64 bit PPCs are backwards compatible with 32 bit PPC code, ie. all the current Mac software will still run. Apple won't "migrate" to the 64 bit PPC any more than they migrated from the G3 to the G4. The jump to an x86 processor is a *much* bigger jump. This is really apples and oranges.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
IBM Power4 and Apple
- Thread starter arn
- Start date
- Sort by reaction score
64 bit PPCs are backwards compatible with 32 bit PPC code, ie. all the current Mac software will still run. Apple won't "migrate" to the 64 bit PPC any more than they migrated from the G3 to the G4. The jump to an x86 processor is a *much* bigger jump. This is really apples and oranges.
There are other pitfalls of moving to Wintel or AMD, but beyond that, you will probably not see such a move until there are more COCOA apps. And that is a few years down the road as all major apps want the OS 9 Market as well as OS X, for several reasons major apps have to be legacy apps, carbon.
IBM has as much of a future as AMD or intel with chip manufacture. AMD/Intel will eventually have to ditch x86, but they may find that much more difficult to do than the switch from 68k to ppc, or from os9 to os x due to the ubiquity and inbeddedness of the PC market.
IBM knows how to get chips out in volume, and on time and has state of the art R and D pushing the envelope. Motorola is still run by people whose dream business model is selling TVs through furniture stores as they did 30 years ago, they are only in the tech business by an accident of hiring some guys to create a remote control for them. They have a handful of smart guys on the payroll drowning in a corporate culture with 1950's values, like a kernal of corn in a turd. The best deal about the development of the PPC is that with IBM in the mix, it always created another possible supply of corn for Apple, but Motorola has always made sure anyone has to climb through a lot of **** to get to it.
Originally posted by G4scott
Well, all of the news is fitting together nicely. A new chip plant, a scaled down Power4, and discussions between IBM and Apple... I think Apple's been somewhat pissed off at Moto because of 1) their first flop with the G4 and not being able to produce enough (where IBM had to come in and help) and 2) performance is falling behind
One thing about Apple is that they try to reach ahead, and develop new technologies that they think will catch on, and work out for them. They will seem to be falling behind the competition, but instead of working on just catching up, they're trying to find a way to surpass the competition. Right now, it looks like Apple has 2 options, the Motorola G5 or the IBM scaled down Power4 (what would they call it? The Power4 is a big, power hungry chip, and this isn't, maybe the miniPower4?). Motorola has been working on the G5 for 2 years or even more, and they still haven't gotten a desktop version on the market... IBM has been making Power4's, and this new chip is just a matter of taking off what a desktop CPU doesn't need... If the new Power4 does have alti-vec, great, but if not, it may be fast enough that you don't need special instructions for better performance...
Personally, I think that Apple's going to have Motorola try to make the G4 a lower power, cooler, yet faster chip for portables, and use this new Power4 for their desktops... Now, the day you get a laptop with a Power4...
Yeah, pissed is an understatement. LOL. They have to be nice in public but I'm sure Steve is not happy about having to use a router chip (hehe, or a Nintendo chip) in his top of the line workstation. Actually, the PowerPC is a very nice platform, it's just been lacking in it's high end implementations. Ever since the original roadmap with the mythical PowerPC 620 (the chip that, at the time, was even supposed to blow the doors off of the then-speed-champ Alpha) fell off the radar, it's been that way.
And yes, the chip fab in Fishkill is almost definitely for this chip. IBM has stated that it's for external customers and for the high end. Cell is not ready for any sort of production yet and so of course that would be one of the biggest mismanagements of capital to have a fab (which cost billions of dollars) sitting idle. The only other application in high end for external customers would be desktop/workstation (read Apple).
Apple needs to surpass IA-32 pretty quickly if they want to be taken seriously in the vertical markets they are trying to target. Why spend more for Apple hardware to run Shake and Maya when cheap IA-32 hardware running Linux does a better job for less money. Apple's a nice platform but money talks.
Also, don't disregard a vector/SIMD unit as being trivial. Good floating point performance is going to be a given (seeing how the POWER4 screams on SPECcpu...something like 1042 where Itanic was doing 800-something). That's all good for technical computing like GIS, some CAD/CAE, and visualization stuff. But SIMD is extremely useful for doing some of the meaty realtime video encoding, effects processing, etc. And considering that Apple wants to be the new Hollywood powerhouse, I think a good SIMD unit is a necessary.
One of the interesting things about the Power4 is that the system bus is always at half the processor speed, and isn't fixed. I wonder if this feature will carry on to this new chip. Of course, then we'd all have to buy PC500 or a special type of RAM to work with the bus. That's going to be interesting, and I'd like to see how the bus speeds for these new processors turn out...
Originally posted by Foocha
I'm not questioning the fact that IBM makes servers. I'm questioning IBMs commitent to developing PowerPC chips that meet Apple's needs. I don't believe that it is not correct to say that IBM has always made what Apple wanted. If this was the case, I think we would have seen IBM supporting Altivec a long time ago.
Whilst I recognise that Apple will not release a PC compatible version of OS X any time soon, that does not preclude the possibility of Apple migrating the Mac platform to an Intel processor. Clearly software would need to be recompiled and some software developers (Microsoft) may not be willing to do that.
All I'm saying is that if the Mac has to migrate to a new CPU - Apple might as well choose one with a future.
This is a discussion, not an arguement - I'm interested to know your views
Yeah, I have to agree that Intel is not the way to go. IBM is the poop when comes to fab technology. They're going to be working on 90 nanometer process soon and their SOI, low-K dielectric, and SiGe technology is pretty much unmatched in the industry.
And saying PowerPC doesn't have a future is a very uninformed statement. PowerPC is just behind on it's roadmap. IBM is driving the platform with POWER which is pretty much one of the only three contenders for high end chips (as well as the current performance champ, thank you very much). The others would be UltraSPARC and IA-64. IA-64 is unproven at this point and there's a lot of risk associated to moving to that platform until it's solid or disproven, one or the other. UltraSPARC would involve less risk but IBM is usually ahead of Sun and TI with both performance and manufacturing technology.
There's pretty much two schools of thought in processor technology right now. Let the chip find parallelism and schedule instructions as best it can, or let it be done at compile time. IBM happens to be on the more conservative hardware scheduling side. Intel is moving most scheduling to the compiler with it's EPIC/VLIW design. The problem with VLIW design is that it hasn't worked out as well in execution as in theory. Transmeta's Crusoe hasn't been able to keep up and Sun's MAJC didn't pan out for a general purpose processor. So staying with RISC might be prudent for Apple at this point. IBM and Sun are both working on instruction and thread level parallelism and IBM is doing neat things with POWER5 and POWER6 to move commonly used calculations (like handling TCP/IP) into hardware and out of software (can we say CISC? VAX?). I think IBM is going to be right up there with cutting edge processors. So don't say PowerPC doesn't have a future.
not to be skeptical, but...
I know we all would love to see a 64 bit chip. I know the rumors are strongly suggesting that Apple is going with IBM and that new models may be appearing soon. Here is my only question: what about the applications? What I mean is, wouldn't programs like Final Cut Pro or Photoshop have to be programmed to specifically take advantage of the 64 bit chip? And what about OS X? Is it possible that one of Jaguar's secret little bonuses is that it is already 64 bit ready or could we expect that with 10.3? There just seems to be a lot of questions that haven't been answered yet regarding a 64 bit chip. Either way, if all the rumors are true, God speed Apple, and good luck.
I know we all would love to see a 64 bit chip. I know the rumors are strongly suggesting that Apple is going with IBM and that new models may be appearing soon. Here is my only question: what about the applications? What I mean is, wouldn't programs like Final Cut Pro or Photoshop have to be programmed to specifically take advantage of the 64 bit chip? And what about OS X? Is it possible that one of Jaguar's secret little bonuses is that it is already 64 bit ready or could we expect that with 10.3? There just seems to be a lot of questions that haven't been answered yet regarding a 64 bit chip. Either way, if all the rumors are true, God speed Apple, and good luck.
Re: not to be skeptical, but...
What if such a 64 bit chip was able to act like two 32 bit ppcs, in a manner more purely additive than todays dual processors, which due to bottlenecks of being two separate chips, add up to 1.4 or 1.5 times rather than 2 times the processing power.
I even tend to look at multiprocessing problems (or one chip able to bite off more) this way-- One particular program may NOT be set up to take full advantage of a particular set up for however the data may be queued about, but with the OS, I could possibly have -more apps open at once- without such a processor hit- have something intense processing in the background without slowing down my foreground work as much. It may not be as important that one program be able to use a processors power fully, as it may be for Several programs to use as much as they can without having to share.
What if such a 64 bit chip was able to act like two 32 bit ppcs, in a manner more purely additive than todays dual processors, which due to bottlenecks of being two separate chips, add up to 1.4 or 1.5 times rather than 2 times the processing power.
I even tend to look at multiprocessing problems (or one chip able to bite off more) this way-- One particular program may NOT be set up to take full advantage of a particular set up for however the data may be queued about, but with the OS, I could possibly have -more apps open at once- without such a processor hit- have something intense processing in the background without slowing down my foreground work as much. It may not be as important that one program be able to use a processors power fully, as it may be for Several programs to use as much as they can without having to share.
Re: not to be skeptical, but...
The apps might have to be re-written to take advantage of the new chip, but they will certainly still work, and they will work faster than they ever have before, so it's probably going to be an optional thing for a couple of years...
The apps might have to be re-written to take advantage of the new chip, but they will certainly still work, and they will work faster than they ever have before, so it's probably going to be an optional thing for a couple of years...
Re: not to be skeptical, but...
The hardware has to come first. This is exactly like the switch from 68K to PPC (in more ways than one). The applications can't be written and tested for 64-bit without a working computer. The solution is to release the 64-bit Mac but have it run mostly in 32-bit mode. Now they have a base to work from and test against. Same thing for OSX, except that Apple has the hardware in advance of its release.
BTW, when they moved from 68K to the PPC, the main reason was that Motorola couldn't get their act together and release new processors fast enough. By the time the 68060 came out, Apple was well on their way to the PPC. And today, Motorola can't get their act together and release new processors fast enough.... Dump that dog already!
The hardware has to come first. This is exactly like the switch from 68K to PPC (in more ways than one). The applications can't be written and tested for 64-bit without a working computer. The solution is to release the 64-bit Mac but have it run mostly in 32-bit mode. Now they have a base to work from and test against. Same thing for OSX, except that Apple has the hardware in advance of its release.
BTW, when they moved from 68K to the PPC, the main reason was that Motorola couldn't get their act together and release new processors fast enough. By the time the 68060 came out, Apple was well on their way to the PPC. And today, Motorola can't get their act together and release new processors fast enough.... Dump that dog already!
I didn't mean that PowerPC in general has no future. I mean that it seems to have no future for Apple, and that the Mac's dependance on PowerPC is starting to do the platform harm.
Wry Cooter - are you saying that Apple could not develop the Carbon framework for an Intel-based Mac?
Wry Cooter - are you saying that Apple could not develop the Carbon framework for an Intel-based Mac?
No. I might be saying that if all apps were Cocoa, you could see OS X on Intel TODAY. (Ever notice most cocoa apps are really old NextStep apps, which ran on Intel when Steve returned?) Its the carbon code, as well as other classic apps, that will not port without being recoded, as far as I know. There are many reasons for carbon being the transitional phase during an OS shift, and why carbon might not work on x86- it makes platform dependant calls. Say PPCs have an instruction set, as do x86 designs. Carbon is a subset that works both in OS X and earlier MacOS. Cocoa needs OS X. But that cocoa app and OS X, might not need a PPC... savvy? But hell, why not just run Gimp on Linux then....
Someone else could explain this more clearly- with all the colored boxes on top of Mach, etc... I don't think my impression is a complete misconception.
I don't think the PowerPC, the G3 and G4 family and its future decendants as we know it today, is in itself bad- Its just that it is supplied by people who can't seem to make enough, make them fast, or make them fast enough.
What's good for the goose is good for the gander. A computer is a computer and a faster computer is a faster computer. The two company's goals are exactly the same as far as what they want from the PowerPC. IBM and Apple both want performance and IBM is going to drive performance into the POWER/PowerPC line. Why? Because it's used for their RISC-based Unix workstations, servers, and supercomputers. IBM needs to keep up with Intel and Sun in order to sell these machines in that very lucrative market. And if Apple stays along for the ride, good things will come of it. (Hell, if only we were using straight up POWER chips right now...) The only tough part is what they are going to do about the portable market.
Don't blame the PowerPC. Blame Motorola for wimping out and retargeting their efforts toward embedded (where all the money is).
The problem is that for over a decade Motorola has been unable to see mutual advantage in properly serving a large high profile customer, or several other market advantages for that matter. This is a company that repeatedly could not fill the early orders from their largest customer because they were selling every bit of their production trickle to smaller competing customers. They simply don't have very good business sense.
I rememberas a young child reading articles by Mac evangelists claiming the x86 hardware would hit a brick wall. It has yet to do so.
With AMDs new x86-64 I am left asking, when will they? Yes Digital introduced x86 as a s hort-term solution but, digital was wrong. Dont even bother claiming RISC is superior to CISC, that argument is a non sequitur.
New PowerMacs
-Originally posted by Faeylyn-
"Personally, I think it will be out by MWSF -- at the latest -- with a good chance of being announced as being part of the next PowerMac. This processor was announced almost a year ago. They're probably already very far along in development, if not finished. There is NO RULE that states one must first present a paper on the chip, then design the chip, then produce the chip, then put it in a Mac. They can do it in whatever order they want. Okay?"
From what I saw from those smuggled pictures of the new power macs they look as if they can hold a power4 (fingers crossed tightly) processor already; huge fans, extra memory slots, etc.
-Originally posted by Faeylyn-
"Personally, I think it will be out by MWSF -- at the latest -- with a good chance of being announced as being part of the next PowerMac. This processor was announced almost a year ago. They're probably already very far along in development, if not finished. There is NO RULE that states one must first present a paper on the chip, then design the chip, then produce the chip, then put it in a Mac. They can do it in whatever order they want. Okay?"
From what I saw from those smuggled pictures of the new power macs they look as if they can hold a power4 (fingers crossed tightly) processor already; huge fans, extra memory slots, etc.
Yes, but judging by all past performance, Applie will find a way to cripple the performance of the system to increase profit margin. It will include old, half-implemented technology, just like all the G3's and G4's have had. Don't think that the G5 will suddenly mean Apple's m.o. is going to change. because it isn't.
but the power4 is a step in the right direction.
i hate to say it but i believe you are correct. Apple's method of "getting blood from a stone" can't let me believe otherwise. i wish they would just release the technology they keep hidden. but if they did they would die because they are such a niche market and rely on Mac "fanatics" to buy into minor upgrades over and over. if they were a larger company things would be different. oh well. i hate Quark too, with a passion.
I"m not sure...
...where people are getting the idea that this is a scaled down Power4. It's a new chip that takes a lot of its design ideas from the Power4. This isn't just Power4 minus all the big price and heat generating parts. As for Apple not using it... if they don't (assuming the chip really is what it seems) I'll be pissed. This is the first really desktop oriented PPC chip since, what, the 604? It has TONS of memory bandwidth (Apple's current performance problem), it has a VPU (It seems reasonable that it's IBM's implementation of the Altivec/VMX instruction set. I certainly hope so.), it's made on IBM's fabs (.13 micron SOI, the most advanced manufacturing process I've heard of any current chip using), and it's based on a chip designed for multiple cores (easy scaling when it gets old?). Apple would have to a) be idiots, or b) know something about the chip that we don't, to not take this opportunity. As for all G3s and G4s being crippled... perhaps you're forgetting when they first came out. The Sawtooth G4 blew the pants off the Pentium 3s of its time. The beige G3 did the same to the P2s of its time <looks at the beige G3 on his desk, smiles>. I'm hoping/guessing that the G5 is going to do the same to the P4 (and more importantly, the Hammer). This makes it a bit more likely, imo.
...where people are getting the idea that this is a scaled down Power4. It's a new chip that takes a lot of its design ideas from the Power4. This isn't just Power4 minus all the big price and heat generating parts. As for Apple not using it... if they don't (assuming the chip really is what it seems) I'll be pissed. This is the first really desktop oriented PPC chip since, what, the 604? It has TONS of memory bandwidth (Apple's current performance problem), it has a VPU (It seems reasonable that it's IBM's implementation of the Altivec/VMX instruction set. I certainly hope so.), it's made on IBM's fabs (.13 micron SOI, the most advanced manufacturing process I've heard of any current chip using), and it's based on a chip designed for multiple cores (easy scaling when it gets old?). Apple would have to a) be idiots, or b) know something about the chip that we don't, to not take this opportunity. As for all G3s and G4s being crippled... perhaps you're forgetting when they first came out. The Sawtooth G4 blew the pants off the Pentium 3s of its time. The beige G3 did the same to the P2s of its time <looks at the beige G3 on his desk, smiles>. I'm hoping/guessing that the G5 is going to do the same to the P4 (and more importantly, the Hammer). This makes it a bit more likely, imo.
Carbon, being more portable that the original Mac Toolbox code, could be ported to another platform. Since the 680x0 code is gone, it should be much simpler. It might not be as simple as Cocoa because NeXT had been through that with OpenStep. Then again, without seeing the source code, who here can say for certain?
CarbonLib on Mac OS 8.x and above, hides some of the system details for the Carbon-ised application.
IBM p690 servers (Power4)--GM's supercom
Found an interesting article/thread at just-auto.com: 30 Aug 2002-
USA: GM's news suprecomputer will "speed vehicle design"
IBM has announced that GM has selected a supercomputing infrastructure based on theIBM pSeries 690 servers to power the company's vehicle design applications.
Now here's a message posted in reply:
from Texas Bill--The IBM p690 systems (also known as Regatta) will be deployed in detroit as wellas at Opel in Russelsheim and Saab in Trollhattan.
When linked they will have a total processing power of 2.3 terraflops, making the system one of the 10 largest supercomputers in the world.
The pseries is an 8 to 32 way symmetrical multiprocessing system and has the industry's currently most powerful processor - the Power4. TheThe p690 also has true logical partitioning. This allowsthe regatta to be divided into up to 16 virtual servers each with its own processor, memory and I/ODaimlerChrysler and Ford also use IBM systems. IBM is the system integrator for DaimlerChrsyler"s new web-based production system.
Note: IBM is a client of mine. But the p690 isreally a pretty spiffy machine. It can even detect components that may be on the verge of failing and reroute processes around them on the fly.
Incedentallly the the Power4 is from the same family of processors that power the Apple Macintosh and some of the new technologies derveloped for the Regatta will be appearing in tyhe next generation of Apple processors.
Found an interesting article/thread at just-auto.com: 30 Aug 2002-
USA: GM's news suprecomputer will "speed vehicle design"
IBM has announced that GM has selected a supercomputing infrastructure based on theIBM pSeries 690 servers to power the company's vehicle design applications.
Now here's a message posted in reply:
from Texas Bill--The IBM p690 systems (also known as Regatta) will be deployed in detroit as wellas at Opel in Russelsheim and Saab in Trollhattan.
When linked they will have a total processing power of 2.3 terraflops, making the system one of the 10 largest supercomputers in the world.
The pseries is an 8 to 32 way symmetrical multiprocessing system and has the industry's currently most powerful processor - the Power4. TheThe p690 also has true logical partitioning. This allowsthe regatta to be divided into up to 16 virtual servers each with its own processor, memory and I/ODaimlerChrysler and Ford also use IBM systems. IBM is the system integrator for DaimlerChrsyler"s new web-based production system.
Note: IBM is a client of mine. But the p690 isreally a pretty spiffy machine. It can even detect components that may be on the verge of failing and reroute processes around them on the fly.
Incedentallly the the Power4 is from the same family of processors that power the Apple Macintosh and some of the new technologies derveloped for the Regatta will be appearing in tyhe next generation of Apple processors.