Separate names with a comma.
Discussion in 'General Mac Discussion' started by pc_convert?, Jan 23, 2002.
What does this mean for apple?
Interesting article ideas about who will fab the G5 here.
Motorola & IBM
it's a damn shame that Motorola can't keep afloat....but in their defense, there are many companies in the same boat as them....
the PowerPC processor is just a fraction of what they produce....so I'm finding it interesting that they are showing a $4 billion loss....
does anyone think IBM plans to use the G4 processor as their own....according to the article, it sounds like they are going to use something very similar to it....maybe just renaming it and changing a few minor details so it doesn't appear to be the exact same chip?
I think IBM will probably fab the current G4 i.e. exactly the same Motorola's inc AltiVec. This can only be good for Apple because it increases availability.
If IBM do fab the G5 this will be great. IBM is in a stable position (compared to Motorola) and could easily ramp up the production and get great yields upto 2GHz (They got the G3 to 1GHz).
I don't think it's a matter of IBM replacing Motorola as Apple's main supplier, after all they are all in the AIM alliance that created the PowerPC ISA.
In the long term I think IBM would be a better supplier for Apple, but I don't know what's going behind the scenes between these companies.
I had always hoped that IBM would make the G5 but I think those hopes are gone...
I am fine with moto making our next G5, unless they make it at high enough yields, and we don't end up in the G4 speed dump fiasco we had before...
do u think Apple would have exclusive rights to use the G5 processor if IBM were to manufacture them?
to help keep Mac performance levels ahead of PC?
I don't think Apple would have exclusive rights to use the G5, but Apple doesn't have exlclusive rights to use the G4.
When the PowerPC was first developed the AIM group prepared a couple of standards PReP and CHRP (I think these are the anacronyms). The idea was that these standards would allow motherboard manufacturers to make PowerPC PC's, destroying the x86 stranglehold on the PC market. Problem was no-one was willing to make the boards becuase PC users didn't want them.
It meant that PC users would have to change all there software (and wait windows! on PowerPC). Apple pushed ahead and had 68k emulatation under the PowerPC, successfully migrating to a new ISA, which itself is a remarkable feat.
I don't think the G5 would be picked up by anyone else. As far as I know the only other OS that runs on PowerPC is Linux (BeOS is dead now).
A while ago IBM realesed a protype design for a PowerPC board for free use but nothing really happened despite many companies saying they were developing boards based on it. I thought about having a go but I'm still at Uni and don't have much time. Anyway it's easier to get a Mac
if motorola can't stay in business, someone else will step in
...but it will be hard since apple may not be big enough for ibm to take seriously enough to keep speeds up and apple will fall that much more behind
perhaps it was a mistake that apple went with the g4 chip in the first place
i really think if motorola is in deep, then they'd be willing to sell part of their r&d dept geared more towards the consumer microprocessor market. by the dept i mean research already done and patents received. then apple could revamp the whole process and build on the good. altho, this is one department where Jobsian micromanagement would be extreme;y detrimental and timeconsuming.
Apple Employee mirrors Moto's demise
My own Apple source told me a similar rumor last night, before I saw this thread. She states that Moto will likely be leaving the Chip fabrication industry and Altivec/AIM production rights would be liscenced or sold to IBM. Whether Moto would also be leaving the R&D arena is anybody's guess. As Moto has been the most consistant bottleneck in Apple's production schedule this could only be a good thing.
I don't forsee any problems if Motorola drops out. IBM has made a large investment in the PowerPC architecture and would continue to develop it. I haven't checked any of the IBM PowerPC roadmaps but I'm sure they exist.
I don't think IBM would see Apple as a small customer, I believe IBM recently dropped out PC manufacturing, the margins on PC's have been squeezed so much in the last few years. They must be looking to make up this revenue, and fabbing Gx's for apple would surley contribute to this.
Also I'm not sure Apple is the only company that uses G4's. I thought Cisco routers use G4's (or maybe an embedded version), anyway I'm sure it would be worth it for IBM.
I've seen other people suggest that Apple buy up Motorola's IP and take over PowerPC R&D, but this gets very expensive (one reason why the AIM consortium was started - to spead the cost). I don't think this is practical solution though. In my opinion Motorola have consistently been "the weakest link" for Apple, I really think that having a second fab partner (IBM) would help Apple's development cycle.
Still this is all hypothetical, maybe Motorola has the G5 ready and in production...
The new Amiga if it ever appears is supposed to use PowerPC's, so IBM would have another customer.
Also I've found a company here that is about to start producing PowerPC motherboards. This may mean Linux users will switch to PowerPC boards to break away from Intel/AMD, so potentially another customer for IBM....
I'd like to point out something of a "big picture" jigsaw puzzle, here. I don't think we're going far back enough, in recent history. When the G4 processor was launched, it was openly acknowledged that it constituted the end of the original development agreement between Apple, IBM and Motorola. Not forgetting, either, that IBM came to Apple with the original 601 processor; not Motorola. The three got together at an R&D facility in Austin, TX (circa late '91 or early '92). There, the 601 was redesigned. Motorola didn't have much to contribute. But, strangely, Apple did. By the time they were through tinkering and tweaking, IBM had to redesign their manufacturing process for the chip. It might as well have taken on a 601e or 602 designation, by the accounts of the Apple techs involved. The greatest contribution Motorola had, evidently, was supportive chips (which, by the 604, was taken over by Texas Instruments) on the circuit board. Motorola was a more important player in the design of the 603 and 604. Apple, conversely, had minimal input on the design of the G3 and G4 chips. The G4 was almost entirely IBM's baby. Motorola's greatest contribution was the scalding hot Velocity Engine, which IBM's G4 doesn't need; that and a pipeline design error keeping the Motorola-built G4's confined to 500MHz for so long. This is irrespective of the 3 month long mechanical failure at Motorla's manufacturing plant. IBM easily and gladly produced the Mac G4s, during that time, much to Moto's protest. The fact is this: Motorola's presence -in this venture- has become largely irrelevant, may not be beneficial and may not be needed!
Apple was reasonably public (for Apple, that is...) about their search for a new development partner. They were obviously looking for a high volume manufacturer. So much so, that, they even interviewed with Intel. Talk of Motorola's inclusion of future developments, during this time, had atrophied to practically nil. Enter an article I found on theregister.co.uk, last summer.
In June or July I read a report that AMD was entering the Power PC fray. This caused confusion and exhileration combined for me. I had, just two months earlier, caught an article hinting that Motorola was out of the Macintosh picture for the future. The catch is that I can't dig up the AMD article anywhere, now! AMD's site has had nothing to say about the matter, either.
This isn't out-of-line with Apple's practical gag order level of secrecy, however, over developing tech. I, actually, agree with such policies, anyway. Plus, siding with AMD over Moto would make sense. They have made the switch to .13 micron production and are preparing for .10. Moto has had trouble with this. AMD is also producing copper-on-silicon chips and has been for almost 2 years. Intel has just gotten that worked out this past August (and only for a variant of the Pentium III). AMD's proven mass-production capabilities would be a better match for IBM's design work than Moto could provide. In addition, there is the very announcement of PPC compliance.
Since Apple and IBM jointly developed the Power PC specifications, nobody can just start producing chips in the name PPC compliance. They almost have to be invited. It's already a known fact that the Sledgehammer chip is a true 64 bit RISC processor, much like the G3. Add to that the fact that AMD can't/wont' afford to involve themselves in a venture of PPC compliance without a reasonable guarantee of return profit, it's resonable to say that some agreement has been reached. But, exactly WHAT?
I'm guessing that the new development agreement will comprise IBM, Apple and AMD tackling research, development and mass-production for future Mac processors, Power PC and beyond (?). I expect that Motorola will be kept on for embedded processors in smaller products like the Air Port, iPod and whatever forthcomming digital media appliance Mr. Jobs has lurking in the rafters (ripening on the branch?).
I think this was the register article you mentioned. But this was back in '99 when Motorola's cash flow was nice and healthy, so it's unlikely to happen now.
I belive you are wrong about this. The 'Hammer' family are based on AMD's extension to the x86 architecture so-called x86-64. This is not a pure a RISC design by virtue of the fact it must be backwardly compatible with x86. Athlons and Pentiums are CISC/RISC hybrids they integrate the best features of both design philosophies.
I hope you right about a new development agreement between Apple and IBM , but I don't believe AMD will be involved.
Sadly, no. That was not the article. It was listed as a headline article on their main page, at the time. A friend of mine (major AMD/Linux fan until I ran Mandrake on a dual G4 500MHz) had messaged me at work about something else on their site. I would imagine it had to do with the Hammer family of processors or the Athlon out-performing the PIIIs. The article was current for the time. It annoys me that I can't find it now. I personally consider it capitally important. Not just for this discussion; but, for the Linux power user community. After seeing how frighteningly well Mandrake 8 did on dual G4s, while itself being optimized for the G3, I'm inclined to throw rocks at the highest clocked Pentium/Linux (Lintel?) systems. And, I'm not as impressed with the Athlon systems anymore, either; but I'd still take it over the P4 any day.
As for RISC/CISC debate about the Sledgehammer, it turns out we are both correct.., sort of. AMD desktop processors have been RISC at their core since the K6. They, simply, tacked on Intel's MMX instruction set to it. Case in point, here. In this article they explain "The K6 was built around the original NexGen 686 core which was inherited by AMD when it bought the small firm in 1996. This allowed AMD to recover after its problems with its own K5 processor and inital difficulties in the development of its "in house" K6 design. NexGen had developed a powerful "six issue" RISC processor core which gave the chip impressive sixth generation performance. AMD enhanced this design by the addition of Intel's new MMX instructions, allowing the K6 to be compatible with the latest software". This having been said, they took a good RISC processor and made it Microsoft friendly, in essence. Effectively speaking, the Sledgehammer, being their 64bit extension to the the PC specifications' lease on life, is a mated pair of 32 bit RISC chips with extended instruction set. I have been looking and cannot find any indication that the Pentium chip is RISC. This would explain the performance matches from slower Athlons. I haven't bothered with researching the Itanium as it doesn't interest me and is steadily losing ground due to the severe and unacceptable production delays. A company the size of Intel shouldn't be experiencing this kind of apparent incompetency. No matter what, though, I was incorrect to call the Sledgehammer a true RISC processor. I wonder, though, how hard it would be for them to ditch the excess instructions and get down to business with those bad boys.
The irony to all this is that, both, AMD and Intel do produce various dedicated RISC processors of varying degrees of complexity. Though, they don't appear to rival the complexity of IBM's chips. Of course, if complexity were the only meaningful criteria, then, we'd have to look at Sun Microsystems' or Hewlitt-Packard's RISC chips.
No matter what, Apple seems intent on the third wheel concept for their little party. It's a shrewd safety precaution. That way if your main vendor's factory craps out -like Motorola's did, at launch- they can immediately turn to the second vendor and start taking up the slack -like IBM did. Now, if they could get "Big Blue" to establish a dedicated Mac chip factory, like the G4 factory they built for Nintendo, this wouldn't really be an issue. The trick is to A) convince IBM they can profit, quickly, from this; and B) Turn a profit, quickly, for their trouble.
What with the PC market's over-saturation and no major PC manufacturers reaching their targeted earnings, this quarter, now seems as good a time as any for IBM (or whomever) to invest some trust and hope in Apple. So far, they are the only one reaching (actually exceeding) their projected earnings.
I have thought quite a bit about this production issue. I developed a list of companies I thought would be a good replacement for Motorola. AMD was in a tie with Texas Instruments for my first pick. Both companies have the facilities in place and ready to go. Both have approximately the same length of history with similar degrees of success and quality. And, neither one is a shrinking violet when crunch is on. TI kind of one-upped them, IMHO, as they dared to produce a "whole widget" (sorry Mr. Jobs) in the '80s. People my age learned programming on either an Apple II or some version of a Trash 80. AMD has had good, recent, success with economizing their operatons and keeping quality control equal to or greater than thier direct rival. TI, on the other hand, has been making serious headway in the DSP chip market. They have a better feel, I think, for maximizing on the silicon being used. The only DSP vendor they haven't outright trounced is my second place contender, Analog Devices Inc.. ADI is the manufacurer of various types of high end DSP chips. They are most well known in the electronic music circuit for their SHARC chips; a fantastic, gutsy little processor. They are a direct rival to Morotola and their Cold Fire DSP unit. This rivalry extends to the cellular phone industry. ADI managed to win the rights to 3rd generation cell phone DSP specs. They have cut cell phones' physical depth to 1/2 of last year's designs thanks to the chipset they developed. Samsung was the first cell phone vendor to get on board with this new standard and has, already, been selling the new phones for a little better than a year in Japan. Anything past these three, I think, would be a dime-a-dozen.
"establish a dedicated Mac chip factory, like the G4 factory they built for Nintendo "
your talking about the gamecube right? i knew it was PPC but had no idea it was a G4!
The game cube does not use a G4, it uses a custom PPC chip called Gekko.
I take artisan002's point about RISC. Both the Pentium(s) and Athlons/K6 take CISC instructions and decompose them into RISC-like instructions which are executed in a 'RISC' processor. This was why I mentioned them being CISC/RISC hybrids. This CISC instructions stream is broken up into a RISC instruction stream.
Hammer will not be able to ditch the CISC instruction in the current generation because it's major seeling point is backwards compatibility. However the 64-bit instructions within Hammer could be entirely RISC. This means that in the future when all softawre is 64 bit Hammer could ditch the X86 compatability and x86-64 would be a fully RISC system.
As for the Fab partners you suggested, I don't really know much about them, except that Samsung Fab's Alpha's, and now that COMPAQ ahs effectively killed off Alpha, they could be looking for a new partner.
I don't like to say this but maybe it's time to move on... and dump the PowerPC... and move onto the Gigaflop PC... I mean apple phased out 601 and other processors earlier, moved to powerpc, normally I would think it would be time to move on...
But that I think won't happen. Look at the perfect example of Apple, they still have "one" line called the Macintosh which expanded into two sublines, Powermac & the iMac. Other companies like IBM have different PC's that branch from different categories.
"I don't like to say this but maybe it's time to move on... and dump the PowerPC... and move onto the Gigaflop PC... I mean apple phased out 601 and other processors earlier, moved to powerpc, normally I would think it would be time to move on..."
A few thoughts:
The macintosh line started with Motorola's 68000 processor, and proceeded on to the rest of the 680x0 processors as they came out. The PowerMac began with the PPC 601 and eventually came to the current PPC 7450.
The architecture of the PPC allowed for emulation of the 68k. The current PPC's are still of the same architecture as the original PPC 601, just several derivatives later.
There is no way to emulate PPC on a PC with any kind of reasonable speed. The reason why Apple was able to transition to the PowerPC was because their PowerMacs could run all of the 68k code, but it ran native PPC code faster. (Heck, even the MacOS wasn't PowerPC only until version 8.5, long after the last 68k mac!) To make a switch to the PC would require *everything* to be recompiled for the PC. There would be no way to run *any* of the Mac's current software. It would go from having several thousand software titles to zero.
It's easier to phase out older processors slowly when your new processors can run the older processor's code. To make a jump to the PC would be nigh to impossible because it would be a very sharp transition, without the ability to run older code.
(Also, what gigaflop PC? My Athlon 1.2GHz clocks in around 400 megaflops - on it's best day, that is... and since Athlon's are better then Intel P4's megahertz-to-megahertz, I would doubt any PC is actually operating at 1 gigaflop!)
You have to remember when comparing RISC processors to CISC processors that 'flops' are two very different things between the two systems.
RISC is 'Reduced Instruction Set'
CISC is Complex Instruction Set'
1 'flop' on a CISC processor may multiply 6 times 10 whereas the RISC processor may have to add 10 to an accumulator 6 times, using 6 'flops' for the same result. This is only a simplified example, but it illustrates the basic difference and why 'flops' are not a meaningful comparison when architectures are so different.
The RISC processor may be able to feed the 6 operations to multiple execution units in parallel and beat the CISC timewise, or the CISC may be executing parallel CISC instructions itself and win.
Either way, IMHO, a 'flop' is as meaningless when comparing differing architectures as 'MHz'. A CPU with 100 execution units operating at 10 MHz can do the same work as a CPU with 10 execution units running at 100 Mhz AND that assumes that the execution units are the same and doesn't take into consideration such things as pipelines, caches, etc, etc.
I think the CISC/RISC debates are mute now anyway. All the current processors are hybrids. Most x86 systems are using multiple RISC cores to execute CISC instructions, and 'ALTIVEC' is by no stretch of the imagination a 'RISC' thing to do.
IBM is the best manufacturer for Apple, Motorola has done nothing but screw them over.
Why do away with the PowerPC alltogether?
I don't think it's time to scrap the PowerPC....not as long as there is still potential there for it to grow....which there obviously is. I think there's more of a problem with manufacturing on Motorola's end which is causing a lot of holdups for greater speeds and whatnot. There's been a lot of discussion about IBM manufacturing future chips....why wouldn't that be a better solution?
The G4 has finally made it's way into the consumer end machines. If we should see a processor get scrapped it would have to be the G3 first.....but that would only be once we see the G4's predecessor.
No way should Apple scrap the PPC. PPC architecture is releatively new and has plently of potential for growth, changing architectures is a desperate measure.
The problem is not with PPC architecture, but rather with Motorola's ability to get the 'goods' developed and ramp up production. The PPC should have been a 1GHz before Pentiums et al. but Motorola mis-manged its development.
But you have to ask yourself why should Motorola push themselves? Apple is almost dependent on Motorola so Motorola are really in control, sure Apple can complain but theres not alot they can do.
If Apple were to take IBM on as a second supplier however, Motorola has a bit of competition, it will either drive them to be more committed, or drop out, in which case IBM supplies Apple.
Oops, my mistake. I don't mean true FLOPS as in processor-level flops, I mean high-level flops. As in, how many adds per second can a program expect? How many multiplies per second? Etc? THAT is a meaningful comparison. When I say my Athlon rates 400 megaflops, I don't mean processor-level flops. I mean high-level flops. Although the technical term "flop" refers to the processor-level, "flop" has started to have unofficial low-level and high-level definitions.
from what we're hearing about Motorola's troubles....maybe we'll see them scrap production and another suitor will take over?
how much of a chance does IBM have in completely taking over PowerPC production for Apple?
I'm guessing Motorola's not ready to throw in the towel just yet. It really depends well the G5 alpha silicon is doing. Maybe they've got there act together and the G5 will ramp up nicely.
Either way I think Apple should take on a second supplier, and IBM seems the most logical partner.