What would an Apple designed processor look like?
And would it be in their best interest? Apple has the technical know-how. They've been designing custom asics since the early Mac days, custom peripheral controllers since the first PCI Macs, and more recently, complicated memory and PCI controllers for their computers. And any thing else they may need can be easily acquired. Obviously they would need a manufacturing partner but there are many Taiwanese and American companies that would be appropriate such as UMC and TSMC that have 130nm plants and plans for smaller (90-70nm) in the near future. My questions are, would it be in their best interest economically and strategically? And, is Apple working on this as we speak?
If Apple was working on a processor to replace the G4 it would look something like this:
Processor core based on Book-E architecture.
4 32bit integer units
2 64bit double precision FPU?s
14 stage pipeline
40bit memory addressing (up to 1TB or addressable RAM)
128 Altivec unit
16+16KB L1 data and instruction cache (1 cycle latency)
768KB L2 cache on die connected to core with a 256bit wide data-path
RapidIO or HyperTransport bus to motherboard 500MHz
DDR-SDRAM or DRDRAM(Rambus) memory controller on die
DDR support ? 266MHz or 333MHz
DRDRAM support ? 800MHz or 1066MHz
Memory controller interfaces with processor core at half core speed DDR (effectively full core speed) and 128bit. External interface with memory (DIMMS) would be 64bit (to reduce amount of pins on DDR-SDRAM) if you use DDR and 32bit if you go with dual channel Rambus.
Made on 130nm process with SOI and Lo-K dielectric.
Speeds: 1.6 - 2.6GHz
I don?t think for a PowerMac Tower they would worry too much about power dissipation, but I would imagine a processor like this would use about 30-35 watts at 2GHz. For iMacs and iBooks Apple can stick with the 7455/45 and the PowerBook might make due with a lower power version (running at 1.4GHz and 1.6GHz) of the above processor with about 14-18 watts being used by the processor.
[ 02-05-2002: Message edited by: Outsider ]</p>
If Apple was working on a processor to replace the G4 it would look something like this:
Processor core based on Book-E architecture.
4 32bit integer units
2 64bit double precision FPU?s
14 stage pipeline
40bit memory addressing (up to 1TB or addressable RAM)
128 Altivec unit
16+16KB L1 data and instruction cache (1 cycle latency)
768KB L2 cache on die connected to core with a 256bit wide data-path
RapidIO or HyperTransport bus to motherboard 500MHz
DDR-SDRAM or DRDRAM(Rambus) memory controller on die
DDR support ? 266MHz or 333MHz
DRDRAM support ? 800MHz or 1066MHz
Memory controller interfaces with processor core at half core speed DDR (effectively full core speed) and 128bit. External interface with memory (DIMMS) would be 64bit (to reduce amount of pins on DDR-SDRAM) if you use DDR and 32bit if you go with dual channel Rambus.
Made on 130nm process with SOI and Lo-K dielectric.
Speeds: 1.6 - 2.6GHz
I don?t think for a PowerMac Tower they would worry too much about power dissipation, but I would imagine a processor like this would use about 30-35 watts at 2GHz. For iMacs and iBooks Apple can stick with the 7455/45 and the PowerBook might make due with a lower power version (running at 1.4GHz and 1.6GHz) of the above processor with about 14-18 watts being used by the processor.
[ 02-05-2002: Message edited by: Outsider ]</p>
Comments
I doubt it will change in the future, Apple is not the only customer of PPC chips, the chips are developped for a bigger market than Apple
. I a m not a specialist, but microprocessor chips are much more complicated than asics or even graphic chips (even if the lattest of them are hybrid including geometrical acceleration), they clock also at higher mhz rate, (to my humble advice there is not any ghz graphic chip in the market), all this features demand an expertise skill. I doubt that Apple got it.
Apple choose to work in the AIM alliance, we do not know what is their role in this alliance. I believe personnaly that Apple is involved in the direction where he wish that the chip evolved (example: i wish that the G5 will be based upon IO architecture in order to work with the new mobo i want to design, i want some new specific instructions in order to improve some performance ...), but i doubt that Apple is personnaly the designer of the chip. I have no proof of this , but if i take an image i'll say Apple is like a customer that ask to an architect to make the house of his dreams, and not simply a customers that will buy a house already designed.
Those "specs" you pulled out of thin air are just that... out of thin air. As was stated, microprocessor design is really complicated and how can you even know enough about it to generate a spec list from nothing?
G-News
It would look like the G5 - Apple working with it's own chip designers, as well as Motorola's, have designed a next generation chip to take over for the G4.
[ 02-06-2002: Message edited by: Aphelion ]</p>
I used to think that on-die memory controllers were a good idea, but I now realize that memory performance will never keep up with processor performance, so there's really no point. Put the effort into designing a super duper cache system and use HT or RapidIO to interface with main memory.
It'd be white with Ti trim.
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
I can see people avoiding Apple's chips, because Apple and Jobs can be, uh, mercurial. They could be left out in the cold wen Jobs decides to change.
So - how expensive is it to design chips? Would it be worth it for Apple to do it, assuming that no one else buys them?
How do you know this as fact? Besides MOSR and theregister we have no concrete info.
Those "specs" you pulled out of thin air are just that... out of thin air. As was stated, microprocessor design is really complicated and how can you even know enough about it to generate a spec list from nothing?
I know I pulled them out of thin air. I didn't claim this as truth but rather what a processor based on the PowerPC ISA designed by Apple for desktop use would look like. It is pure fiction, but maybe Apple should take the lead in it's own future. They can design it and have someone else manufacture it.
<strong>If MIPS and ARM can do it, Apple can do it (design their own processor, that is).
</strong><hr></blockquote>
i'm not so sure. SGI is bleeding to death under the costs of MIPS development. according to the register they're even interested in the G5 as a potential MIPS replacement. i don't think apple has the resources to really be a player in microprocessor design. it takes too much money and when you're through it ends up costing more because you're gonna sell fewer of them. apple's best bet is to find a patron (like IBM) that can fabricate whatever they want easily and pay them to do just that.
<strong>If MIPS and ARM can do it, Apple can do it (design their own processor, that is).
</strong><hr></blockquote>
i'm not so sure. SGI is bleeding to death under the costs of MIPS development. according to the register they're even interested in the G5 as a potential MIPS replacement. i don't think apple has the resources to really be a player in microprocessor design. it takes too much money and when you're through it ends up costing more because you're gonna sell fewer of them. apple's best bet is to find a patron (like IBM) that can fabricate whatever they want easily and pay them to do just that.
<strong>
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" /> </strong><hr></blockquote>
Michael Grey, could I add your mock-up to my web site (http://homepage.mac.com/benvp/)? Please eMail me at [email protected]
Thanks.
i'm not so sure. SGI is bleeding to death under the costs of MIPS development.</strong>
SGI was and is bleeding to death because they had that Microsoft shill Rick Belluzzo as CEO for awhile. It's not because of MIPS development. When your company loses its core competency (graphics), makes stupid acquisitions (Cray), and tries to move the platform over to Intel (first Pentium than Itanium) from MIPS, I would imagine there was no MIPS development going on whatsoever except for doing what is necessary to fab MIPS R12000 chips on more advanced technology. Ie, SGI planned on phasing out MIPS for Itanium. They lost.
On the other hand, Sony is doing a bang up job with the Emotion Engine and Sibyte (now Broadcom) is shipping (I think) a MPC 8540 competitor a year before Moto even ships the 8540. So, CPU development isn't much of an issue. About 60M to 80M dollars per year.
<strong>according to the register they're even interested in the G5 as a potential MIPS replacement.</strong>
The Register is no better than AI. If SGI wanted to move away from MIPS, Itanium is the right way to go. Problem is that SGI is just a shell of company right now. They've got nothing in the tank. MIPS was separated from SGI in 1998. All the graphics people left or were moved to Nvidia. Cray was sold for a piddling.
<strong>i don't think apple has the resources to really be a player in microprocessor design.</strong>
No one's saying Apple should enter the fab business. They design the CPU. Intel, IBM, UMC, TSMC, TI, AMD, and Moto can fab it.
<strong>SGI was and is bleeding to death because they had that Microsoft shill Rick Belluzzo as CEO for awhile. It's not because of MIPS development. When your company loses its core competency (graphics), makes stupid acquisitions (Cray), and tries to move the platform over to Intel (first Pentium than Itanium) from MIPS, I would imagine there was no MIPS development going on whatsoever except for doing what is necessary to fab MIPS R12000 chips on more advanced technology. Ie, SGI planned on phasing out MIPS for Itanium. They lost.</strong><hr></blockquote>
Yeah, SGI is really just another victim of the compression of scale in processor - hmm, actually hardware - performance. The gap between SGI's high-end workstations and the massing herds narrowed so much as to make its position impossible to maintain. When you're developing your own hardware, processors, graphics chipsets, operating system, and (at one time) the software that almost 100% of machines were bought to run, you can't compete with systems constructed from off-the-shelf components. Hmm, sound familiar?
(Though this isn't really on-topic, SGI was also crippled by some seriously questionable management, bad business decisions - like the buyout of Cray - and the loss of most of its talent).
[quote]<strong>On the other hand, Sony is doing a bang up job with the Emotion Engine and Sibyte (now Broadcom) is shipping (I think) a MPC 8540 competitor a year before Moto even ships the 8540. So, CPU development isn't much of an issue. About 60M to 80M dollars per year.</strong><hr></blockquote>
It depends. $60 to 80 million annually is a lot if the processor is only going to end up in your own products. R&D plus third-party fab costs would be more, I suspect, than Apple currently pays out for supplies of G3s and G4s.
Apple would need to know that the performance of a homegrown chip would be enough to attract new customers to Macintosh.
I think it would also have to look at selling the chip - or some derivative - for purposes other than computers.
Sony's Emotion Engine is a good example - it was developed for the Playstation 2 (and apparently scaling for 3 and 4), but was designed (I assume from the start) to be a huge source of income for Sony through sales to third parties.
[ 02-06-2002: Message edited by: Belle ]</p>
If they don't have the talent to design CPUs in house right now, they can get it pretty easily. It might even be worth it for Apple to take over entirely, but I doubt it. I imagine they're still working closely with Motorola and IBM, as they have been since the AIM alliance was founded.
"If SGI wanted to move away from MIPS, Itanium is the right way to go."<hr></blockquote>
That's an interesting statement. A lot of what I read in these forums slams the Itanium pretty severly.
It only clocks to 800MHz
Performs at the equivalent of a PIII
etc.
Am I being lied to?
<strong>
That's an interesting statement. A lot of what I read in these forums slams the Itanium pretty severly.
It only clocks to 800MHz
Performs at the equivalent of a PIII
etc.
Am I being lied to?</strong><hr></blockquote>
I may be mistaken, but I take it those comparisons with the PIII are in respect to running IA-32/x86 code which is handled by hardware emulation. Code compiled for IA-64 will perform somewhat better.
And don't you listen to the Apple hype? MHz are an outdated measure of performance.