The next generation Macintosh
First the disclaimers:
This is my first post to AppleInsider, but I'm a long-time Mac user and dedicated Apple supporter. I'm posting because I mentioned my little theory to a friend and AppleInsider reader, and he said, "Put your money where your mouth is, post on AppleInsider, and watch your ass get flamed off." (Obviously he doesn't buy my theory.)
I just started reading the forums here, so if I'm treading on old well-beaten ground, please forgive, and point me at the relevant threads.
I don't have any insider info. This is all the purest speculation right out of my own head.
The background, as I see it:
Point 1 -- Apple is losing the performance battle to Wintel PCs. PCs are, once again, faster and cheaper than Macs. Macs have OS X going for them, but if it takes longer to get your work done, a really cool OS isn't much consolation. Performance is less of an issue with consumer machines like the iMac, but Apple can't survive just selling thin-margin consumer products. They need those $4k top-end machines selling in significant numbers. But at the top end, buyers demand state-of-the-art, fastest-on-the-planet performance, and Macs don't offer that anymore.
Point 2 -- There is no G5. Not for at least a year, maybe never. The Apollo G4 might be coming, but it will only be a little faster than the current G4, not fast enough. Remember my disclaimer; I don't have anything more to go on here than rumor sites and news stories, plus common sense. The way I see it, Motorola took 2 years just to get the G4 from 400MHz to 800MHz. Now I am being asked to believe they will magically produce a 1.5-2GHz G5, the speed Apple needs for competitive Macs, a year or more sooner than expected. I don't believe it, and if it happens you could knock me over with a feather.
Point 3 -- From what I've read, the new iMac hardware design has been finished for 6-12 months, and has just been waiting for the cost of LCD panels to drop. But Apple's hardware team has been doing _something_ for the last 6 months, something big if the rumors are any indication. If it wasn't working on the iMac, obviously new high-end Macs are coming. Also the new iMac pricing yanks the rug out from under the low- and mid-range Quicksilver PowerMacs, suggesting that those models are going away soon. But the new machines won't have G5s because there is no G5, and G4s which are only a little bit faster would spell death for Apple. They have to do something big.
And so, my prediction:
In less than 6 months, maybe even this quarter, Apple will announce a new line of Macintosh computers which contain AMD processors running at 2GHz and up.
Now, put the flamethrower down for just a minute...
I most definitely do _not_ mean that Apple is going to port OS X to the Wintel PC hardware architecture. That would be fatally foolish. Apple is a hardware company, OS X exists only so they can sell more computers. What I mean is that Apple will build Macs with proprietary system boards that just happen to have an AMD processor instead of a PowerPC. This does not mean these new Macs will run Windows, nor does it mean you can run OS X on a Wintel box. These will still be _Macintosh_ computers, and only Macs will ever run OS X.
Porting OS X to another processor is a relatively minor effort, everyone seems to agree on that. The developers will whine a bit, but it will really be no big deal, OS X is ready-made to be a multi-platform OS. Most applications should be a one-click re-compile once the new dev tools are available. They don't even need to provide separate installers thanks to the app package scheme. With all the new native apps like MS Office X, many users can give up OS 9 entirely (I know I'm close to doing that). So there is no serious problem with losing the Classic environment, at least not for the power users who buy top-end machines and can upgrade to native OS X apps. And these new machines will just be the top one or two models, there will still be a G4-based low-end PowerMac in addition to the iMac and laptops.
Why AMD and not Intel? Partly because switching to Intel is too much like going over to the "dark side", the hard-core Mac fanatics will be able to accept AMD a bit more easily. But mainly because I think AMD will give Apple a much better deal, meaning lower costs. On the other hand, why not Intel? To paraphrase His Steveness from several years ago, there is no reason that Intel has to lose in order for Apple to win! But in fact there are other processor choices. How about an Alpha-based Mac? Too bad Compaq killed further development on that processor when they bought DEC, it was already at 1GHz in '98, by now it would be up around 5GHz, and kicking the P4's butt like you wouldn't believe. Ah, well.
Whatever the alternative chosen, if Motorola can't deliver some processor running at a minimum of 1.5GHz within 3-6 months, which I don't believe they can do, Apple has to drop them like a rotten tomato.
So there's my $0.02. Flame away!
Bob
This is my first post to AppleInsider, but I'm a long-time Mac user and dedicated Apple supporter. I'm posting because I mentioned my little theory to a friend and AppleInsider reader, and he said, "Put your money where your mouth is, post on AppleInsider, and watch your ass get flamed off." (Obviously he doesn't buy my theory.)
I just started reading the forums here, so if I'm treading on old well-beaten ground, please forgive, and point me at the relevant threads.
I don't have any insider info. This is all the purest speculation right out of my own head.
The background, as I see it:
Point 1 -- Apple is losing the performance battle to Wintel PCs. PCs are, once again, faster and cheaper than Macs. Macs have OS X going for them, but if it takes longer to get your work done, a really cool OS isn't much consolation. Performance is less of an issue with consumer machines like the iMac, but Apple can't survive just selling thin-margin consumer products. They need those $4k top-end machines selling in significant numbers. But at the top end, buyers demand state-of-the-art, fastest-on-the-planet performance, and Macs don't offer that anymore.
Point 2 -- There is no G5. Not for at least a year, maybe never. The Apollo G4 might be coming, but it will only be a little faster than the current G4, not fast enough. Remember my disclaimer; I don't have anything more to go on here than rumor sites and news stories, plus common sense. The way I see it, Motorola took 2 years just to get the G4 from 400MHz to 800MHz. Now I am being asked to believe they will magically produce a 1.5-2GHz G5, the speed Apple needs for competitive Macs, a year or more sooner than expected. I don't believe it, and if it happens you could knock me over with a feather.
Point 3 -- From what I've read, the new iMac hardware design has been finished for 6-12 months, and has just been waiting for the cost of LCD panels to drop. But Apple's hardware team has been doing _something_ for the last 6 months, something big if the rumors are any indication. If it wasn't working on the iMac, obviously new high-end Macs are coming. Also the new iMac pricing yanks the rug out from under the low- and mid-range Quicksilver PowerMacs, suggesting that those models are going away soon. But the new machines won't have G5s because there is no G5, and G4s which are only a little bit faster would spell death for Apple. They have to do something big.
And so, my prediction:
In less than 6 months, maybe even this quarter, Apple will announce a new line of Macintosh computers which contain AMD processors running at 2GHz and up.
Now, put the flamethrower down for just a minute...
I most definitely do _not_ mean that Apple is going to port OS X to the Wintel PC hardware architecture. That would be fatally foolish. Apple is a hardware company, OS X exists only so they can sell more computers. What I mean is that Apple will build Macs with proprietary system boards that just happen to have an AMD processor instead of a PowerPC. This does not mean these new Macs will run Windows, nor does it mean you can run OS X on a Wintel box. These will still be _Macintosh_ computers, and only Macs will ever run OS X.
Porting OS X to another processor is a relatively minor effort, everyone seems to agree on that. The developers will whine a bit, but it will really be no big deal, OS X is ready-made to be a multi-platform OS. Most applications should be a one-click re-compile once the new dev tools are available. They don't even need to provide separate installers thanks to the app package scheme. With all the new native apps like MS Office X, many users can give up OS 9 entirely (I know I'm close to doing that). So there is no serious problem with losing the Classic environment, at least not for the power users who buy top-end machines and can upgrade to native OS X apps. And these new machines will just be the top one or two models, there will still be a G4-based low-end PowerMac in addition to the iMac and laptops.
Why AMD and not Intel? Partly because switching to Intel is too much like going over to the "dark side", the hard-core Mac fanatics will be able to accept AMD a bit more easily. But mainly because I think AMD will give Apple a much better deal, meaning lower costs. On the other hand, why not Intel? To paraphrase His Steveness from several years ago, there is no reason that Intel has to lose in order for Apple to win! But in fact there are other processor choices. How about an Alpha-based Mac? Too bad Compaq killed further development on that processor when they bought DEC, it was already at 1GHz in '98, by now it would be up around 5GHz, and kicking the P4's butt like you wouldn't believe. Ah, well.
Whatever the alternative chosen, if Motorola can't deliver some processor running at a minimum of 1.5GHz within 3-6 months, which I don't believe they can do, Apple has to drop them like a rotten tomato.
So there's my $0.02. Flame away!
Bob
Comments
<strong>"Porting OS X to another processor is a relatively minor effort"
Are you serious? Exactly how much do you know about processor architecture and low-level programming?</strong><hr></blockquote>
Not all that much, I'll admit. But more than half the work is already done, Darwin has already been ported, you can run it on your Wintel box right now. How hard can it be to finish off the job with the rest of the higher-level OS stuff? Other than Altivec-optimized code, which apps can't depend on anyway because they have to run on the G3, there shouldn't be any architecture-specific code above the OS kernel and device driver level.
The only thing that might stand in the way is device drivers, the developers might get a wee bit annoyed at having to re-write drivers again after having just finished switching from OS 9 to OS X.
Bob
<strong>"Porting OS X to another processor is a relatively minor effort"
Are you serious? Exactly how much do you know about processor architecture and low-level programming?</strong><hr></blockquote>
ehehe. I was about to say the same thing but you beat me to it
Seriously guys if you don't believe in the G5 I wouldn't be believing in a switch to AMD's chips (they may produce chips but won't design). It just won't happen.
As for the repeated notion of the G5, or any next generation chip not existing, whether it is the actual G5 or another chip is a matter of symantics, you can be certain development is well on the way. Apple isn't stupid and the second they had issues with the G4s they would have begun development even had they not started before then.
I won't even touch on the politics and other issues raised by any switches either.
[ 01-10-2002: Message edited by: Telomar ]</p>
Alex
<strong>more than half the work is already done, Darwin has already been ported</strong><hr></blockquote>
that is about 10-20% of OS X, and let me add that Darwin is the easy 10%.
Take a look at this graph
and you get a pretty good idea of what would be on the todo list.
bye.
[ 01-10-2002: Message edited by: GnOm ]</p>
BUT::: if you have to worry whether the G5 is backward-compatible to G4 - you'll never get that thing with an AMD CISC-ATHLON !!!! NEVER.
just try to re-compile your assembler coude *laugh*
<strong>And developing a 3Ghz CPU is also childs play...
Children have done NO WORSE at releasing even a 1 Ghz G4 than have Motorola engineers.
Because of: 1/ Altivec and 2/ the new iMac
Do you think Apple will made the switch to only the pro tower? Foolish !
You're right about the fact that in 2 years will only jump from G4 350 to G4 867.
I think/hope will see the motherboard with DDR / RapidIO / ATA 133 as soon as possible with Apollo or GoldFish. I hope also that they will provide the kind of stuff like SGI propose for they pro consumers.
I've heard Cocoa is ready to go and that Carbon might actually be as well. I'd only bet on Cocoa myself, but also that at least a migration plan for Carbon exists if it's not semi-completed.
Classic would obviously be dead, unless Connectix is ready for the challenge.
After that Aqua and Apple Script use the foundations listed above so they could probably be ported "relatively simply" if Quartz and Cocoa/Carbon were done. I still think Quartz would be tough though.
I don't think this is going to happen, but I think it's an option Apple has prepared for in one way or another.
-The G5, if there is a version for Apple or from Apple even, might well start at above 1GHz, even if the G4 doesn't clock well at all, simply because the G5 is a completely new architecture, it'sn not an evolutionary step as from the 603 to the G3 and from the 604 to the G4, the G5 is based on the Bluebook (or similar) called architecture idea that is a co-development of IBM and Moto.
-AMD is currenty having trouble getting the athlon up to speed as well. They're still stuck at 0.18 micron tech and high core voltage combined with enormous die size, at least compared to PPC chips as we know them. Switching to 0.13 will give them some headroom, but won't dramatically change the challenges that architecture bears. Intel's P4 has a brighter future there. The current Athlon CP 2000+ is in fact a 1.67GHz chip, thus 2GHz is most certainly quite a few months away.
Then I am asking some open questions:
-Why does everyone believe Apple would choose AMD over Intel? Because YOU like AMD better than Intel, which has become the brother of the Devil Microsoft for some reason, even though we owe A LOT to Intel's engineers, who have generally done a great job (PCI, AGP, USB etc etc).
The fact that Darwin and some other Apple tech is available for x86 platform does not mean that you can just glue those components together with some magic trick and boom, there you have a brand new x86 capable OS.
Also a switch in CPU, even if Apple kept it propietary would bring with it a lot of consequences and problems:
-Tinkerers and third parties would soon manage to get OS X to run on non-apple machines too. This was already the case with PPC clones. Some guys found ways how to circumwent the ROM issue and run Mac OS on third party PPC platforms without the Apple ROM.
This would not be to Apple's liking or benefit.
-Porting the OS alone isn't the whole job, in fact it's the smallest problem: You'd have to adapt all the software to the new platform, or run it in emulation, which wouldn't make sense, as you'd lose all the gained performance right away.
If you like theories about conspiracies, future Apple decisions etc, try to think out of Apple's point of view. It's much more likely that Apple will for example start developing, or already has, their own PPC chip. That is about 3 times more likely than having them switch to a PC platform. It would rip their face off in front of their users and fans..."OMG my favorite company has run over to the enemy".
I'm sorry, but in my opinion the whole "switch over to WINTEL" discussion is both pointless and childish. It's been invented by jealous teenagers and publicity hungry money sharks who call themselves "analysts". We're not gonna see it happen.
G-News
<strong>Two things that might clarify a few questions:
-The G5, if there is a version for Apple or from Apple even, might well start at above 1GHz, even if the G4 doesn't clock well at all, simply because the G5 is a completely new architecture, it'sn not an evolutionary step as from the 603 to the G3 and from the 604 to the G4, the G5 is based on the Bluebook (or similar) called architecture idea that is a co-development of IBM and Moto.
G-News</strong><hr></blockquote>
Go join the "There is no G5" thread if you're interested in this topic....
There is an excellent article in Arstechnica comparing these two processors in detail. you can read it here:
<a href="http://www.arstechnica.com/cpu/1q00/g4vsk7/g4vsk7-1.html"; target="_blank">http://www.arstechnica.com/cpu/1q00/g4vsk7/g4vsk7-1.html</a>;
here is an excerpt from the intro.
"Throughout this article, I'll emphasize the fact that regardless of marketing spin, stone-age controversies between platform factions, and general hysteria surrounding the terms "RISC" and "CISC", the K7 and the G4 are remarkably similar.Â* They face similar problems, and they solve them in similar ways.Â* The majority of the differences between these two CPUs are not the result of any sort of fundamental difference between "RISC" or "CISC" design philosophies -- most of the K7's CISC baggage is taken care of in its front end, and the back ends of both CPUs are fully post-RISC designs.Â* Rather, their differences stem from perfectly ordinary design tradeoffs of the kind all engineers and architects make, tradeoffs that involve price, power dissipation, performance, speed, support for legacy technology, and a whole host of other considerations.Â* AMD's K7 and Motorola's MPC7400 (the unit which, when paired with a particular chipset, becomes known as the G4) are, in fact, cousins with common ancestry".
The "front-end" of the Athlon takes care of all the x86 baggage via emulation. Removing the front-end would essentially make the Athlon a RISC CPU. This however does not make it anymore easier to run PPC OS's or apps on the Athlon. There would be some consderable work needed to run MacOS X or LinuxPPC on the Athlon. but it is worth mentioning that if Apple were to drop the PPC (which is unlikely due to the amount of resources and money spent on it already), the switch would not be as difficult as it would really seem.
Also note that this article was written in Jan 2000. It is 2 years old and some of the design of the Athlon has channged (as well as the PPC). In fact they have used a lot more of Motorola's PPC technology in the Athlon in recent years. Especially in taking care of problems in heat and power consumption. While the PPC architecture is considered more elegant, the Ahtlon is far from considered an inferior chip. One has to remember that AMD doesn't have the R&D budgets that Intel, IBM, or even Motorola have the luxury of holding. The work done to design and produce the Athlon is remarkable.
My apologies if this post was too long. I just wanted to clear up the misconception that the Athlon was a CISC chip. I also think the Athlon is an incredible processor. Not just in speed but in design as well. It is also worth noting that it doesn't really matter what chip is used to power the Mac. It's dark in the box. Product design aside. The Mac experience is the MacOS.
trowa.
[ 01-10-2002: Message edited by: trowa ]</p>
1) PowerPC compatible ==> no instruction set conversion nastiness (not a small issue as others have pointed out)
2) 64 bit (with 32 bit compatibility)
3) multi-core (2 microprocessors per die)
4) faster than *any* AMD or Intel CPU (32 or 64 bit)
5) does not incorporate AltiVec
6) extremely power hungry (~150 W)
If Apple/IBM can find a way to solve 5) and 6), the Mac would have hardware that very few people would be able to quibble with performance-wise.
A division of PowerPC design between Motorola for low power dissipation CPUs suitable for laptops or the iMac (which plays to Motorola's core interest in embedded CPUs) and IBM for high-performance CPUs suitable for pro desktops and servers (which plays to IBM's core interest in technical workstations, servers and mainframes) makes a *lot* more sense to me than porting to AMD, Intel, UltraSparc, etc.
[ 01-10-2002: Message edited by: gopher ]
[ 01-10-2002: Message edited by: gopher ]
"4) and 5)" swapped to "5) and 6)" (doh!)
[ 01-10-2002: Message edited by: gopher ]</p>
<strong>My apologies if this post was too long. I just wanted to clear up the misconception that the Athlon was a CISC chip.</strong><hr></blockquote>
The Athlon is a CISC chip. The Athlon uses a complex instruction set whether it is translated at the back end or not. The RISC ops are still hidden from the compiler/program and therefore become simply an implementation choice, not an architectural differentiation.
<strong>
4) faster than *any* AMD or Intel CPU (32 or 64 bit)
*snip*
If Apple/IBM can find a way to solve 4)
*snip*
</strong><hr></blockquote>
Maybe they could put it on a 25MHz bus?
Modern CISC designs from AMD and Intel decode the x86 instructions into a proprietary very-long-word instruction stream that is then fed to a core that is of modern design. The modern design arose from what was the RISC movement of the late 80s. That movement started out with the philosophy of having a small hardwired instruction set, but evolved to become something else. The PowerPC instruction set certainly isn't small, and it has some quite complex instructions... but it was designed with modern processor techniques in mind and thus is more easily implemented in high performance chips. This is in contrast to the x86 chips which have to have a "decoder" stuck on the front of them to translate x86 to their internal instructions. This costs transistors, power, and development time. The G4 has been stuck at lower clock rates because Motorola wanted shorter pipelines (and their fab problems), while the WIntel guys got caught up in a race to push clock rates higher and higher... even at the cost of crazy pipeline lengths. The G4 is not stuck as slower speeds because its a PowerPC.
<strong>In all fairness, Darwin, Quicktime, OpenGL, and Java 2 all run on AMD processors right now. </strong><hr></blockquote>
Darwin runs, true, QT not really, sure the playersoftware runs but in OS X QuickTime is a whole Framework, not just the player and some codecs. OpenGL and Java exist on x86 of course but Apple has (not really has to but after all it would be easier than trying to intergrate the existing versions) to port over their own versions because those are part of the core Technologies and interact heavily with each other.
So were still at only darwin running on x86.
bye.
<strong>
The Athlon is a CISC chip. The Athlon uses a complex instruction set whether it is translated at the back end or not. The RISC ops are still hidden from the compiler/program and therefore become simply an implementation choice, not an architectural differentiation.</strong><hr></blockquote>
Its not really a choice. They have to implement that front end, to provide compatiblity with x86 apps.