Apple and IBM Working Together on 64-bit CPUs

Currawong writes "eWeek reports that IBM Microelectronics is working with Apple on a 64-bit PowerPC processor called the GigaProcessor Ultralite (GPUL). Unlike previous reports, eWeek now reports that Apple is testing the chip for use with future hardware. IBM apparently also plans to use the processor in linux-based servers. It's believed IBM will disclose some details of the processor in October at the upcoming Microprocessor Forum in San Jose, California. While this story is similar to recent stories about Apple using Power4-based IBM chips in future Macs, the GPUL, unlike the Power4, is smaller, runs cooler and consumes far less power, making it suitable for desktop machines and small servers. The processor is described as having the same 8-way superscalar design fully supporting Symmetric MultiProcessing." We had a previous story about these new chips.

Comments

Cooler?

[greenhide]

the GPUL, unlike the Power4, is smaller, runs cooler and consumes far less power, making it suitable for desktop machines and small servers

Does anyone know if the chip would actually be cool enough so that it would not require a fan? One of my favorite features of the G4 is that it requires no fan whatsoever. My PowerMac G4 makes so little noise that sometimes it's hard to tell if its running or not without looking at the little glowing power button on the front.

I think this is one of the nicest features of Macintosh computers and if they need to add a fan I think that will be a real shame. On the other hand, Motorolla really hasn't gotten their act together, so Apple may not have a choice.

Re:Cooler?

[Lumpy]

BAH! the requirement of a fan on the processor is based on very poor heatsink design. remember a small plastic fan is cheaper than a large block of copper and aluminum. and ANY processor including the cook-your-egg AMD's can use a fanless heatsink IF the heatsink is properly designed and sized, AND your case has a poper heat chimney and vents in it's design so that convection will promote cooling.

Using fans is the cheaters way out or the cheap way out.

Will it have DRM built-in?

[sphealey]

A key question: will this chip have DRM (aka Digital Rights Reduction) features built-in? If NOT, there could be a good market here for IBM as the free alternative to Intel.

sPh

Re:Will it have DRM built-in?

[SmittyTheBold]

If Apple sticks to their old game, there will be no DRM whatsoever.

After all, iTunes rips audio into MP3 formats instead of some "protected" format. QuickTime does not (IIRC) support DRM, except for (weak) protections on streamed movies to prevent a person from saving the movie.

Apple has made a market by keeping a user's options open. Closing that up is not a priority for them. The infrastructure to do such things is not only not there, it would take a lot of time to implement. I am sure Apple is more interested in getting a new processor to market than they are in restricting the rights of their target market - content creators.

Re:Will it have DRM built-in?

[gclef]

I was at a talk recently given by one of the security guys from Apple. He was asked about the whole TCPA thing, and his response was that Apple wasn't participating in it at present, and didn't really see what they could offer to it. Unless some sort of TCPA-like thing became law, or unless someone came up with some way for Apple to contribute, they were going to stay out of it.

So, at least for now, they're staying out of the DRM wars. Of course, this is all subject to management whims, but that's the state as of now.

I can see why Apple hates rumors

[Anonymous Coward]

I can really see why Apple hates rumor-mongering like this. They go through a lot of trouble to get a machine design done and out in the marketplace, and two weeks later someone posts a rumor somewhere saying "G5 systems will be announced in three months!" so the user goes "well, I was going to buy a new machine, but I don't want to get screwed so I'll wait for the G5".

This chips' project doesn't even complete until summer 2003, that doesn't even imply it'll be ready to fabricate or be in any kind of production then, even if it DOES pan out to be a useful design. I imagine by tomorrow Macosrumors will be touting it to be in the new uber-G4 to be released next month.

How long has the G5 been 'almost ready' as far as rumor sites go? Two years now? It's great to spin up your readership with crap like that, but it really does a disservice when it's untrue.

Re:I don't see the landscape changing too much...

[Matthias Wiesmann]

(although, if you want the most speed, you'll eventually need to upgrade to the 64-bit versions of your apps).

Why would code that uses 32 bit pointers be slower than code that uses 64 bit pointers?
Having 64 bits pointers is needed to address more than 4 gigabytes, but why would there be a performance gain? I would think that longer pointers imply moving more data into the CPU, and therefore would consome more memory bandwidth. Am I missing something?

You're kidding, right?

[bill_mcgonigle]

"New processor Z has just been released. Sources say the processor is so fast typical users won't have a need for it, but is expected to be popular among engineering and CAD users."

I first started reading this line when the 386/25 came out. Replace CAD with 3D Graphics for this decade. Every time a new processor comes around, they say almost exactly the same thing - watch for it in the press. So far the prediction hasn't shown to be true.

Re:You're kidding, right?

[starseeker]

Actually, I'd argue for most users the Pentium II was the point where things got fast enough to be usable. For paper writing, email and web, a Pentium II will do just fine. I know because I've been able to do all of these on a Pentium 200, which is significantly slower. (granted I was using Linux, but still.)

Where processor speed helps in my experience is a) heavy duty mathematical software and b) compiling software. For graphics, acceleration cards do far more than a processor upgrade, and memory is also a common bottleneck (or was - with the really cheap memory we have now I suspect it's less of a problem.) A fast processor can help if you have lots of excess toys running, but for doing your job the Pentium II was when that task was effectively solved.

There is a reason the computer market is saturating. People don't feel the need to upgrade so much. If they upgrade their software, it may demand more resources, but people don't feel the need to use Office XP or whatever if 97 does the job. And despite what we all think of Microsoft, it does do the job. Hence Microsoft's consideration of subscription licenses - their revenue stream is likely falling off somewhat, or at least not growing as fast.

Don't confuse Want with Need. From a marketing standpoint they may look the same, but they actually aren't. In a recession we notice that fact more.

Re:I'll believe it when it's on the shelf at CompU

[sphealey]

Again thank you Motorola for screwing us! I have a small feeling that IBM can be counted on a little more that motorola, because IBM sells its power pc based chips to more that just apple, where if I am not mistaken motorola only sells to Apple

The whole PowerPC thing was one of the most amazing displays of corporate loyalty I have ever heard of. Apple needed a new chip but was unwilling to abandon their historical supplier, so they forced IBM and Motorola to the table and knocked heads until they got a joint production agreement.

Most companies would have said: "sorry Motorola - you are out of gas. We just signed with Digital (Alpha) [or IBM or Intel]. Thanks for the memories". Instead Apple force-fed the entire PowerPC thing.

I wonder what their motivation was? And did Apple truely benefit in the long run?

sPh

Re:Shades of PowerPC

[ZigMonty]

This being Apple, one can infer the following future events:

1) That "state-o-the-art" Powerbook you just bought won't run the next version of the OS.

2) All of your current software will still work but in some sort of wierd "Compatibility Mode" that is ten times slower than it runs today.

3) Developers will get screwed (again).

Look, I'm sorry but I'm sick of these posts. The PPC instruction set was designed to be a 64bit architecture. There is a 32bit subset that all current mac programs use and Mac CPUs understand. Theoretically, running 32bit code on a 64bit PPC should be as simple as setting a bit in a special register in the CPU, putting it in 32bit mode.

In fact it might make sense to make 64bit mode an option to the developer. If they don't need very large integers or 4+GB of address space, they could use 32bit mode. This would mean that you don't waste RAM and memory bandwidth using 64bit pointers when you don't need them. The OS would still be 64bit of course.

All applications should run flawlessly (if they did before :-). There is no emulation. And even if there was, how would that hurt the developers? The only time Apple has switched processor architectures before was 68k->PPC. I can still run a 1984 68k copy of MacPaint in Mac OS X's Classic environment. Hell, their 68k emulator was so good that they didn't update all of the OS to PPC straight away! Yes, the jump from OS9 to OSX was difficult for developers but this wont be, even if Apple had to use some sort of emulator (which they wont).

Re:Shades of PowerPC

[hey!]

I think he's alluding, not to the OS9->OSX upgrade, but the 68K -> PPC conversion. The compatibility mode was, in my opinion, a tour de force, but they screwed developers by the Lisa Pascal to C switch. It wasn't just that C became the preferred development enviornment, it was because they decided not to support Pascal at all. This was a horrible miscalculation, because it put developers using what up to then was the preferred development environment at a huge disadvantage. They had to retrain their programmers and port their applications. This left some applications stranded in emulation land for two years or more.

I think he may also be referring to the death of OpenDoc, which badly burned many developers and for which I too still have not forgiven them. OpenDoc was brilliant and so, so close to being ready for prime time when it was killed. This was a one-two punch for many small developers -- once they spent perhaps eighteen months in their C conversion, they then spent another eighteen months or two years redesigning their application for an architecture that simply went up in smoke. I knew some small innovative software developers that had, perhaps, a two or three year lead over similar applications on the Windows end, who ended up behind, a place you simply can't afford to be if you are on a niche platform like the Mac. This experience soured many developers on Apple, and prepared many of them to be well disposed to open source.

Bitterness for past misdeeds aside, I expect a 32 bit to 64 bit conversion to go more smoothly than the 68K to PPC conversion, or the equivalent conversion on the Windows side.

Not quite the next best thing.

[Phoukka]

Except that the GPUL is not the next best thing. If you read the eWeek article, you'll find that the projected time-line reads, basically, the G5 first and then the next best thing after that. And it is very much up in the air what that next best thing will be. I know that Apple has had a long history of working with IBM and Motorola, and that adds a certain amount of probability to the conjecture that the GPUL will be the next best thing, but the existence of Apple's Marklar project shows that we cannot discount the possibility of a switch to x86 architecture. I think the most likely candidate within the x86 world is AMD's Hammer -- it will be available at desktop-processor-level prices, and will also be available in versions more suitable for servers. Since both markets are areas Apple has targeted, this makes the Hammer more appropriate than, say, a combination of Intel's Pentium4 on desktop and Itanium for servers.

Again, though, let me reiterate that this is all just conjecture until "The Steve" makes some sort of formal announcement.

Re:OS X only handles dual processors

[Graymalkin]

What are you smoking? The Darwin kernel can scale up to 32 processors. The 2 processor limit is definitely not in the kernel itself. It is actually a probably with the design of the G4. Instead of a point to point link to the memory controller the G4s are on a shared bus. Stick more than two processors on a shared bus topology like that and your overhead is going to eat any extra performance you can manage to get.

I don't think that means what you think it means..

[Rui del-Negro]

I think you've been playing with consoles a bit too much. 64-bit here refers to the size of general-purpose registers and memory addressing.

With 64-bit you can address over 4 terabytes. Do you feel the need for more than that?

You can also work with integers up to 18.446.744.073.709.551.615, and floating-point numbers up to 1.7976931348623158 E+308. Feel the need for more than that?

There are wider registers in the CPU (such as the dedicated SSE2 or Altivec registers), but for normal operation I think 64-bit should keep us going for quite a few years.

RMN
~~~

The funny thing is I'm going to wait for a G5

[BoomerSooner]

So it may be a long wait! I got my G4 Tower a few months ago to see if I even would like Apple OSes. To my delight I love OS X (hell even OS 9) and OS X is everything X-Windows/Linux should have been striving for. I was going to sell my G4 and get a dual 1.25 but the one I have is more than enough for now and 1 to 1 1/2 years isn't too long to wait for the next Mac (besides I've still got to save for the 22" display!).

I've tried to use Linux on the desktop since 0.98 (Slackware in '96) and never found it to my liking. I don't like to tweak and read man pages for hours, I just want the damn thing to work. That being said all my companies servers run Linux (killed the SPARC the other day) and being able to sftp/ssh to my servers from a terminal in OS X was great. Plus using Dreamweaver to do my JSP development makes a great environment.

Hopefully 1 to 1 1/2 years is all I'll have to wait. I'm patient so I'll start saving now.

Don't use FCP, do you?

[edremy]

G4 chips have more than enough "under the hood" to comfortable kick the likes of Photoshop and Illustrator around, not to mention the iApps, and everybody's favorite Final Cut Pro.

You have *got* to be kidding. Enough power for FCP? Dude, I routinely run 30+ minute renders for a 3 minute chunk of video on a 933MHz G4, and I'm not even doing all that much. A few filters, some text generation, a mask or two and it's walk away from the machine time.

Apple could be shipping 8-way 2GHz G4s and it still wouldn't be enough.

Re:Don't use FCP, do you?

[thatguywhoiam]

You do realize that this sort of thing took hours and hours on a $100,000 Avid previously. And now you're doing it on (approx.) $5k worth of Apple hardware with no special boards or drives.

I feel your pain, but let's get some real perspective. Video is almost always going to need some sort of rendering, especially when dealing with uncompressed (or nearly) video. That's upwards of 600K per frame, times 30 per second. Just for the data.

I used to have all these stats for explaining to clients why 'video rendering' always takes so long. My favourite: one minute of Cinepak (old-school!) video requires more math than the Apollo missions did. Sure, it's a whack stat, but it get's the point across, eh?

The G4 is no slouch. Realtime Video Everything requires a massive bank of DSPs, or a CPU that does not yet live.

new bus is the interesting part

[smagoun]

The really important part here is that Apple would be using a new bus with these machines. What the bus is doesn't matter so much as the fact that it's not the Maxbus, which is what the G4 and its ilk use. Maxbus is designed for routers + other embedded apps, not high-performance desktop computers. Currently Maxbus runs at 167Mhz, which is about as far as Motorola is willing to push it (167Mhz single-pumped, mind you). As a result, even a single G4 can more than saturate the bus, and the dualies spend a *lot* of time idling (they share one memory bus). Big caches help the problem, but there's still a fundamental issue.

Even if the new chips are clock-for-clock identical to the current G4, the mere fact that they're running on a newer bus will make the machines much more powerful.

For more info about this, head over to Ars and check out the posts in the Mac Achaia by BadAndy from earlier this summer ("Altivec, anyone?" I think it was titled). He knows a hell of a lot more about this stuff than I do; it makes for fascinating reading, and you can really understand why faster CPUs alone won't cut it for Apple.

Re:I don't see the landscape changing too much...

[Graymalkin]

There's little to worry about with porting of apps. Unless you've got some seriously processor dependent assembly in your PPC binary there's little that will stop it from running on a POWER chip. The PowerPC instruction set is a subset of the POWER one meaning POWER ostensibily has more instructions besides the ones PowerPC has. It is trivial to compile an app for generic PPC code that will run on every PPC chip you can find.

I don't get what you mean by the G4 "showing its age", it isn't some ancient chip pulled out of a tar pit. It's performance problems come from the low clock speed and the lack of multiple floating point pipelines. That is more of an implementation issue than an overall design issue. The Athlon has 3 FP pipelines, the G4 has one. AltiVec is fine if you can fine the parallelism it is good at in your code. Most people for go that effort and stick to simple floating point operations. Hence the Athlon's high floating point performance.

Please people, 64-bits does not equal performance, instructions per second is the important factor. With 8 way superscalar goodness the POWER4 design gets stuff done not with its 64-bit GPRs but the fact it can suck down multiple integer and floating point operations at once and out of order. You've got the potential of 4 FLOPs per cycle in the POWER4, at just 1.25GHz that's 5 GFLOPS of plain old floating point performance. That is twice the Athlon's performance at the same clock speed. A second core would effectively double that rate since the cores on a POWER4 share their L2 cache making them look like a single chip.

Re:Apple working on a CPU? Not likely...

[overunderunderdone]

Since when does Apple have any hardware engineers?

Umm... Since Woz started working in Steve Jobs garage? One of their divisions is the "Hardware Engineering Division"

Even their boards are outsourced

I'm pretty sure that the design is done in-house. some manufacturing may be outsourced.

let alone the actual chips.

I don't know if they STILL have any chip designers (I sort of doubt it) but when AIM first got started the Somerset chip design facility was a joint venture between all three partners including Apple. I believe some of the chip designers at the facility were technically on the books as Apple employees. At the very least the chip designers at Somerset worked closely with Apple.

If Apple had any ability to develop their own CPUs they wouldn't still be stuck with the pre-historic G4, they would simply ditch IBM and use their own chips.

Despite the fact that they DO have hardware engineers, and may even have a few that specialise in chip design to evaluate & work with the other two AIM partners it is obvious that they are not themselves, and are unlikely to become, a chip designers. Though because of the way patent and license agreements between the AIM partners they probably could get into it. But that would be a nightmare, they would bear all the costs and still be stuck with a single supplier (themselves) that would likely fall behind the competition.

Re:This could be good.

[jweatherley]

i'd be happy to use a ppc box as my desktop while not paying apple prices.

After you've checked out IBM's prices for PPC boxes you might not mind Apple's pricing so much...

Re:Actually the new Dual Systems have a fan it'sOT

[Oculus Habent]

Fans aren't that bad until you realize you can't hear as well as you used to. I work in the server closet alot (moving offices right now) and I sometimes grab a pair of earplugs.

The whine isn't bad until you realize you used to watch TV on 12, and now it's got to be 15.

In fact, our whole world (mine, anyway) is like this - far more noise than we were intended to hear regularly, and it slowly causes us to lose frequencies and ranges...

Do you find yourself trying to figure out what people said?

Incidentally...

[artemis67]

I don't have a DVD player at home, but I just got the new Monster's Inc. DVD (yes, I know I need to buy a player, but I'm cheap...). I happened to bring a brand-new ThinkPad home from the office to do some work. No RCA out, just S-Video. Cool, I can work with that.

So I pull out my S-Video cable, my computer speakers, and subwoofer, and get it all hooked up. Pop in the DVD and play it. Hmm... the TV is mirroring the laptop screen, but the video doesn't show up. After playing around with it for half an hour (and trying two different software players), I finally notice this little warning that says that "Copy protected DVD's will not output to the S-Video port" (or something like that).

WTF? Why even have a DVD drive and an S-Video port if I can't combine them? Note to everyone: Don't buy a ThinkPad if you think that there's EVER a chance you'll want to play a DVD through the S-Video port. If IBM is so damned concerned about DRM, they need to put a big sticker on the laptop that this is a DRM-enabled system. I guarantee that I will never buy another ThinkPad.

Anyway, next night, I bring home the Apple PowerBook. Hook everything up, pop in the DVD, hit play. No problemo.

Re:Computer noise, it's not so bad

[nettdata]

Apparently the white noise is supposed to be similar to the sound of the womb.
I'm not at all sure what this says about you. Perhaps you want to go back... :-).

Been trying non-stop since I was 14. ;)