good.
Apple will also have to maintain two seperate hardware development lines. That's expensive.
I can believe that they will maintain the PPC code base for OS X the way they did with x86, but only in case disaster strikes.
ps. call me SR - that's my initials 8)
Originally posted by melgross
. . . Don't forget that this universal binary thingy is not easy. If a program is coded for the G5 and Altivec it will be difficult to re-encode. This is in the documents. It's a matter of maintaining two code bases.
What you say is true if the application was not developed with Apple tools. In anticipation of the move to Intel, Apple has been preparing its tools to make the transition easy. The only exception I know of is accelerate framework, which provides only the most common functions for AltiVec code. If a developer makes heavy use of AltiVec for operations outside of accelerate framework, he or she will have some work to do. Even so, it is no more work than the developer would do today if both Mac and Windows versions of the product are made.
I think the whole point of WWDC was to show developers how easy their task can be, using proper tools. I got a call from some developers just before Steve came on stage. They didn't believe the rumors and had a lot of time and energy invested in the PPC. After the keynote, their spirits were very low, but by the end of the conference most were even enthusiastic about the move to Intel. I can check back, but the last I heard it was not a big deal to maintain two Mac versions. My real concern at the time was that developers would drop PPC support too quickly, making the PPC Macs almost worthless.
Apple will also have to maintain two seperate hardware development lines. That's expensive.
IBM maintains a line of PPC and Intel servers, giving customers a choice of what they want. Using two different CPUs in hardware is no big deal. Each product is individually designed. The majority of Macs using an Intel CPU will be faster and easier to design because Intel provides complete chip sets. Apple may even use a few Intel motherboards on some products. Yet this will not make the design of a PPC Mac any more difficult than it is today. If a PPC chip is good enough to warrant inclusion in the Mac lineup, Apple could easily do it. And don't forget, this would be years away. With IBM's slow but sure push of the PPC, IBM may begin to make some supporting chips for systems designers.
Originally posted by snoopy
What you say is true if the application was not developed with Apple tools. In anticipation of the move to Intel, Apple has been preparing its tools to make the transition easy. The only exception I know of is accelerate framework, which provides only the most common functions for AltiVec code. If a developer makes heavy use of AltiVec for operations outside of accelerate framework, he or she will have some work to do. Even so, it is no more work than the developer would do today if both Mac and Windows versions of the product are made.
I think the whole point of WWDC was to show developers how easy their task can be, using proper tools. I got a call from some developers just before Steve came on stage. They didn't believe the rumors and had a lot of time and energy invested in the PPC. After the keynote, their spirits were very low, but by the end of the conference most were even enthusiastic about the move to Intel. I can check back, but the last I heard it was not a big deal to maintain two Mac versions. My real concern at the time was that developers would drop PPC support too quickly, making the PPC Macs almost worthless.
IBM maintains a line of PPC and Intel servers, giving customers a choice of what they want. Using two different CPUs in hardware is no big deal. Each product is individually designed. The majority of Macs using an Intel CPU will be faster and easier to design because Intel provides complete chip sets. Apple may even use a few Intel motherboards on some products. Yet this will not make the design of a PPC Mac any more difficult than it is today. If a PPC chip is good enough to warrant inclusion in the Mac lineup, Apple could easily do it. And don't forget, this would be years away. With IBM's slow but sure push of the PPC, IBM may begin to make some supporting chips for systems designers.
What I'm seeing is that it's difficult to transition Altivec, only some of it has an equivalent command. Some of the rest has to be painfully redone.
In many cases, code written for the G5 has to be hand tuned afterwards.
Insofar as maintaining a code line for Windows and OS X, you are now asking developers to also maintain a code line for PPC. That's three. If OS X had a much larger marketshare, fine. But it doesn't. It can't cost more to maintain code for OS X than it for for the much larger market that Windows represents.
As far as hardware lines go, IBM is a much larger company than Apple is, almost $100 billion vs 13 billion. Don't forget that IBM is also about equal with MS as far as software goes. They are much LARGER than MS is in every category except for OS and Office. All of these lines of computers they make come with hefty maintenance contracts as well, not just HW maintenance, but software to run a business as well.
Apple has nothing to compete, and has no interest in doing anything to make a business easer to run using their machines.
Apple's software sales will be a bit over $1 billion this year vs IBM's $50 billion. As software is much more profitable than HW, IBM can afford to cater to various customers desires in HW in order to sell their their business services. They are a service company. Apple is not.
Originally posted by melgross
What I'm seeing is that it's difficult to transition Altivec, only some of it has an equivalent command. Some of the rest has to be painfully redone. . .
This is true. The developer who call just before the keynote has a lot of AltiVec code that must be converted because there is nothing in Accelerate framework to do the job. However, when it is complete he will have that part of a Windows version of his application finished. So it is not "extra" work, but something he would have to do anyway. I'm not saying that keeping a PPC Mac version will be free, but it should be relatively easy. In the worst case scenario, some developer will simply not optimize the PPC code after a while and PPC performance will suffer. However, if Apple introduce another PPC model several years from now, I'm sure those doing mission critical code will put in the extra effort.
As far as hardware lines go, IBM is a much larger company than Apple is, almost $100 billion vs 13 billion.
My main point was that using two different CPUs in hardware is not a big deal. Each product is individually designed. A PPC will cost more to develop because IBM does not now make chip sets to go with it. That is why Apple would stay with Intel for the majority of Macs. However, if the PPC performance warrents it, and the market can bear the little extra cost burden of a PPC Mac, why not do it?
Originally posted by snoopy
This is true. The developer who call just before the keynote has a lot of AltiVec code that must be converted because there is nothing in Accelerate framework to do the job. However, when it is complete he will have that part of a Windows version of his application finished. So it is not "extra" work, but something he would have to do anyway. I'm not saying that keeping a PPC Mac version will be free, but it should be relatively easy. In the worst case scenario, some developer will simply not optimize the PPC code after a while and PPC performance will suffer. However, if Apple introduce another PPC model several years from now, I'm sure those doing mission critical code will put in the extra effort.
My main point was that using two different CPUs in hardware is not a big deal. Each product is individually designed. A PPC will cost more to develop because IBM does not now make chip sets to go with it. That is why Apple would stay with Intel for the majority of Macs. However, if the PPC performance warrents it, and the market can bear the little extra cost burden of a PPC Mac, why not do it?
Because it's not as easy as you're making it out to be.
If a developer makes improvements to a program, adds features, or rewrites the code from the groung up as they have to do after a while, it's a major job. As people migrate over to the new machines it will become less attractive to maintain an increasingly improved and complex program for both platforms (Mac platforms, that is). After a while the PPC software will become a problem child, as OS 9 programs have begun to be.
If after several years Apple suddenly announces a new PPC machine, that announcement will land with a thud. Developers will be angry and frustrated. If they have Windows products they may leave altogether. This has happened before. Developers are still leaving.
No, I hope that Apple never feels as though it has to do that.
As fas as hardware goes, it costs a fortune to develop support chips. If Apple is only going to use them on one line of machines they won't be able to make enough to get the cost down. Two series of mobo's with different architectures, etc. I don't think so. If IBM built controllers into the cpu as AMD does, it might be different. But IBM doesn't have the interest in the G5 for Apple's use. Apple bought about 75% of that chip's production. What will IBM do with it now> Apple will still be using it for another 2 years or so. After that - poof!
Originally posted by Hiro
What THT is saying is most general user apps have several threads, but not all of them are actively running at the same time. At any particular time most of the work is being done by one thread and that will run faster on a faster single processor than on a slower multi-processor box.
Yes, most of the web browsers, office apps and Internet apps, which consumers spend 99% of their time in, seem to have only 1 compute-bound thread while the others are waiting for input.
But if you try to do many things simultaneously, you would finish the collective tasks faster on a multi-CPU box, than a faster singler CPU box.
Well, it would depend on how much of a clock rate advantage the single CPU box has.
If Apple ever comes out with a quad box, I would think it would have to be a dual 2.5 GHz 970mp at the top end rather than 2 GHz. With the 2 GHz, I'm not too sure it would really provide enough of a performance improvement over the dual 2.7 970fx PowerMac G5.
Originally posted by THT
Yes, most of the web browsers, office apps and Internet apps, which consumers spend 99% of their time in, seem to have only 1 compute-bound thread while the others are waiting for input.
Well, it would depend on how much of a clock rate advantage the single CPU box has.
If Apple ever comes out with a quad box, I would think it would have to be a dual 2.5 GHz 970mp at the top end rather than 2 GHz. With the 2 GHz, I'm not too sure it would really provide enough of a performance improvement over the dual 2.7 970fx PowerMac G5.
Agreed.
Originally posted by melgross
. . . As fas as hardware goes, it costs a fortune to develop support chips. If Apple is only going to use them on one line of machines they won't be able to make enough to get the cost down. . . .
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
Originally posted by snoopy
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
It would cost significantly more to build chips for perhaps one quarter the number of machines. how much more would people pay for these machines?
When my old company (I was a partner in a Pro/hi end audio manufacturing concern) decided to come out with a new product, we had to determine what the purpose of this product was. How many we would sell, and what we could charge, based on whatever competing products were already out there, as well as possible new ones. All of these factors are interrelated.
Apple has the same problem.If other products are decreasing in cost, then Apple must also decrease the costs of its products.
I can't go into all of the complexities of the design and manufacturing process here, but every item adds cost in different ways. If the item is off the shelf, then the cost of the item might be doubled in the final product pricing.
If R&D is required for a part, then the cost is going to be much higher. The overall R&D costs have to be recovered. If it costs $50 million for the R&D, the resulting price has to be determined by simple math. How many products are to be assumed as being sold within a specific time to cover those costs? It also has to cover start up costs involved in production, etc.
If total costs are $100 million, and it's expected to sell 500 thousand units during that time, then it will add $200 to the final product's cost.
If other competetitors don't have those costs, then you are at a disadvantage.
Apple's costs are much higher because they already have costs that others don't. The R&D for the OS, for example. That cost is pretty much fixed. Whether Apple sells 3 million machines or 10 million, the cost to develop it remains the same. MS sells about 200 million copies a year, so even if their costs are much higher, it's less per copy.
To burden Apple with two different lines makes it more difficult for them to recover those costs. You can't just say that they can do it. You have to cost it out. Have you done that?
As far as IBM goes, you don't know what they will do. The 970 is a derivative of the Power, but it is not the Power. The Power chips are a chip set, and they are far more expensive than the G5's. Will they have a controllor on board? Maybe. But most likely not. Could IBM build such a chip with the thermal and pricing characteristics that Apple would need? Why should they?
If IBM finds that they don't need such a chip, they won't built it. Remember the R&D costs. How many of those chips would they have to sell? To what market would they be selling it? And what would be the guarantee that they would have any more success than they are having now in meeting Apple's needs?
Originally posted by snoopy
Yes it does. Yet this is exactly what Apple is doing today, providing the support chip at great expense. It has been said over and over that Apple's development costs are high compared to those who use Intel chip sets. If the Mac customer base is willing to pay the difference today for a Mac, why would they not be willing in the future too, if nothing changes? But IBM will make changes, and an on-chip memory controller would be just one such improvement that would appear in a Power 6 derivative. If you do not believe this will happen, I'd be interested in hearing your reasons.
if IBM offers Apple a realistic and deliverable 9xx Power-6 derivative that meets it needs, Apple would give it some thought. But performance/watt has gotta be high, yields have got to be promising (how it is promised i have no idea
Originally posted by sunilraman
but who knows
Well, I would hope so!
Originally posted by melgross
[B]Because it's not as easy as you're making it out to be.
If a developer makes improvements to a program, adds features, or rewrites the code from the groung up as they have to do after a while, it's a major job. As people migrate over to the new machines it will become less attractive to maintain an increasingly improved and complex program for both platforms (Mac platforms, that is). After a while the PPC software will become a problem child, as OS 9 programs have begun to be.
If after several years Apple suddenly announces a new PPC machine, that announcement will land with a thud. Developers will be angry and frustrated. If they have Windows products they may leave altogether. This has happened before. Developers are still leaving.[...]/B]
I agree that it's highly unlikely Apple will switch back, or offer its own line of PPC hardware as an alternative once Intel-based versions of all its hardware lines are on the market a couple of years down the road. However, I believe software support for PPC Macs will continue for years to come. Correct me if I'm wrong, but I don't believe Xcode in its current form even allows developers to create an Intel-only binary for Mac OS X; just a PPC binary or a PPC+Intel "fat" binary. So, in Xcode at least, maintaining support for legacy PPC hardware should be automatic.