IBM, Motorola sign on to single PowerPC chip

Toddius Maximus writes "The on-again off-again partnership between IBM and Motorola to design a single PowerPC architecture appears to be on again. The companies made a joint announcement Wednesday that they will work together on an embedded version of the PowerPC that will be targeted at networking and telecommunications devices.Story here. "

Current fabrication?

[Anonymous Coward]

1. These sorts of deals tend to have some technology swaps on the side that may be the real reason for it all.
2. My understanding of the fabrication practices of manufacturers of chips is that they basically fill up a wharehouse with chips and then sell them off--often emptying the wharehouse years after the end of production. If Motorola was, say, running out of 68040 chips and wanted to make more for some customer it might very well have to in effect design the chip from scratch for a new process technology. Hence, why not use something from the PPC line instead that can be made more economically using existing fabrication facilities (perhaps a foundry that is .22 nm needs work as things move to .18 nm foundries for their cutting edge work, etc.) And there would be the possible performance advantages of the PPC core versus the 68xxx core, etc.

3. I can't wait for my 800 MHz PPC governed toaster! But seriously, there are some embedded applications, like laser printers, where the PPC could really help. When you up pixel count, you really increase the computational demands, for instance, and this is an area where PPC rules for a given clocking. Why use a foundry process which should give you a 500 MHz capable core for a core whose performance doesn't scale with clock speed.

AFAIK nope, it doesn't

[Chris Johnson]

For low power you're basically looking at the 603e- I'm not sure what they have of more recent lineage which is comparable to the 603 but I do know that I raised the issue of really low-cost Macs a long time ago, believing that 68040s were cheaper than PowerPC. Turns out they are not because PPC was where the volume was.
By the same token, I wouldn't assume 68K is lower power than PPC- I'll grant that it seems possible it could be, but I wouldn't automatically conclude it was. I don't actually know what Coldfire is- is this some sort of StrongArm-ish thing? The thing to remember about PPCs is that they aren't much like Pentia- they don't consume anywhere near as much power and they are less of a disposable part (I saw a fellow on a mailing list today, talking about celerons and overclocking, who basically gave computer parts a lifespan of a couple years before replacement- something I find hard to understand, though it's possible he _didn't_ mean 'then it burns up from normal use')
Either Coldfire or the PPC runs on a lot less power than modern x86, but that doesn't mean the PPC uses that much more :)

Re:Enh

[jafac]

--- moto DOES need to "get with the program" on clock speeds. They're seriously behind intel in raw clockage, and they've got to realize by now that no benchmark is going to convince anyone that their CPU is faster if the megahertage ain't there.

Lies, damn lies, statistics, and benchmarks.

-The number of suckers born each minute doubles every 18 months.

Re:Enh

[Millennium]

'Course, I'd rather have both - I won't deny that Altivec is cool, I just think that higher clock speeds would be better overall.

Why? Clock speed, when you think about it, isn't much more than a number. You can't compare across architectures, and you can't even really compare across chips within the same general architecture. Quality, not quantity, is what's important.

I'll grant, most people are very dense when it comes to measuring chip speeds based on clockrate; they think a MHz is a MHz is a MHz, no matter the chip (therefore, a 300-MHz Celeron would be as fast as a 300-MHz PIII, if such a thing existed, and we all know that's not true). I see this as nothing but proof that clockrate is an inappropriate mesaure of speed.

Re:Enh

[cpt kangarooski]

Yeah, I know all that. But unless you already do know it, then all you think about is the clock.
Ever wonder why Apple switched from calling it's processors by name (e.g. 68040, 604e) to vague psuedonames (i.e. G3, G4 - which are not the actual processor names). I think it's to compete with the Pentia. The G3 sounds better than the P2. The G4 sounds better than the P3 (and pretty certainly is too, software notwithstanding)
First, higher MHz does yield a performance benefit all across the board, without having to rewrite software (which tend to have their own speed governors if they really need them). New instructions are also good, but do not provide an immediate benefit. They require effort on the part of programmers. Like subprocessors (DSPs are the best example, but there are others) their popularity waxes and wanes. For decades there's been oscillations between speeding up systems through higher efficiency CPUs, or speeding up systems through lots of ancillary chips. I like the idea of software 3D accellerators better than HW ones. At the moment this is not the best way to go, but at some point it probably will be. Then it'll become too much overhead, and we'll be back to chips.
I like the Altivec stuff, and I think it will make the system better. But marketing is important. If no one buys a good system because of confusion (..cough..Mac..cough..Amiga..cough..Connection Machine..cough) then it's screwed. A faster clock and continued high percieved performance (including bencharks and other lies) would probably benefit the mac more than an MMX like set of instructions. Even if they kick ass. Again, I'd _really_ like to have both. And I realize that a lot of the support for Altivec comes not from programmers per se, but just recompiling with the new libraries and such. (fine tuning can't hurt though)
Having sold Macs before, I know that it is annoying as hell to get someone clueless to believe that a slower clock is not necessarily bad, as all other things are not equal. Same goes for multiple processors - Intel is probably annoyed at the complexities added by P3 vs. P2 vs. Celeron. Trust me, if AIM brings out chips that are faster than the P3, similarly priced or better, with good performance they'll sell a lot. Slower, higher performance chips, similar cost will not work quite as well. (A cryo-cooler option would be nice !_!)
So go Motorola! Go IBM! Go Apple! Go banana!

Enh

[cpt kangarooski]

The announcement is about an embedded chip. That's nice but I'm more interested in the larger processors... G4 and such. Honestly I side with IBM, in that faster chips are better (that is in that it will speed up all operations w/o rewriting anything & makes for better marketing) than additional instructions. 'Course, I'd rather have both - I won't deny that Altivec is cool, I just think that higher clock speeds would be better overall.

Re:RS/6000

[Xunker]

Depending on model, but I believe the Northstar boards are part number 1125-AA565. Don't quote me though.

Re:AIX on iMac

[Xunker]

Thats as laughable as running MacOS on an RS/6000 or AS/400. Gee, I wanna see OS/400 on my daVinci. :)

What about...

[sinator]

This partnership seems nice on the exterior, but what about Motorola's Coldfire line of 68xxx processors for embedded work? That's a big money maker, especially in *Motorola* cellular phones and other embedded processors. I can see the motivation perhaps in the realm of interoperability with Bigger Computers (tm) but AFAIK the Coldfire runs on a lot less power than the PowerPC for comparable performance (remember, these are embedded apps)...

QNX might be a better choice.

[Thag]

Linux ports to practically everything, so it's probably doable, but AFAIK, QNX is probably a better choice for an embedded systems device. It's expressly designed for that purpose, and it's tiny.

Jon

Re:QNX & Linux

[joetee]

Neutrino will be used as the kernel RTOS on future Amiga computers. a VERY crisp response to user input is expected on an Amiga. Having a small un-improvable assembler core, just for that CPU, yields the best task switching, with the least waste of cache. Fitting the entire kernel in cache, and leaving room for apps, is a good thing. It's something like the old rule of:
It's slower to run off of tape, then off of disk,
then off of memory, than off of cache, than in all_those_registers. Thats where Amigas heading.
Praise Dr. Havemose!

The article's a little off

[webslacker]

It doesn't mention the new architecture's name or main purpose.

The architecture's name is Book E, and it was designed to allow IBM and Motorola to make their own customizations and still stay compatible. Embedded processors were the main drive for Book E, but it's also useful to the PowerPC processors for use in PowerMacs. However, the effects of Book E will not be felt to Mac users until after the G4 processors have collected a little dust.

On the other hand.

[Gumber]

So, perhaps we should drop hardware FPUs too and put that silicon into improving integer performance?

We will have to see, but I think things like AltiVec and KNI are probably a wise use of silicon in that they can improve the performance of some of the most computationaly intensive tasks likely to run on a microprocessor these days.

Look for more Baby ATs

[ripcrd]

You get this group to start making those PC's on a chip, maybe buy out Cyrix or IDT and you got yourself the next wave of small computers. Or they could make a deal with some motherboard mfg. to use their PC on a chip (CPU, chipset, etc.) on mobo w/sound, video, SCSI, 56K or cable modem built in and we're talkin' cutting edge, small footprint computing. I for one would like to clear some crap off my desk (work and home). They could add an AGP slot for upgrading video and totally drop the ISA slot. Maybe two or three PCI and an AGP. And make it a sub $250 system. Wishful thinking.

Re:AIX on iMac

[NII Link]

Actually, that might not be so farfetched. AIX isn't only for supercomputers. Linux runs on the iMac, so does OpenBSD and MacOS X (obviously). There's really no reason that AIX couldn't run with some modifications.