I totally believe Jobs on this one when he said that "We are through with Power."
There is no reason for Apple to ever look back on this.
The POWER 6 chipset is a several thousand dollar design. Apple would have to have at least a $10,000, or more, machine, to accommodate it.
Keeping the software alive for that would be a waste.
They kept it alive for x86 for a very good reason. They saw that Moto, then later, Freescale and IBM, simply weren't interested in advancing the lines in a useful way for Apple.
Once Intel showed their future pathway, Apple jumped.
There is simply no way that they will jump back.
Freescale hasn't made a desktop processor for years, and IBM has scaled back PPC development.
Even if Apple were to buy 20 million cpu's a year from IBM, it still wouldn't be enough to pay for the R&D costs.
IBM is now concentrating their chip producing energies towards producing chips for the game machines. Apple's needs simply can't compete with that.
The POWER family is a high performance, heavy iron chipset that's produced in low quantities. Even this new efficient design uses much more power than two 4 core x86 cpu's.
This is a non-event as far as Apple is concerned, and no reason for them to keep the PPC OS, which would have to be re-compiled, and optimized, to run on it.
At the moment Apple are content with only being as good as the mainstream consumer competition performance-wise but there's no reason to not keep their options open.
Didn't we go through that somewhere else just now?
Quote:
At the moment Apple are content with only being as good as the mainstream consumer competition performance-wise but there's no reason to not keep their options open.
Sure there is. How many developers do you think will be willing to go through this mess of moving to a new chip?again?
Moving from PPC to x86 made sense, and was an easy choice for most developers. They could visualize Apple's sales going up much more with Intel chips. But, I don't see any way that Apple could ever convince them to go back the the PPC, or POWER.
Apple simply doesn't sell enough machines even now for that to happen. If they come out with one machine, abet a high end one, using either PPC or POWER, Apple could never sell enough of them for any developer to spend the time and money to write to it.
Sure there is. How many developers do you think will be willing to go through this mess of moving to a new chip?again?
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
Apple have then gone on to add a third architecture - ARM - so I'd hope developers have learned their lesson by now that coding to a specific architecture and not keeping up with Apple's development is going to get you in to trouble later on.
I don't think the changes for developers are over yet either with Apple's interest in LLVM. It's pretty clear that they're looking to support OSX on multiple architectures and aren't putting all their eggs into one architecture. LLVM solves some of that problem.
POWER6 is actually a die shrink and mass production away from being a <$500 CPU, ***if*** there was the demand. Sure, you'd drop the L3 cache, and the chipset would be more mass market (you probably use a variant of HyperTransport to use commodity chipsets) so you'd lose some performance, but overall it would still be awesome. Not amazingly hot on single threaded apps, but they're getting less common.
Hell, the die size and power consumption is comparable to the AMD R600, and that's in $500 graphics cards.
IBM have 6GHz POWER6 operating in the lab. It's likely they'll release a 5GHz+ variant by the end of the year (the initial launch is in June for the 3.5GHz - 4.7GHz machines). POWER6+ on 45nm in 2009 will be 6GHz or above...
It's great that someone is pushing the clock speed envelope finally, and still achieving decent IPC to boot.
It will be interesting to see how process knowledge from IBM will aid AMD's CPUs given their partnership. AMD have been rumoured to be sandbagging on Barcelona clock speeds, but it's at 45nm that things get interesting.
Apple should just provide an OEM version of OS X Server with Power6 hardware. That way they get into the high end server space, leverage OS X processor agnostic nature, and not strain their relationship with Intel or waste money on dual hardware support.
The high-end server market is mostly one of services. Apple's not just going to jump into that market. It would be more practical for them to license OS X to IBM's business unit, but that's not going to happen either.
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
Apple have then gone on to add a third architecture - ARM - so I'd hope developers have learned their lesson by now that coding to a specific architecture and not keeping up with Apple's development is going to get you in to trouble later on.
I don't think the changes for developers are over yet either with Apple's interest in LLVM. It's pretty clear that they're looking to support OSX on multiple architectures and aren't putting all their eggs into one architecture. LLVM solves some of that problem.
But they did have to move to a new chip architecture as well. Coding for one is better because they need only one programming team. That's much cheaper. Even though they have X Code, they still have to optimize for each chip, if they care. If they don't, then the PPC versions will be increasingly out of touch. Considering that Apple's old PPC machines won't have the hardware for new features, that will be harder to do as time goes on. At some point in time, these developers will simply stop coding for the PPC version.
If Apple then does come back with a PPC machine, it will be too late. My prediction is that most large programs will see one, or at most two more major upgrades for the PPC platform. After that, their PPC development teams will be broken up, never again to put out another product, though a small team might remain for a while to fix any last bugs. That's the way it went before, and the way it will go again.
I doubt very much if you asked Adobe, or any other large Mac developer, whether they will keep those PPC teams together, they would say yes.
Apple's ARM products should, if everything goes well, have a rather large customer base, I'm sure you would agree. That will make development worthwhile.
LLVM does look interesting, but it would still require separate teams. What must be remembered is that Intel, as I mentioned before, is implementing more and more features within it's cpu's and chipsets. While Apple may not be using all of them right off the bat, they will use more of them as time goes on. As the PPC has no equivalent instructions, or features, it will be much more difficult to keep PPC machines concurrent with newer x86 models. At some point, it will become impossible.
POWER6 is actually a die shrink and mass production away from being a <$500 CPU, ***if*** there was the demand. Sure, you'd drop the L3 cache, and the chipset would be more mass market (you probably use a variant of HyperTransport to use commodity chipsets) so you'd lose some performance, but overall it would still be awesome. Not amazingly hot on single threaded apps, but they're getting less common.
Hell, the die size and power consumption is comparable to the AMD R600, and that's in $500 graphics cards.
IBM have 6GHz POWER6 operating in the lab. It's likely they'll release a 5GHz+ variant by the end of the year (the initial launch is in June for the 3.5GHz - 4.7GHz machines). POWER6+ on 45nm in 2009 will be 6GHz or above...
It's great that someone is pushing the clock speed envelope finally, and still achieving decent IPC to boot.
It will be interesting to see how process knowledge from IBM will aid AMD's CPUs given their partnership. AMD have been rumoured to be sandbagging on Barcelona clock speeds, but it's at 45nm that things get interesting.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
But they did have to move to a new chip architecture as well. Coding for one is better because they need only one programming team. That's much cheaper. Even though they have X Code, they still have to optimize for each chip, if they care. If they don't, then the PPC versions will be increasingly out of touch. Considering that Apple's old PPC machines won't have the hardware for new features, that will be harder to do as time goes on. At some point in time, these developers will simply stop coding for the PPC version.
The above is entirely in the domain of Apple to support though obviously that doesn't stop dumb-ass developers from implementing their own architecture specific code. Apple provides frameworks to isolate developers from the hardware. It's up to the developers to use them.
Quote:
Originally Posted by melgross
If Apple then does come back with a PPC machine, it will be too late. My prediction is that most large programs will see one, or at most two more major upgrades for the PPC platform. After that, their PPC development teams will be broken up, never again to put out another product, though a small team might remain for a while to fix any last bugs. That's the way it went before, and the way it will go again.
I wouldn't disagree with that prediction for some of the developers out there - I'm looking at you Adobe, Microsoft...
Quote:
Originally Posted by melgross
I doubt very much if you asked Adobe, or any other large Mac developer, whether they will keep those PPC teams together, they would say yes.
I think I'd be asking them why they were STILL writing their own architecture specific code. It's bitten both Adobe and Microsoft in the past and they don't seem to have learned yet. eg. not being able to carry VBA into Intel because of their needlessly stupid compiler.
Quote:
Originally Posted by melgross
Apple's ARM products should, if everything goes well, have a rather large customer base, I'm sure you would agree. That will make development worthwhile.
LLVM does look interesting, but it would still require separate teams.
At the moment, yes. Eventually, probably not. I'm sure Apple will want to get it down to the 'just tick this box in XCode' level.
Quote:
Originally Posted by melgross
What must be remembered is that Intel, as I mentioned before, is implementing more and more features within it's cpu's and chipsets. While Apple may not be using all of them right off the bat, they will use more of them as time goes on. As the PPC has no equivalent instructions, or features, it will be much more difficult to keep PPC machines concurrent with newer x86 models. At some point, it will become impossible.
Really? Just what features do you mean that affect developers? Most of the changes from PPC to Intel have been downgrades in features like AltiVec to SSE, 64bit to 32bit etc which Intel are gradually correcting, but those only affected people if you were rolling your own CPU specific code and not using VecLib or other Apple provided frameworks.
If you're a true Mac developer, coding in Cocoa, you're mostly isolated from hardware issues which is why those developers were able to get universal binaries out weeks after WWDC 2005 and not years after.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
I don't think that's true anymore. The first POWER implementation in the RS/6000 was multi-chip but subsequent implementations have been single chip.
Really? Just what features do you mean that affect developers? Most of the changes from PPC to Intel have been downgrades in features like AltiVec to SSE, 64bit to 32bit etc which Intel are gradually correcting, but those only affected people if you were rolling your own CPU specific code and not using VecLib or other Apple provided frameworks.
If you're a true Mac developer, coding in Cocoa, you're mostly isolated from hardware issues which is why those developers were able to get universal binaries out weeks after WWDC 2005 and not years after.
I'm sure that at some point Apple will be supporting features such as Intels' built-in WiMax, and others. SSE 4 takes no back seat to Alti-Vec. What would have been the promise that IBM would have made strides forward with Alti-Vec? Intel is moving with SSE.
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
That would be nice, if that theory reflected reality. But there are some major publishers (like Adobe) that have already released Intel-only products, along with official statements that there will be no PPC versions.
Quote:
Originally Posted by aegisdesign
Apple have then gone on to add a third architecture - ARM
The iPhone is a closed platform. Nobody is going to be developing a cross-platform app that runs on Macs and iPhones. Even if you could get the code to port, the UI would need to be so radically different that it would undermine the entire point.
Sure, Apple is porting OS X to a lot of platforms, but that's a far cry from supporting Mac apps on all those platforms.
I don't think that's true anymore. The first POWER implementation in the RS/6000 was multi-chip but subsequent implementations have been single chip.
I can't find anything about that particular aspect, at this moment, but the main chip itself is about 341mm2 in size, and uses 100 watts in LOW power mode! Somewhere, I remember reading that at full speed, the 2 core chip (there is no 4 core version) will use as much as 250 watts. And this is using 65 nm. The POWER line has always been one process generation behind for the purpose of robustness.
I'm sure that at some point Apple will be supporting features such as Intels' built-in WiMax, and others. SSE 4 takes no back seat to Alti-Vec. What would have been the promise that IBM would have made strides forward with Alti-Vec? Intel is moving with SSE.
IBM of course HAVE moved forward with Altivec. They moved forward with ViVA and ViVA2 where they can use multiple cores as one huge incredibly fast vector engine.
You could argue they've also moved forward with Cell SBEs.
Intel have nothing like either.
Quote:
Originally Posted by melgross
Cocoa can isolate from hardware "issues", but not from the need to optimise.
There's very few operations that need optimization and generally sticking with Apple's already optimized frameworks is enough and shields you from architecture changes. If you're Adobe or Microsoft though, writing cross platform software, you may not have that luxury as your code may have to use your own frameworks. Just look at the horribly cludgy CS3 UI that tries to look like Aqua and you'll see what sort of mess that method of development gives us.
Then of course there's LLVM whose whole mission is to provide cross platform optimizing compiler technology so that developers don't have to. It's early still with that though.
Quote:
Originally Posted by shamino
That would be nice, if that theory reflected reality. But there are some major publishers (like Adobe) that have already released Intel-only products, along with official statements that there will be no PPC versions.
That's Adobe being lazy and seeing a quick buck in quickly porting a Windows product back to the Mac that they left to go stale some years ago in the OS9->OSX transition. To get it out quick they've slapped in Windows x86 code.
Joost I think is the only other Intel only product I've come across and they were saying a PPC version will be produced eventually.
Quote:
Originally Posted by shamino
The iPhone is a closed platform. Nobody is going to be developing a cross-platform app that runs on Macs and iPhones. Even if you could get the code to port, the UI would need to be so radically different that it would undermine the entire point.
Sure, Apple is porting OS X to a lot of platforms, but that's a far cry from supporting Mac apps on all those platforms.
You've got to develop the software somewhere so building cross platform development tools is par for the course. I'm pretty sure they don't want developers buying PCs for ARM development and I really don't believe Apple will keep the iPhone platform closed or that the iPhone is the entire platform they'll be using OSX on ARM for.
They just wouldn't be so publicly supporting the LLVM project with tool development if the iPhone platform was closed. They'd just be using Windows in house with Codewarrior.
Anyway, all I'm saying is that it'd be a little silly for Apple, with a mature cross platform architecture, to burn it's bridges entirely.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
Take a look at the entry level hardware: "The p5-505 Express is a 1- or 2-core server with 1.9 or 2.1 GHz POWER5+ processor cores." In this configuration IBM is using a Dual-Chip Module: 1 dual-core POWER5, the external L3 is optional. Several diagrams and pictures of POWER5 modules are shown here.
The POWER6 will replace the PPC970 in IBM's thin blade servers: "Today, IBM's blade servers are available with the company's PowerPC 970 processors. But the Power6 will replace those lower-end sibling in blade servers, Tim Doughtery, IBM's BladeCenter strategist, said in an interview. [?] Dougherty wouldn't say the PowerPC line has no future, but with AltiVec now incorporated into Power6, "I don't see it right now," he said."
Quote:
Originally Posted by melgross
the main chip itself is about 341mm2 in size, and uses 100 watts in LOW power mode!
The 341mm2 die is huge for a 65nm chip but there are 2 on-chip memory controllers plus an on-chip L3 controller and 8MB of L2 cache (diagram). Contrary to the POWER4 and 5, each core includes an Altivec unit.
The dual-core PPC 970MP was only 154mm2 with 2x1MB of L2 cache and of course it would be even smaller at 65nm. If Apple was committed to the PowerPC architecture they would need some kind of POWER6 derivative. The POWER4 was huge too, yet IBM produced a desktop class chip with the PPC 970. I'm not convinced that a POWER6-lite would fit in a notebook, though. It's far easier to switch to Intel, all the chips that Apple needs (from notebooks to servers) are already in mass production, it's a "one stop shop."
Take a look at the entry level hardware: "The p5-505 Express is a 1- or 2-core server with 1.9 or 2.1 GHz POWER5+ processor cores." In this configuration IBM is using a Dual-Chip Module: 1 dual-core POWER5, the external L3 is optional. Several diagrams and pictures of POWER5 modules are shown here.
The POWER6 will replace the PPC970 in IBM's thin blade servers: "Today, IBM's blade servers are available with the company's PowerPC 970 processors. But the Power6 will replace those lower-end sibling in blade servers, Tim Doughtery, IBM's BladeCenter strategist, said in an interview. [?] Dougherty wouldn't say the PowerPC line has no future, but with AltiVec now incorporated into Power6, "I don't see it right now," he said."
The 341mm2 die is huge for a 65nm chip but there are 2 on-chip memory controllers plus an on-chip L3 controller and 8MB of L2 cache (diagram). Contrary to the POWER4 and 5, each core includes an Altivec unit.
The dual-core PPC 970MP was only 154mm2 with 2x1MB of L2 cache and of course it would be even smaller at 65nm. If Apple was committed to the PowerPC architecture they would need some kind of POWER6 derivative. The POWER4 was huge too, yet IBM produced a desktop class chip with the PPC 970. I'm not convinced that a POWER6-lite would fit in a notebook, though. It's far easier to switch to Intel, all the chips that Apple needs (from notebooks to servers) are already in mass production, it's a "one stop shop."
It's good that you l,ooked all of that up. POWER 6 is too much for Apple.
I'm not surprised at the likely demise of the 970. IBM is abandoning the low end for good, it seems.
That's Adobe being lazy and seeing a quick buck in quickly porting a Windows product back to the Mac that they left to go stale some years ago in the OS9->OSX transition. To get it out quick they've slapped in Windows x86 code.
Of course that's the case. That's not the point.
Adobe is a major player in the Mac software market. People alter hardware purchasing decisions based on the availability of Adobe products.
Apple can't go around changing processors, telling developers to "just recompile" when there are companies like Adobe that routinely respond "no" or "only when we're good and ready."
If Apple would release a new PPC system, some of the die-hard Mac users would appreciate it, many more would not. They'd respond "we just spent all this time and money upgrading our apps and now it was all for nothing?" They'd say "we just finished migrating to all-Intel to take advantage of the newest Adobe apps and now the next generation won't run any of them?"
Mac OS isn't even close to big enough to be able to ship workstations with multiple processor architectures, and keep them all properly supported at once. And I doubt the big software vendors will want to play ball either.
Quote:
Originally Posted by aegisdesign
... I really don't believe Apple will keep the iPhone platform closed or that the iPhone is the entire platform they'll be using OSX on ARM for.
So how is that iPod software development kit coming along? What you don't have one? Why not? Oh that's right, it's still a closed platform after all these years.
What makes you think the iPhone will be any different?
Unless you have access to some confidential Apple information, your belief is nothing more than wishful thinking.
If Apple would release a new PPC system, some of the die-hard Mac users would appreciate it, many more would not. They'd respond "we just spent all this time and money upgrading our apps and now it was all for nothing?" They'd say "we just finished migrating to all-Intel to take advantage of the newest Adobe apps and now the next generation won't run any of them?"
The latest Adobe apps are Universal binaries, not Intel only. There's no reason why the apps would stop working if Apple just came out with another PPC Mac.
Quote:
Originally Posted by shamino
Mac OS isn't even close to big enough to be able to ship workstations with multiple processor architectures, and keep them all properly supported at once. And I doubt the big software vendors will want to play ball either.
But that's exactly what Apple are doing now or did I miss a memo where Leopard wasn't supported on PowerPC?
Quote:
Originally Posted by shamino
So how is that iPod software development kit coming along? What you don't have one? Why not? Oh that's right, it's still a closed platform after all these years.
If you're the right person Apple will give you an iPod SDK.
Quote:
Originally Posted by shamino
What makes you think the iPhone will be any different?
Because the iPhone isn't an iPod. It's a much broader, more computer like environment than an mp3 player. Every smartphone platform going relies to some extent on 3rd party developers. Sure, Apple could do it all themselves but that seems just too stupid for words.
The beauty of it would be that future games are optimized for it. The consoles remain static and the only way that games work well is that they are written for it - you'd probably have to do this to make any return. For games that are not optimized specifically for it, the fact that it would be a generic PC gives you a good level of compatibility.
Concerning updates, it would last at least 3 years. The GeForce 6600GT is still a decent card and it's 3 years old. Also, newer games like Double Agent, although they require DirectX shader model 3 and things, a version has been written for the PS2, which performs about the same as a Mac Mini.
Wow. 3 years old? ...Yeah the 6600GT is *awesome*. Best graphics card I've ever owned. (And this goes back to the halcyon days of Matrox).
The latest Adobe apps are Universal binaries, not Intel only. There's no reason why the apps would stop working if Apple just came out with another PPC Mac.
But that's exactly what Apple are doing now or did I miss a memo where Leopard wasn't supported on PowerPC?
Aegis, you know that that's all true now.
We aren't denying that it's true now.
I'm saying, for example, and so are others, that Apple will come out with a 10.6 for PPC. That will happen about two and a half to three years from now.
After that introduction, Apple will keep it current with 10.6 for x86 until 10.7 comes out, another two and a half years, or so, later. The likelihood that Apple will have a 10.7 PPC version is extremely slight, almost non existent. That will be five to five and a half years from now though. That's what we're saying.
You don't really think that Apple will produce another PPC machine during that time do you? It seems that the PPC is going away anyway soon.
How many PPC machines will be left that long from now? Very few, I'm sure.
Comments
It's NEVER going to happen!
I totally believe Jobs on this one when he said that "We are through with Power."
There is no reason for Apple to ever look back on this.
The POWER 6 chipset is a several thousand dollar design. Apple would have to have at least a $10,000, or more, machine, to accommodate it.
Keeping the software alive for that would be a waste.
They kept it alive for x86 for a very good reason. They saw that Moto, then later, Freescale and IBM, simply weren't interested in advancing the lines in a useful way for Apple.
Once Intel showed their future pathway, Apple jumped.
There is simply no way that they will jump back.
Freescale hasn't made a desktop processor for years, and IBM has scaled back PPC development.
Even if Apple were to buy 20 million cpu's a year from IBM, it still wouldn't be enough to pay for the R&D costs.
IBM is now concentrating their chip producing energies towards producing chips for the game machines. Apple's needs simply can't compete with that.
The POWER family is a high performance, heavy iron chipset that's produced in low quantities. Even this new efficient design uses much more power than two 4 core x86 cpu's.
This is a non-event as far as Apple is concerned, and no reason for them to keep the PPC OS, which would have to be re-compiled, and optimized, to run on it.
It's NEVER going to happen!
Never say Never.
At the moment Apple are content with only being as good as the mainstream consumer competition performance-wise but there's no reason to not keep their options open.
Never say Never.
At the moment Apple are content with only being as good as the mainstream consumer competition performance-wise but there's no reason to not keep their options open.
Sure there is. How many developers do you think will be willing to go through this mess of moving to a new chip?again?
Moving from PPC to x86 made sense, and was an easy choice for most developers. They could visualize Apple's sales going up much more with Intel chips. But, I don't see any way that Apple could ever convince them to go back the the PPC, or POWER.
Apple simply doesn't sell enough machines even now for that to happen. If they come out with one machine, abet a high end one, using either PPC or POWER, Apple could never sell enough of them for any developer to spend the time and money to write to it.
Sure there is. How many developers do you think will be willing to go through this mess of moving to a new chip?again?
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
Apple have then gone on to add a third architecture - ARM - so I'd hope developers have learned their lesson by now that coding to a specific architecture and not keeping up with Apple's development is going to get you in to trouble later on.
I don't think the changes for developers are over yet either with Apple's interest in LLVM. It's pretty clear that they're looking to support OSX on multiple architectures and aren't putting all their eggs into one architecture. LLVM solves some of that problem.
Hell, the die size and power consumption is comparable to the AMD R600, and that's in $500 graphics cards.
IBM have 6GHz POWER6 operating in the lab. It's likely they'll release a 5GHz+ variant by the end of the year (the initial launch is in June for the 3.5GHz - 4.7GHz machines). POWER6+ on 45nm in 2009 will be 6GHz or above...
It's great that someone is pushing the clock speed envelope finally, and still achieving decent IPC to boot.
It will be interesting to see how process knowledge from IBM will aid AMD's CPUs given their partnership. AMD have been rumoured to be sandbagging on Barcelona clock speeds, but it's at 45nm that things get interesting.
Apple should just provide an OEM version of OS X Server with Power6 hardware. That way they get into the high end server space, leverage OS X processor agnostic nature, and not strain their relationship with Intel or waste money on dual hardware support.
The high-end server market is mostly one of services. Apple's not just going to jump into that market. It would be more practical for them to license OS X to IBM's business unit, but that's not going to happen either.
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
Apple have then gone on to add a third architecture - ARM - so I'd hope developers have learned their lesson by now that coding to a specific architecture and not keeping up with Apple's development is going to get you in to trouble later on.
I don't think the changes for developers are over yet either with Apple's interest in LLVM. It's pretty clear that they're looking to support OSX on multiple architectures and aren't putting all their eggs into one architecture. LLVM solves some of that problem.
But they did have to move to a new chip architecture as well. Coding for one is better because they need only one programming team. That's much cheaper. Even though they have X Code, they still have to optimize for each chip, if they care. If they don't, then the PPC versions will be increasingly out of touch. Considering that Apple's old PPC machines won't have the hardware for new features, that will be harder to do as time goes on. At some point in time, these developers will simply stop coding for the PPC version.
If Apple then does come back with a PPC machine, it will be too late. My prediction is that most large programs will see one, or at most two more major upgrades for the PPC platform. After that, their PPC development teams will be broken up, never again to put out another product, though a small team might remain for a while to fix any last bugs. That's the way it went before, and the way it will go again.
I doubt very much if you asked Adobe, or any other large Mac developer, whether they will keep those PPC teams together, they would say yes.
Apple's ARM products should, if everything goes well, have a rather large customer base, I'm sure you would agree. That will make development worthwhile.
LLVM does look interesting, but it would still require separate teams. What must be remembered is that Intel, as I mentioned before, is implementing more and more features within it's cpu's and chipsets. While Apple may not be using all of them right off the bat, they will use more of them as time goes on. As the PPC has no equivalent instructions, or features, it will be much more difficult to keep PPC machines concurrent with newer x86 models. At some point, it will become impossible.
POWER6 is actually a die shrink and mass production away from being a <$500 CPU, ***if*** there was the demand. Sure, you'd drop the L3 cache, and the chipset would be more mass market (you probably use a variant of HyperTransport to use commodity chipsets) so you'd lose some performance, but overall it would still be awesome. Not amazingly hot on single threaded apps, but they're getting less common.
Hell, the die size and power consumption is comparable to the AMD R600, and that's in $500 graphics cards.
IBM have 6GHz POWER6 operating in the lab. It's likely they'll release a 5GHz+ variant by the end of the year (the initial launch is in June for the 3.5GHz - 4.7GHz machines). POWER6+ on 45nm in 2009 will be 6GHz or above...
It's great that someone is pushing the clock speed envelope finally, and still achieving decent IPC to boot.
It will be interesting to see how process knowledge from IBM will aid AMD's CPUs given their partnership. AMD have been rumoured to be sandbagging on Barcelona clock speeds, but it's at 45nm that things get interesting.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
But they did have to move to a new chip architecture as well. Coding for one is better because they need only one programming team. That's much cheaper. Even though they have X Code, they still have to optimize for each chip, if they care. If they don't, then the PPC versions will be increasingly out of touch. Considering that Apple's old PPC machines won't have the hardware for new features, that will be harder to do as time goes on. At some point in time, these developers will simply stop coding for the PPC version.
The above is entirely in the domain of Apple to support though obviously that doesn't stop dumb-ass developers from implementing their own architecture specific code. Apple provides frameworks to isolate developers from the hardware. It's up to the developers to use them.
If Apple then does come back with a PPC machine, it will be too late. My prediction is that most large programs will see one, or at most two more major upgrades for the PPC platform. After that, their PPC development teams will be broken up, never again to put out another product, though a small team might remain for a while to fix any last bugs. That's the way it went before, and the way it will go again.
I wouldn't disagree with that prediction for some of the developers out there - I'm looking at you Adobe, Microsoft...
I doubt very much if you asked Adobe, or any other large Mac developer, whether they will keep those PPC teams together, they would say yes.
I think I'd be asking them why they were STILL writing their own architecture specific code. It's bitten both Adobe and Microsoft in the past and they don't seem to have learned yet. eg. not being able to carry VBA into Intel because of their needlessly stupid compiler.
Apple's ARM products should, if everything goes well, have a rather large customer base, I'm sure you would agree. That will make development worthwhile.
LLVM does look interesting, but it would still require separate teams.
At the moment, yes. Eventually, probably not. I'm sure Apple will want to get it down to the 'just tick this box in XCode' level.
What must be remembered is that Intel, as I mentioned before, is implementing more and more features within it's cpu's and chipsets. While Apple may not be using all of them right off the bat, they will use more of them as time goes on. As the PPC has no equivalent instructions, or features, it will be much more difficult to keep PPC machines concurrent with newer x86 models. At some point, it will become impossible.
Really? Just what features do you mean that affect developers? Most of the changes from PPC to Intel have been downgrades in features like AltiVec to SSE, 64bit to 32bit etc which Intel are gradually correcting, but those only affected people if you were rolling your own CPU specific code and not using VecLib or other Apple provided frameworks.
If you're a true Mac developer, coding in Cocoa, you're mostly isolated from hardware issues which is why those developers were able to get universal binaries out weeks after WWDC 2005 and not years after.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
I don't think that's true anymore. The first POWER implementation in the RS/6000 was multi-chip but subsequent implementations have been single chip.
Really? Just what features do you mean that affect developers? Most of the changes from PPC to Intel have been downgrades in features like AltiVec to SSE, 64bit to 32bit etc which Intel are gradually correcting, but those only affected people if you were rolling your own CPU specific code and not using VecLib or other Apple provided frameworks.
If you're a true Mac developer, coding in Cocoa, you're mostly isolated from hardware issues which is why those developers were able to get universal binaries out weeks after WWDC 2005 and not years after.
I'm sure that at some point Apple will be supporting features such as Intels' built-in WiMax, and others. SSE 4 takes no back seat to Alti-Vec. What would have been the promise that IBM would have made strides forward with Alti-Vec? Intel is moving with SSE.
http://www.reghardware.co.uk/2006/09...sa_extensions/
Cocoa can isolate from hardware "issues", but not from the need to optimise.
The last big change wasn't really about changing the chip architecture, it was changing the development tool chain from Codewarrior to XCode. Apple wanted everyone on XCode. That was the major pain for some.
Developers have got to support PowerPC still anyway so they've not moved to a new chip at all, just added one.
That would be nice, if that theory reflected reality. But there are some major publishers (like Adobe) that have already released Intel-only products, along with official statements that there will be no PPC versions.
Apple have then gone on to add a third architecture - ARM
The iPhone is a closed platform. Nobody is going to be developing a cross-platform app that runs on Macs and iPhones. Even if you could get the code to port, the UI would need to be so radically different that it would undermine the entire point.
Sure, Apple is porting OS X to a lot of platforms, but that's a far cry from supporting Mac apps on all those platforms.
I don't think that's true anymore. The first POWER implementation in the RS/6000 was multi-chip but subsequent implementations have been single chip.
I can't find anything about that particular aspect, at this moment, but the main chip itself is about 341mm2 in size, and uses 100 watts in LOW power mode! Somewhere, I remember reading that at full speed, the 2 core chip (there is no 4 core version) will use as much as 250 watts. And this is using 65 nm. The POWER line has always been one process generation behind for the purpose of robustness.
I'm sure that at some point Apple will be supporting features such as Intels' built-in WiMax, and others. SSE 4 takes no back seat to Alti-Vec. What would have been the promise that IBM would have made strides forward with Alti-Vec? Intel is moving with SSE.
http://www.reghardware.co.uk/2006/09...sa_extensions/
That's called playing catchup with Altivec.
IBM of course HAVE moved forward with Altivec. They moved forward with ViVA and ViVA2 where they can use multiple cores as one huge incredibly fast vector engine.
You could argue they've also moved forward with Cell SBEs.
Intel have nothing like either.
Cocoa can isolate from hardware "issues", but not from the need to optimise.
There's very few operations that need optimization and generally sticking with Apple's already optimized frameworks is enough and shields you from architecture changes. If you're Adobe or Microsoft though, writing cross platform software, you may not have that luxury as your code may have to use your own frameworks. Just look at the horribly cludgy CS3 UI that tries to look like Aqua and you'll see what sort of mess that method of development gives us.
Then of course there's LLVM whose whole mission is to provide cross platform optimizing compiler technology so that developers don't have to. It's early still with that though.
That would be nice, if that theory reflected reality. But there are some major publishers (like Adobe) that have already released Intel-only products, along with official statements that there will be no PPC versions.
That's Adobe being lazy and seeing a quick buck in quickly porting a Windows product back to the Mac that they left to go stale some years ago in the OS9->OSX transition. To get it out quick they've slapped in Windows x86 code.
Joost I think is the only other Intel only product I've come across and they were saying a PPC version will be produced eventually.
The iPhone is a closed platform. Nobody is going to be developing a cross-platform app that runs on Macs and iPhones. Even if you could get the code to port, the UI would need to be so radically different that it would undermine the entire point.
Sure, Apple is porting OS X to a lot of platforms, but that's a far cry from supporting Mac apps on all those platforms.
You've got to develop the software somewhere so building cross platform development tools is par for the course. I'm pretty sure they don't want developers buying PCs for ARM development and I really don't believe Apple will keep the iPhone platform closed or that the iPhone is the entire platform they'll be using OSX on ARM for.
They just wouldn't be so publicly supporting the LLVM project with tool development if the iPhone platform was closed. They'd just be using Windows in house with Codewarrior.
Anyway, all I'm saying is that it'd be a little silly for Apple, with a mature cross platform architecture, to burn it's bridges entirely.
I doubt that very much. POWER is not one chip. Look it up. It's not even a chip, it requires far more support. PPC was POWER's $500 chip, and will remain so.
Take a look at the entry level hardware: "The p5-505 Express is a 1- or 2-core server with 1.9 or 2.1 GHz POWER5+ processor cores." In this configuration IBM is using a Dual-Chip Module: 1 dual-core POWER5, the external L3 is optional. Several diagrams and pictures of POWER5 modules are shown here.
The POWER6 will replace the PPC970 in IBM's thin blade servers: "Today, IBM's blade servers are available with the company's PowerPC 970 processors. But the Power6 will replace those lower-end sibling in blade servers, Tim Doughtery, IBM's BladeCenter strategist, said in an interview. [?] Dougherty wouldn't say the PowerPC line has no future, but with AltiVec now incorporated into Power6, "I don't see it right now," he said."
the main chip itself is about 341mm2 in size, and uses 100 watts in LOW power mode!
The 341mm2 die is huge for a 65nm chip but there are 2 on-chip memory controllers plus an on-chip L3 controller and 8MB of L2 cache (diagram). Contrary to the POWER4 and 5, each core includes an Altivec unit.
The dual-core PPC 970MP was only 154mm2 with 2x1MB of L2 cache and of course it would be even smaller at 65nm. If Apple was committed to the PowerPC architecture they would need some kind of POWER6 derivative. The POWER4 was huge too, yet IBM produced a desktop class chip with the PPC 970. I'm not convinced that a POWER6-lite would fit in a notebook, though. It's far easier to switch to Intel, all the chips that Apple needs (from notebooks to servers) are already in mass production, it's a "one stop shop."
Take a look at the entry level hardware: "The p5-505 Express is a 1- or 2-core server with 1.9 or 2.1 GHz POWER5+ processor cores." In this configuration IBM is using a Dual-Chip Module: 1 dual-core POWER5, the external L3 is optional. Several diagrams and pictures of POWER5 modules are shown here.
The POWER6 will replace the PPC970 in IBM's thin blade servers: "Today, IBM's blade servers are available with the company's PowerPC 970 processors. But the Power6 will replace those lower-end sibling in blade servers, Tim Doughtery, IBM's BladeCenter strategist, said in an interview. [?] Dougherty wouldn't say the PowerPC line has no future, but with AltiVec now incorporated into Power6, "I don't see it right now," he said."
The 341mm2 die is huge for a 65nm chip but there are 2 on-chip memory controllers plus an on-chip L3 controller and 8MB of L2 cache (diagram). Contrary to the POWER4 and 5, each core includes an Altivec unit.
The dual-core PPC 970MP was only 154mm2 with 2x1MB of L2 cache and of course it would be even smaller at 65nm. If Apple was committed to the PowerPC architecture they would need some kind of POWER6 derivative. The POWER4 was huge too, yet IBM produced a desktop class chip with the PPC 970. I'm not convinced that a POWER6-lite would fit in a notebook, though. It's far easier to switch to Intel, all the chips that Apple needs (from notebooks to servers) are already in mass production, it's a "one stop shop."
It's good that you l,ooked all of that up. POWER 6 is too much for Apple.
I'm not surprised at the likely demise of the 970. IBM is abandoning the low end for good, it seems.
That's Adobe being lazy and seeing a quick buck in quickly porting a Windows product back to the Mac that they left to go stale some years ago in the OS9->OSX transition. To get it out quick they've slapped in Windows x86 code.
Of course that's the case. That's not the point.
Adobe is a major player in the Mac software market. People alter hardware purchasing decisions based on the availability of Adobe products.
Apple can't go around changing processors, telling developers to "just recompile" when there are companies like Adobe that routinely respond "no" or "only when we're good and ready."
If Apple would release a new PPC system, some of the die-hard Mac users would appreciate it, many more would not. They'd respond "we just spent all this time and money upgrading our apps and now it was all for nothing?" They'd say "we just finished migrating to all-Intel to take advantage of the newest Adobe apps and now the next generation won't run any of them?"
Mac OS isn't even close to big enough to be able to ship workstations with multiple processor architectures, and keep them all properly supported at once. And I doubt the big software vendors will want to play ball either.
... I really don't believe Apple will keep the iPhone platform closed or that the iPhone is the entire platform they'll be using OSX on ARM for.
So how is that iPod software development kit coming along? What you don't have one? Why not? Oh that's right, it's still a closed platform after all these years.
What makes you think the iPhone will be any different?
Unless you have access to some confidential Apple information, your belief is nothing more than wishful thinking.
If Apple would release a new PPC system, some of the die-hard Mac users would appreciate it, many more would not. They'd respond "we just spent all this time and money upgrading our apps and now it was all for nothing?" They'd say "we just finished migrating to all-Intel to take advantage of the newest Adobe apps and now the next generation won't run any of them?"
The latest Adobe apps are Universal binaries, not Intel only. There's no reason why the apps would stop working if Apple just came out with another PPC Mac.
Mac OS isn't even close to big enough to be able to ship workstations with multiple processor architectures, and keep them all properly supported at once. And I doubt the big software vendors will want to play ball either.
But that's exactly what Apple are doing now or did I miss a memo where Leopard wasn't supported on PowerPC?
So how is that iPod software development kit coming along? What you don't have one? Why not? Oh that's right, it's still a closed platform after all these years.
If you're the right person Apple will give you an iPod SDK.
What makes you think the iPhone will be any different?
Because the iPhone isn't an iPod. It's a much broader, more computer like environment than an mp3 player. Every smartphone platform going relies to some extent on 3rd party developers. Sure, Apple could do it all themselves but that seems just too stupid for words.
The beauty of it would be that future games are optimized for it. The consoles remain static and the only way that games work well is that they are written for it - you'd probably have to do this to make any return. For games that are not optimized specifically for it, the fact that it would be a generic PC gives you a good level of compatibility.
Concerning updates, it would last at least 3 years. The GeForce 6600GT is still a decent card and it's 3 years old. Also, newer games like Double Agent, although they require DirectX shader model 3 and things, a version has been written for the PS2, which performs about the same as a Mac Mini.
Wow. 3 years old?
The latest Adobe apps are Universal binaries, not Intel only. There's no reason why the apps would stop working if Apple just came out with another PPC Mac.
But that's exactly what Apple are doing now or did I miss a memo where Leopard wasn't supported on PowerPC?
Aegis, you know that that's all true now.
We aren't denying that it's true now.
I'm saying, for example, and so are others, that Apple will come out with a 10.6 for PPC. That will happen about two and a half to three years from now.
After that introduction, Apple will keep it current with 10.6 for x86 until 10.7 comes out, another two and a half years, or so, later. The likelihood that Apple will have a 10.7 PPC version is extremely slight, almost non existent. That will be five to five and a half years from now though. That's what we're saying.
You don't really think that Apple will produce another PPC machine during that time do you? It seems that the PPC is going away anyway soon.
How many PPC machines will be left that long from now? Very few, I'm sure.