No pedestal, please. UnixPoet has a legitimate difference of opinion on the importance of ISA.
As a software guy, I have a clear view of what it costs to modify existing software. What the X on x86 rumor proposes is a massive change to all software in the Mac universe. The cost of doing this will be enormous, even for those companies which simply hit a switch and recompile... if that happens even once. If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked. And on any piece of software of significant size there are going to be issues that require programmer time to fix. Software that is already cross-platform is not 100% cross-platform between OSX on PPC vs. x86 because the x86 variant has not been available for testing. There is much more to software development than just recompiling. Apple has a huge investment in their software, so they must know this.
Hardware is expensive too, but a single project could bring a PPC product into Intel's lineup. This project could be the conversion of an existing PPC to Intel's fab, or it could be the modification of one of Intel's own core designs to support the PPC ISA (hugely simplified by their seperate decoder architecture -- modifying a PPC to decode x86 would be much much harder). Either way it would leave Apple able to buy from 3 processor vendors, covering a very wide range of performance/price/power capabilities. It would also not disrupt the customer base at all, and it is well within even a fraction of the value of the money that Apple has in the bank right now.
We'll see tomorrow what truth there is behind these rumors, and what reasoning Steve's team is using.
I completely agree with you about the difficulty of bringing software over. I've said this over and over again. Some people just don't seem to want to believe it.
I'm not so sure about Intel (or AMD) manufacturing a PPC though. Even though Apple is one third of the PPC group, and has had input on the designs, I don't know if the agreements would allow them to outsource a chip by themselves. They might have to get both IBM and Freescsle to sign off on that, and what would be their incentive? With IBM's Power Everywhere initiative, Apple's defection would have the effect of setting that back severely. Why would others join if the major customer for the chips defects? And Freescale is still advancing their G3 and G4 designs. They would lose sales as well.
With the posts about a 4GHz Mac using Intel's chips out here I'm wondering if it's understood that Apple's 2.7GHz machine has achieved approximate parity with Intel's fastest Xenon and AMD's fastest Opteron? Intel won't be reaching 4GHz, just a small bit further ahead than they are now, until 65nm comes on line next year.
If it's thought that a PPC manufactured by Intel will reach that speed, I can only say to keep dreaming. It's not just the fab, but the chip design that's responsible for the chip speed. Intel only reached 2GHz on the Itanium this year. Hp's fastest mini computer chips have only reached 1.1GHz, though analysts thought they would be a bit higher (Hp could have brought them out at 1.3, but didn't, for marketing reasons, as they are switching to other chips).
I can't (won't?) believe that IBM's GX chips are that far behind schedule. If they do come out with a 3GHz GX and (or) a 3GHz MP this year, then it would be fine. Even by January. Something has to be done about the portable line though, SOON.
I really don't see this happening, but I have to admit that it is always possible.
im scared. i have a part of my brain which is telling me this entire thing is absolutely false, and the only intel announcement we would see tmrw is the possible intel based mac tablet. but if this is all true, steve better be prepared for the mixed reactions, and if there is one thing that companies despise, its a mixed reaction to their business ventures.
If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked.
QA? I think I remember hearing about that. Isn't that what software manufacturers used to do to work the bugs out of their product before releasing it? But I thought that all went out the window, with them now allowing the populace, for the small fee of buying the software, to do that testing for them...
I can't (won't?) believe that IBM's GX chips are that far behind schedule. If they do come out with a 3GHz GX and (or) a 3GHz MP this year, then it would be fine. Even by January. Something has to be done about the portable line though, SOON.
I agree, and I speculated elsewhere that this deal may not be about a move from PPC or the G5 but a replacement for the hoary old G4. While everyone is wetting themselves because the G5 has only scaled from 2.0 to 2.7GHz, in that time the G4 has gone from 1.42 to a whopping 1.67! And still on a 167MHz bus! If IBM has been unable to deliver a promised portable chip, and Freescale seems to be MIA, I can see Apple demanding whatever is necessary for them to take PPC to Intel and have them fab a Centrinoesque PPC chip for portables and the mini.
x86 does not at all make it easier to port Windows software over. They use two completely separate sets of APIs and Frameworks. The only way it's "easier" is if Apple integrates a windows emulator, which would be kinda lame unless it actually displays the programs with an Aqua look.
Based on my experience with NeXT, there is no reason why OSX on Intel wouldn't work. While AltaVec is not available for Intel, this is not necessarily a show stopper. Apple owns AltaVec. It could be ported.
If the report that Intel would start at the bottom is true, one could speculate on an OSX solution that could run most important Windows applications natively but without the maintenance cost, security and virus issues of the Windows OS.
A $400 Mac Mini that runs Windows applications natively, would not have Microsoft seat license, maintenance costs and security concerns. This would be a very big deal in corporate IT. This would allow seamless migration to a far superior system today for a fraction of the cost of an XP seat.
This would give IT a solution to security, their most pressing problem. Now. Not in two years. It would do so for a fraction of the cost and do so with the performance of Longhorn. Now. Not in two years.
I agree, and I speculated elsewhere that this deal may not be about a move from PPC or the G5 but a replacement for the hoary old G4. While everyone is wetting themselves because the G5 has only scaled from 2.0 to 2.7GHz, in that time the G4 has gone from 1.42 to a whopping 1.67! And still on a 167MHz bus! If IBM has been unable to deliver a promised portable chip, and Freescale seems to be MIA, I can see Apple demanding whatever is necessary for them to take PPC to Intel and have them fab a Centrinoesque PPC chip for portables and the mini.
Hoary, I've always liked that word.
The worst thing about the G4 is that performance has not increased much with the speed increase, if at all.
I just sent Powerlogix an e-mail because I'm about to upgrade two of my machines shortly. I think it bears being put in my post for thought. I'm awaiting an answer from then, as I sent it Friday afternoon.
Here it is:
"This is pre-sales, but it's a tech question as well, so I'm not sure if you will want to answer it or pass it on.
There are now two different lines of processors that you have.
The 1.33 7457 and the 1.8 7447a, both duals.
I'm trying to decide what to get for my Digital Audios.
The confusing thing here is the speed vs. cache. I remember the last G4 Macs. One was a dual 1.25 with 1MB cache, and the other was a dual 1.42 with 2MB cache. The 1.42 only beat the 1.25 by about 6%. As I remember that the Mac's bus was limited to about 1.3GHz throughput, the cpu's were starved. The only reason that the 1.42 was faster was because of the larger cache. And that was with a 167MHz bus. The story should be even worse at 133.
So now we have either chips that are slower but with a 2MB cache running at 1.33GHz vs. chips with no L3 cache running at 1.8GHZ.
It doesn't seem to me that there is a performance advantage to the 7447a's, and that they might actually be a detriment.
So, other than a power advantage, what is the true story here?
Which should I use? Do you have tests comparing the two on the same machine, preferably a 133MHZ one?
This isn't easy to decide. I have two machines that I want to upgrade. One now, and one later this summer. I would like some peace of mind about this.
As a software guy, I have a clear view of what it costs to modify existing software. What the X on x86 rumor proposes is a massive change to all software in the Mac universe. The cost of doing this will be enormous, even for those companies which simply hit a switch and recompile... if that happens even once. If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked. And on any piece of software of significant size there are going to be issues that require programmer time to fix. Software that is already cross-platform is not 100% cross-platform between OSX on PPC vs. x86 because the x86 variant has not been available for testing. There is much more to software development than just recompiling. Apple has a huge investment in their software, so they must know this.
Well as a software guy that specialised in OS X for a long time I know that this is not much of an issue. The recompile will take the same time as a compile for ppc ( and could be attempted on a ppc machine if intel binaries are released in a sdk). Testing isn't much of an issue either, the programs are written on top of carbon, or cocoa, or other api ( like openGL), or even the OSX unix layer, all of which do not have to change. The chip will change under the application developer, but so what? That happens with normal transitions to G4/G5 chips. OS X is written to be processor independent. In other words there is no need for developers to really test anything - it should just work. All bugs are Apple bugs - the developer has to be aware of nothing. Apple will probably issue testing machines to certain developers - but not to iron out developer bugs caused by developers - but bugs that Apple has in it's frameworks, mostly.
This is an order of magnitude easier than the transition to ppc for 68k, or the transition to Carbon from the old OS 9 api set for carbon coders. The transition to Carbon forced old OS 9 developers to use a modern version of the old OS 9 api set ( which was a mess) and to replace processor dependent code like assembly - or old api which were dangerous. The carbon api was a large subset of the OS 9 codebase ( and added some new api sets, of course), but was still only 80%, so it took Adobe a long time to port photoshop.
That work -having been done - does not have to be redone. It should Just Work.
By the way this is exactly what Apple will say tomorrow.
(The only fly in the ointment is altivec. I am sure that altivec will be included on the new machine's chipset somewhere.)
Remember the hints that Steve dropped this spring about talks with 3 of the top PC makers. As hard as it may be to believe, the current stories could very well be about licensing. His hints are usually well planned.
Also don't forget the odd quote from Intel's Otellini, " "If you want to fix it tomorrow, maybe you should buy something else."
Based on my experience with NeXT, there is no reason why OSX on Intel wouldn't work. While AltaVec is not available for Intel, this is not necessarily a show stopper. Apple owns AltaVec. It could be ported.
If the report that Intel would start at the bottom is true, one could speculate on an OSX solution that could run most important Windows applications natively but without the maintenance cost, security and virus issues of the Windows OS.
A $400 Mac Mini that runs Windows applications natively, would not have Microsoft seat license, maintenance costs and security concerns. This would be a very big deal in corporate IT. This would allow seamless migration to a far superior system today for a fraction of the cost of an XP seat.
This would give IT a solution to security, their most pressing problem. Now. Not in two years. It would do so for a fraction of the cost and do so with the performance of Longhorn. Now. Not in two years.
This isn't really correct though. If you read back here you will se why it isn't that easy. We seem to be doomed to repeating the same things over and over.
While OS X could undoubledly run on x86, that isn't the problem, it's the programs that run on OS X that are the proble. Not as easily solved.
The other strange thing about this rumor is why Apple would take the lowest machines in their lineup and bring them to an x86, if the question is that of speed for the high end machines. This boggles the mind. Apple could always move the lowest machine upwards in the processor line if they had to.
If the Mini and the eMac were the problems, they could have always put a 1.5GHz G5 in either. It would give a 40% boost over a 1.4 GHz G4. That doesn't consume much power, and doesn't put out that much heat. The eMac would have been fine for this as is. The Mini could have been an inch higher and no one would have commented. It still would have been very small and quiet. A 1.5GHZ would only need 20 watts or less. If you look at the charts at IBM you would see that a 2GH FX chip needs much less power than the old design.
Well as a software guy that specialised in OS X for a long time I know that this is not much of an issue. The recompile will take the same time as a compile for ppc ( and could be attempted on a ppc machine if intel binaries are released in a sdk). Testing isn't much of an issue either, the programs are written on top of carbon, or cocoa, or other api ( like openGL), or even the OSX unix layer, all of which do not have to change. The chip will change under the application developer, but so what? That happens with normal transitions to G4/G5 chips. OS X is written to be processor independent. In other words there is no need for developers to really test anything - it should just work. All bugs are Apple bugs - the developer has to be aware of nothing. Apple will probably issue testing machines to certain developers - but not to iron out developer bugs caused by developers - but bugs that Apple has in it's frameworks, mostly.
This is an order of magnitude easier than the transition to ppc for 68k, or the transition to Carbon from the old OS 9 api set for carbon coders. The transition to Carbon forced old OS 9 developers to use a modern version of the old OS 9 api set ( which was a mess) and to replace processor dependent code like assembly - or old api which were dangerous. The carbon api was a large subset of the OS 9 codebase ( and added some new api sets, of course), but was still only 80%, so it took Adobe a long time to port photoshop.
That work -having been done - does not have to be redone. It should Just Work.
By the way this is exactly what Apple will say tomorrow.
(The only fly in the ointment is altivec. I am sure that altivec will be included on the new machine's chipset somewhere.)
Yes, a recompile will work. That's never been questioned here. But a recompile will result in some very bloated code. Code that will run slowly, because platform crossed re-compiles are inefficient. Even with a compatibility layer.
This leads to painstaking hand tuning. Altivec is not the same thing as SSE, or whatever the initials are, I for get at the moment. As this provides a big boost to numerous areas, even of the OS itself, I don't know how this would be easy.
Here in 10.4 Apple is using Altivec for Extreme as well as its other uses. This would all have to be remapped. Sometimes there will be nothing equivalent to remap it to, and a compromise will have to be made. OS X is not Windows or Linux. It does things differently.
I'm not saying that it can't be done, but it won't be that easy. And going to the low end machines first is ass backwards.
wouldn't it be interesting if apple said it not only was switching to intel-based acrhitecture, but that *gasp* apple would be the first out of the gate with apps that took advantage of such architecture, like new versions of iwork, the ilife suite, the production studio suite, etc.
we'll ignore the fact that apple would anger just about every large software manufacturer out there...
wouldn't it be interesting if apple said it not only was switching to intel-based acrhitecture, but that *gasp* apple would be the first out of the gate with apps that took advantage of such architecture, like new versions of iwork, the ilife suite, the production studio suite, etc.
we'll ignore the fact that apple would anger just about every large software manufacturer out there...
I just wonder what would happen to the performance to programs such as FCP. Apple highly optimized it for the G5.
I've had a *huge* increase in rendering speed from that.
I'm hoping this Intel story is a leaked fake to bring attention to tomorrow's WWDC speech by Jobs where he announces a super fast G6 dual core cell based PPC chip that just flat out blows away anything from Intel or AMD by a magnitude or more.
I suppose everyone gets to post a speculative note this weekend.
CNET is so coy and vague with their article. WSJ isn't any more specific. Still, something must be up with Apple and Intel.
Witness the odd endorsement of Apple by Paul Otellini last week. Sure, he's pimping the security features of Intel's new D Series. But it's a very interesting reference.
Then there's the Intel Mac Mini clone we've all seen. Unless this was done with the blessings of Apple, its design is so impossibly derivative as to invite a common law trademark challenge from Cupertino.
We also have Transitive's teasing reference to a "second company" that's ready to deploy their technology in early 2005. Sounds like this technology is for real, even allowing a modest performance hit.
On the other hand, IBM is already tipping what will become Power 6. They clearly think PPC is still the future.
My opinion is worth what you're paying for it, but here goes:
* Apple will retain PowerPC as its native architecture.
* Intel will supply its dual-core Pentium D chips for use in Apple's laptops. It will run OS X via Transitive's emulation technology. This array will be plenty snappy for the marketplace and provide Apple with modern mobility hardware.
* Pro-level gear will run OS X natively on IBM Power chips. Developers need not change their code. Current customers with investment in PPC needn't jump off the nearest bridge. All is well.
I suppose it is also possible that Apple will begin producing the Mini and eMac with a Pentium or Celeron M chipset, also running OS X via Transitive's emulation. The iMac is selling well as-is on the G5 platform. No need to mess with it.
The upside to all this chipmongering would be to catch-up Apple in the mobility market in very short order. We could have greatly improved Powerbooks and a refreshed iBook almost immediately, product-differentiated by dual- or single-core chipsets. When a true Power alternative is ready for the Powerbook, it gets dropped in without upsetting the developers and the consumer base.
Apple serves notice to IBM that it is very, very serious about chip supply. The next step would be a painful transition to x86, which I still doubt, or a jointly developed high-end Power chip with Intel. I'm not sure Power licensing would allow the latter alternative, though.
I think it is of great interest that Apple is currently running some very expensive, double-truck print ads featuring OS X. You'll notice the only hardware shown is Apple's new display line, which can be plugged into pretty much anything these days.
Since OS X currently requires Apple hardware, these ads are like writing a sales contract without a line for a signature.
My suspicion is that Apple will be primarily known as an entertainment brand a decade from now, marketing appliances that sell and store music and video content on demand. To achieve this, Apple must increase their market share.
If OS X can run via emulation on Intel-chipped Apple hardware, it could also run on these new Mac Mini clones. They were displayed running Windows, but could just as easily represent a careful program to license OS X on non-Apple boxes.
Well, you know it is entirely possible that one day, a year or two ago, Jobs and Gates per chance passed by eachother at a Washington Super WalMart one afternoon and started some friendly chit-chat over a box of Lucky Charms and Gates said," F-IT! Why bicker over whose is bigger anymore? Why don't we just combine forces and license everything to eachother and confuse the hell out of the whole frickin' world!"
Then Jobs said,"You know what, Apple just doesn't seem viable to me anymore, in the configuration that I currently have it as and I don't want to be different anymore (Bill at this moment cuddles Jobs and they weep on each others shoulders) so you know what? I'm going to switch everything over to Intel and we'll work hand in hand now, Bill!"
I think it is of great interest that Apple is currently running some very expensive, double-truck print ads featuring OS X. You'll notice the only hardware shown is Apple's new display line, which can be plugged into pretty much anything these days.
Since OS X currently requires Apple hardware, these ads are like writing a sales contract without a line for a signature.
My suspicion is that Apple will be primarily known as an entertainment brand a decade from now, marketing appliances that sell and store music and video content on demand. To achieve this, Apple must increase their market share.
If OS X can run via emulation on Intel-chipped Apple hardware, it could also run on these new Mac Mini clones. They were displayed running Windows, but could just as easily represent a careful program to license OS X on non-Apple boxes.
Tough to say.
Well, we don't have that many more hours to go. I have a fair amount of stock, and to tell the truth, my stomach is churning.
Comments
Originally posted by Programmer
No pedestal, please. UnixPoet has a legitimate difference of opinion on the importance of ISA.
As a software guy, I have a clear view of what it costs to modify existing software. What the X on x86 rumor proposes is a massive change to all software in the Mac universe. The cost of doing this will be enormous, even for those companies which simply hit a switch and recompile... if that happens even once. If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked. And on any piece of software of significant size there are going to be issues that require programmer time to fix. Software that is already cross-platform is not 100% cross-platform between OSX on PPC vs. x86 because the x86 variant has not been available for testing. There is much more to software development than just recompiling. Apple has a huge investment in their software, so they must know this.
Hardware is expensive too, but a single project could bring a PPC product into Intel's lineup. This project could be the conversion of an existing PPC to Intel's fab, or it could be the modification of one of Intel's own core designs to support the PPC ISA (hugely simplified by their seperate decoder architecture -- modifying a PPC to decode x86 would be much much harder). Either way it would leave Apple able to buy from 3 processor vendors, covering a very wide range of performance/price/power capabilities. It would also not disrupt the customer base at all, and it is well within even a fraction of the value of the money that Apple has in the bank right now.
We'll see tomorrow what truth there is behind these rumors, and what reasoning Steve's team is using.
I completely agree with you about the difficulty of bringing software over. I've said this over and over again. Some people just don't seem to want to believe it.
I'm not so sure about Intel (or AMD) manufacturing a PPC though. Even though Apple is one third of the PPC group, and has had input on the designs, I don't know if the agreements would allow them to outsource a chip by themselves. They might have to get both IBM and Freescsle to sign off on that, and what would be their incentive? With IBM's Power Everywhere initiative, Apple's defection would have the effect of setting that back severely. Why would others join if the major customer for the chips defects? And Freescale is still advancing their G3 and G4 designs. They would lose sales as well.
With the posts about a 4GHz Mac using Intel's chips out here I'm wondering if it's understood that Apple's 2.7GHz machine has achieved approximate parity with Intel's fastest Xenon and AMD's fastest Opteron? Intel won't be reaching 4GHz, just a small bit further ahead than they are now, until 65nm comes on line next year.
If it's thought that a PPC manufactured by Intel will reach that speed, I can only say to keep dreaming. It's not just the fab, but the chip design that's responsible for the chip speed. Intel only reached 2GHz on the Itanium this year. Hp's fastest mini computer chips have only reached 1.1GHz, though analysts thought they would be a bit higher (Hp could have brought them out at 1.3, but didn't, for marketing reasons, as they are switching to other chips).
I can't (won't?) believe that IBM's GX chips are that far behind schedule. If they do come out with a 3GHz GX and (or) a 3GHz MP this year, then it would be fine. Even by January. Something has to be done about the portable line though, SOON.
I really don't see this happening, but I have to admit that it is always possible.
Especially with Jobs in charge.
Originally posted by Programmer
If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked.
QA? I think I remember hearing about that. Isn't that what software manufacturers used to do to work the bugs out of their product before releasing it? But I thought that all went out the window, with them now allowing the populace, for the small fee of buying the software, to do that testing for them...
Originally posted by melgross
I can't (won't?) believe that IBM's GX chips are that far behind schedule. If they do come out with a 3GHz GX and (or) a 3GHz MP this year, then it would be fine. Even by January. Something has to be done about the portable line though, SOON.
I agree, and I speculated elsewhere that this deal may not be about a move from PPC or the G5 but a replacement for the hoary old G4. While everyone is wetting themselves because the G5 has only scaled from 2.0 to 2.7GHz, in that time the G4 has gone from 1.42 to a whopping 1.67! And still on a 167MHz bus! If IBM has been unable to deliver a promised portable chip, and Freescale seems to be MIA, I can see Apple demanding whatever is necessary for them to take PPC to Intel and have them fab a Centrinoesque PPC chip for portables and the mini.
Originally posted by Louzer
x86 does not at all make it easier to port Windows software over. They use two completely separate sets of APIs and Frameworks. The only way it's "easier" is if Apple integrates a windows emulator, which would be kinda lame unless it actually displays the programs with an Aqua look.
Sure it is.
If the report that Intel would start at the bottom is true, one could speculate on an OSX solution that could run most important Windows applications natively but without the maintenance cost, security and virus issues of the Windows OS.
A $400 Mac Mini that runs Windows applications natively, would not have Microsoft seat license, maintenance costs and security concerns. This would be a very big deal in corporate IT. This would allow seamless migration to a far superior system today for a fraction of the cost of an XP seat.
This would give IT a solution to security, their most pressing problem. Now. Not in two years. It would do so for a fraction of the cost and do so with the performance of Longhorn. Now. Not in two years.
Originally posted by BrunoBruin
I agree, and I speculated elsewhere that this deal may not be about a move from PPC or the G5 but a replacement for the hoary old G4. While everyone is wetting themselves because the G5 has only scaled from 2.0 to 2.7GHz, in that time the G4 has gone from 1.42 to a whopping 1.67! And still on a 167MHz bus! If IBM has been unable to deliver a promised portable chip, and Freescale seems to be MIA, I can see Apple demanding whatever is necessary for them to take PPC to Intel and have them fab a Centrinoesque PPC chip for portables and the mini.
Hoary, I've always liked that word.
The worst thing about the G4 is that performance has not increased much with the speed increase, if at all.
I just sent Powerlogix an e-mail because I'm about to upgrade two of my machines shortly. I think it bears being put in my post for thought. I'm awaiting an answer from then, as I sent it Friday afternoon.
Here it is:
"This is pre-sales, but it's a tech question as well, so I'm not sure if you will want to answer it or pass it on.
There are now two different lines of processors that you have.
The 1.33 7457 and the 1.8 7447a, both duals.
I'm trying to decide what to get for my Digital Audios.
The confusing thing here is the speed vs. cache. I remember the last G4 Macs. One was a dual 1.25 with 1MB cache, and the other was a dual 1.42 with 2MB cache. The 1.42 only beat the 1.25 by about 6%. As I remember that the Mac's bus was limited to about 1.3GHz throughput, the cpu's were starved. The only reason that the 1.42 was faster was because of the larger cache. And that was with a 167MHz bus. The story should be even worse at 133.
So now we have either chips that are slower but with a 2MB cache running at 1.33GHz vs. chips with no L3 cache running at 1.8GHZ.
It doesn't seem to me that there is a performance advantage to the 7447a's, and that they might actually be a detriment.
So, other than a power advantage, what is the true story here?
Which should I use? Do you have tests comparing the two on the same machine, preferably a 133MHZ one?
This isn't easy to decide. I have two machines that I want to upgrade. One now, and one later this summer. I would like some peace of mind about this.
Best regards, Mel Gross."
As a software guy, I have a clear view of what it costs to modify existing software. What the X on x86 rumor proposes is a massive change to all software in the Mac universe. The cost of doing this will be enormous, even for those companies which simply hit a switch and recompile... if that happens even once. If nothing else you have to go back through your quality assurance testing (you do have that right) to make sure that it all worked. And on any piece of software of significant size there are going to be issues that require programmer time to fix. Software that is already cross-platform is not 100% cross-platform between OSX on PPC vs. x86 because the x86 variant has not been available for testing. There is much more to software development than just recompiling. Apple has a huge investment in their software, so they must know this.
Well as a software guy that specialised in OS X for a long time I know that this is not much of an issue. The recompile will take the same time as a compile for ppc ( and could be attempted on a ppc machine if intel binaries are released in a sdk). Testing isn't much of an issue either, the programs are written on top of carbon, or cocoa, or other api ( like openGL), or even the OSX unix layer, all of which do not have to change. The chip will change under the application developer, but so what? That happens with normal transitions to G4/G5 chips. OS X is written to be processor independent. In other words there is no need for developers to really test anything - it should just work. All bugs are Apple bugs - the developer has to be aware of nothing. Apple will probably issue testing machines to certain developers - but not to iron out developer bugs caused by developers - but bugs that Apple has in it's frameworks, mostly.
This is an order of magnitude easier than the transition to ppc for 68k, or the transition to Carbon from the old OS 9 api set for carbon coders. The transition to Carbon forced old OS 9 developers to use a modern version of the old OS 9 api set ( which was a mess) and to replace processor dependent code like assembly - or old api which were dangerous. The carbon api was a large subset of the OS 9 codebase ( and added some new api sets, of course), but was still only 80%, so it took Adobe a long time to port photoshop.
That work -having been done - does not have to be redone. It should Just Work.
By the way this is exactly what Apple will say tomorrow.
(The only fly in the ointment is altivec. I am sure that altivec will be included on the new machine's chipset somewhere.)
Also don't forget the odd quote from Intel's Otellini, " "If you want to fix it tomorrow, maybe you should buy something else."
Originally posted by shawk
Based on my experience with NeXT, there is no reason why OSX on Intel wouldn't work. While AltaVec is not available for Intel, this is not necessarily a show stopper. Apple owns AltaVec. It could be ported.
If the report that Intel would start at the bottom is true, one could speculate on an OSX solution that could run most important Windows applications natively but without the maintenance cost, security and virus issues of the Windows OS.
A $400 Mac Mini that runs Windows applications natively, would not have Microsoft seat license, maintenance costs and security concerns. This would be a very big deal in corporate IT. This would allow seamless migration to a far superior system today for a fraction of the cost of an XP seat.
This would give IT a solution to security, their most pressing problem. Now. Not in two years. It would do so for a fraction of the cost and do so with the performance of Longhorn. Now. Not in two years.
This isn't really correct though. If you read back here you will se why it isn't that easy. We seem to be doomed to repeating the same things over and over.
While OS X could undoubledly run on x86, that isn't the problem, it's the programs that run on OS X that are the proble. Not as easily solved.
The other strange thing about this rumor is why Apple would take the lowest machines in their lineup and bring them to an x86, if the question is that of speed for the high end machines. This boggles the mind. Apple could always move the lowest machine upwards in the processor line if they had to.
If the Mini and the eMac were the problems, they could have always put a 1.5GHz G5 in either. It would give a 40% boost over a 1.4 GHz G4. That doesn't consume much power, and doesn't put out that much heat. The eMac would have been fine for this as is. The Mini could have been an inch higher and no one would have commented. It still would have been very small and quiet. A 1.5GHZ would only need 20 watts or less. If you look at the charts at IBM you would see that a 2GH FX chip needs much less power than the old design.
Originally posted by asdasd
Well as a software guy that specialised in OS X for a long time I know that this is not much of an issue. The recompile will take the same time as a compile for ppc ( and could be attempted on a ppc machine if intel binaries are released in a sdk). Testing isn't much of an issue either, the programs are written on top of carbon, or cocoa, or other api ( like openGL), or even the OSX unix layer, all of which do not have to change. The chip will change under the application developer, but so what? That happens with normal transitions to G4/G5 chips. OS X is written to be processor independent. In other words there is no need for developers to really test anything - it should just work. All bugs are Apple bugs - the developer has to be aware of nothing. Apple will probably issue testing machines to certain developers - but not to iron out developer bugs caused by developers - but bugs that Apple has in it's frameworks, mostly.
This is an order of magnitude easier than the transition to ppc for 68k, or the transition to Carbon from the old OS 9 api set for carbon coders. The transition to Carbon forced old OS 9 developers to use a modern version of the old OS 9 api set ( which was a mess) and to replace processor dependent code like assembly - or old api which were dangerous. The carbon api was a large subset of the OS 9 codebase ( and added some new api sets, of course), but was still only 80%, so it took Adobe a long time to port photoshop.
That work -having been done - does not have to be redone. It should Just Work.
By the way this is exactly what Apple will say tomorrow.
(The only fly in the ointment is altivec. I am sure that altivec will be included on the new machine's chipset somewhere.)
Yes, a recompile will work. That's never been questioned here. But a recompile will result in some very bloated code. Code that will run slowly, because platform crossed re-compiles are inefficient. Even with a compatibility layer.
This leads to painstaking hand tuning. Altivec is not the same thing as SSE, or whatever the initials are, I for get at the moment. As this provides a big boost to numerous areas, even of the OS itself, I don't know how this would be easy.
Here in 10.4 Apple is using Altivec for Extreme as well as its other uses. This would all have to be remapped. Sometimes there will be nothing equivalent to remap it to, and a compromise will have to be made. OS X is not Windows or Linux. It does things differently.
I'm not saying that it can't be done, but it won't be that easy. And going to the low end machines first is ass backwards.
we'll ignore the fact that apple would anger just about every large software manufacturer out there...
Originally posted by rok
wouldn't it be interesting if apple said it not only was switching to intel-based acrhitecture, but that *gasp* apple would be the first out of the gate with apps that took advantage of such architecture, like new versions of iwork, the ilife suite, the production studio suite, etc.
we'll ignore the fact that apple would anger just about every large software manufacturer out there...
I just wonder what would happen to the performance to programs such as FCP. Apple highly optimized it for the G5.
I've had a *huge* increase in rendering speed from that.
CNET is so coy and vague with their article. WSJ isn't any more specific. Still, something must be up with Apple and Intel.
Witness the odd endorsement of Apple by Paul Otellini last week. Sure, he's pimping the security features of Intel's new D Series. But it's a very interesting reference.
Then there's the Intel Mac Mini clone we've all seen. Unless this was done with the blessings of Apple, its design is so impossibly derivative as to invite a common law trademark challenge from Cupertino.
We also have Transitive's teasing reference to a "second company" that's ready to deploy their technology in early 2005. Sounds like this technology is for real, even allowing a modest performance hit.
On the other hand, IBM is already tipping what will become Power 6. They clearly think PPC is still the future.
My opinion is worth what you're paying for it, but here goes:
* Apple will retain PowerPC as its native architecture.
* Intel will supply its dual-core Pentium D chips for use in Apple's laptops. It will run OS X via Transitive's emulation technology. This array will be plenty snappy for the marketplace and provide Apple with modern mobility hardware.
* Pro-level gear will run OS X natively on IBM Power chips. Developers need not change their code. Current customers with investment in PPC needn't jump off the nearest bridge. All is well.
I suppose it is also possible that Apple will begin producing the Mini and eMac with a Pentium or Celeron M chipset, also running OS X via Transitive's emulation. The iMac is selling well as-is on the G5 platform. No need to mess with it.
The upside to all this chipmongering would be to catch-up Apple in the mobility market in very short order. We could have greatly improved Powerbooks and a refreshed iBook almost immediately, product-differentiated by dual- or single-core chipsets. When a true Power alternative is ready for the Powerbook, it gets dropped in without upsetting the developers and the consumer base.
Apple serves notice to IBM that it is very, very serious about chip supply. The next step would be a painful transition to x86, which I still doubt, or a jointly developed high-end Power chip with Intel. I'm not sure Power licensing would allow the latter alternative, though.
I think it is of great interest that Apple is currently running some very expensive, double-truck print ads featuring OS X. You'll notice the only hardware shown is Apple's new display line, which can be plugged into pretty much anything these days.
Since OS X currently requires Apple hardware, these ads are like writing a sales contract without a line for a signature.
My suspicion is that Apple will be primarily known as an entertainment brand a decade from now, marketing appliances that sell and store music and video content on demand. To achieve this, Apple must increase their market share.
If OS X can run via emulation on Intel-chipped Apple hardware, it could also run on these new Mac Mini clones. They were displayed running Windows, but could just as easily represent a careful program to license OS X on non-Apple boxes.
Tough to say.
Then Jobs said,"You know what, Apple just doesn't seem viable to me anymore, in the configuration that I currently have it as and I don't want to be different anymore (Bill at this moment cuddles Jobs and they weep on each others shoulders) so you know what? I'm going to switch everything over to Intel and we'll work hand in hand now, Bill!"
I guess we'll see tomorrow won't we?
Originally posted by paperfrog
... and one more thing.
I think it is of great interest that Apple is currently running some very expensive, double-truck print ads featuring OS X. You'll notice the only hardware shown is Apple's new display line, which can be plugged into pretty much anything these days.
Since OS X currently requires Apple hardware, these ads are like writing a sales contract without a line for a signature.
My suspicion is that Apple will be primarily known as an entertainment brand a decade from now, marketing appliances that sell and store music and video content on demand. To achieve this, Apple must increase their market share.
If OS X can run via emulation on Intel-chipped Apple hardware, it could also run on these new Mac Mini clones. They were displayed running Windows, but could just as easily represent a careful program to license OS X on non-Apple boxes.
Tough to say.
Well, we don't have that many more hours to go. I have a fair amount of stock, and to tell the truth, my stomach is churning.
TIGERTEL
...has a nice ring to it.