Future of the PPC? (970, 980, 990, 9900)

[MacRumors]

The following information is from an unconfirmed and anonymous source. As such, authenticity is always uncertain, but due to the content of the piece, was felt to be of sufficient interest for publishing. Of interest, MacBidouille has posted similar details in their unconfirmed rumors of the PPC's timeline. This may represent corroboration -- or simply a common source. Take, as with all rumors, an appropriate amount of skepticism.

Apple and IBM have been working on parallel development of the Power5 and PPC 980. The PPC 980 is a single core version of the Power 5. While prototype forms of this chip exist, it is almost a year away from shipping in Macs.

Improvements in the PPC 980 include Hyperthreading, eLiza error correction, and more massive parallelism. IBM's implementation of hyperthreading provides a 30% gain over Intel's. eLiza technology will reduce the bottlenecks when the branch prediction unit fails. Altivec will split into 3 pipelines (vs 2 in the 970), 4 Integer and 4 Floating point units. 980 will have to be built on a 90nm processor due to heat dissipation requirements.

Steve's comment of 3GHz in 1 year will not be accomoplisedh by the G5 - which will top out at 2.6-2.8GHz. The PPC 980 will start at speeds of 2.6-3GHz and top out around 4.5-5GHz. The G5 will make it's way into PowerBook lines in Jan/Feb, Xserve's later this year, and iMacs in approximately one year.

Marklar's project size has decreased, but remains ongoing. There are four generations of the PowerPC including and beyond the 970 that are in development and planning. Besides the 980 chips (targetted at end of 2004), there are plans for 990 chips on a 65nm process in 2005/2006 @ 6GHz and scaling up to near 10GHz. Beyond this, the PPC 9900 starting on a 45nm process is targetted in 2007/2008 starting at 9-10GHz and reaching up to 20-25GHz by 2010-2011.

[arn]

This could be true... it could be made up. I don't know. But is an interesting read... so enjoy.

I guess it's something to keep in mind as more details surface over the next year.

arn

[k2k koos]

it's good to see they are not resting on their b*tts, but keepp on developing products for the future, to stay current, and perhaps even widening the performance gap...

[evil]

20-25 ghz
hot damn

[QCassidy352]

good lord, makes me not even want a g5 anymore! I hope it's all true. That sounds great.

One question though- can someone explain "power 4" and "power 5" as compared to 970 and 980? The 970 is a power 4 derivative, while the 980 is a single core power 5? Is the G4 also a power 4 derivative? Is the 990 also a power 5 derivative? Thanks for the help.

[ZildjianKX]

Well, some interesting PB news. It would really suck for the towers to go to the PPC 980 chip and the PBs to still be G4s... I guess I'll have to hold out on the PBs for another 6 months... I think it will be worth it in the long run, maybe a new ipod by then anyways... not to mention preloaded with Panther.

[Ja Di ksw]

Not to be rude, but what does someone need 25 Ghz for? Honestly, that is just insane, especially if they are dual processors. I know people making music or animation or whatever would like them, but for your average Joe, do we really need that much? I guess that's what the iMac's and the like are for, though by then they will have 10 or 20 Ghz or whatever.

Remember this?

"640K ought to be enough for anybody"
- Satan, um, I mean, Bill Gates.

[Kermit]

I'm no fan of either Bill Gates nor Microsoft but that "640 kb is good enough for everyone"-quote is false. Bill Gates never made such a remark and it is time that we lay that myth to rest.

[macdop]

I am no expert on chips etc., but a 45nm sounds a little imposable, as I believe that would mean like 3-4 atom gates, which I believe is imposable even with UV fabrication technologies, and would also cause random electron jumps from gate to gate hence making transistors imposable, as I said, I am no expert, and I could very well be wrong, and I have just come back from a bar, and am very drunk, but I think simple physics makes this impossible

[GregA]

Originally posted by Ja Di ksw
Not to be rude, but what does someone need 25 Ghz for? <snip>

Remember this?
"640K ought to be enough for anybody"
- Satan, um, I mean, Bill Gates.

Ah, 10 years from now, a friend walks in and says "nice garden out your window" - you say "no, actually that's my 3d Mac interface, with garden themes - pull out that carrot and watch what happens".

Or maybe they'll use the 25Ghz for a complex AI which only weeds out Spam and ads.

[Veldek]

I'm quite sure that I read some days ago, that IBM has a 970 working at 3.2GHz dissipating 82W, which is as much as the Intel chips at the same frequency. So it doesn't seem impossible for them to reach the 3Ghz barrier with a 970.

Anyway, as IBM is working on the Power5 (the 980 will be a derivative of it, the G4 has nothing to do with either of them) and on the Power5+, I think we may expect a lot in the near future.

[C14ru5]

What's the point of having eLiza error correction in a consumer processor? I thought error correction only was necessary in server processors like Power4 and Power5. I mean, how often do calculation errors occur in a processor chip?

[BaghdadBob]

20-25 Ghz in four-five years?

Wow.

I wonder how true that is.

To anyone who wonders what you would do with all that power, just remember that the more power your computer has, the smarter it can act. There are many things computers all do today that never would have been practical on sub 100 Mhz processors.

Trust me, if the power is there, someone will find a way to use it. Like Longhorn and the wavy window effect!

Actually, I'm waiting for the day when my desktop can fluidly animate like in Lawnmower man.

[LostPacket]

Originally posted by Macrumors
Beyond this, the PPC 9900 starting on a 45nm process is targetted in 2007/2008 running up to 20-25GHz.

Correct me if I'm wrong, but at 25 GHz an electron will only have 4E-11 seconds to complete a cycle and could travel about 1.1cm (assuming 90% speed of light).

How large will these chips be? Will the electon even be able to get across the CPU in time to complete its cycle?

I don't know anything about chip design, but these frequencies seem a little far fetched to me.

[MyLeftNut]

20-25Ghz?

Wow...you know what will happen though, your new PowerMac becomes self aware as of 12.35:48pm on Jan 02, 2009 and then we all vaporize in a blinding flash of light....

Sorry couldnt help myself...

[h'biki]

Yeah, I bet Steve will love to announce the QMac, the world's first personal quantum computer.

The comes the aiMac, the world's first artificially intelligent, fully aware personal computer

And then Steve downloads his consciousness into the computer and calls himself Steveomancer.

[midifarm]

Wahoo! You're all clear kid. Now blow this thing and let's get out of here!

It'll be a wonderful day when even the ./ers will wonder why they ever thought Intel would do the job. Go Big Blue and the Core!

[sososowhat]

All the physics aside (and way beyond my expertise), so long as Moore's Law continues, there's no news here.

Doubling every 18 months (yes, I know it has to do with transistor counts, not GhZ, but historically it works for either):

3 years = 4x
5 years = 10x
10 years = 100x
15 years = 1000x
20 years = 10000x
30 years = 1 Million X

So, by 2033 your Mac should be running around 1 Million times the speed of a G5.

And - more to the point - in 5 years, it should be running at 20 GhZ, which is what this "breaking news" says.

1986, my first PC:

6 MhZ 286
1 Meg RAM
30 Meg disk
1200 baud modem

15 years later (2001) - by Moore's Law 1000x - curiously it applies to everything, not just the chip:

6 GhZ chip (ok, we're off by 3 years)
1 Gig RAM (yup, right on)
30 Gig disk (right on schedule)
1.2 Mbaud(?) is about 1.5x my DSL

[LostPacket]

Originally posted by MyLeftNut
Wow...you know what will happen though, your new PowerMac becomes self aware as of 12.35:48pm on Jan 02, 2009 and then we all vaporize in a blinding flash of light....

I'd like to give the future race of machines the benefit of the doubt.

After all, any self respecting intelligence would spend its time searching for porn. Maybe a PowerMac G10 would search for a screencap of a dual-G5 with the side panel open ("Oh baby, look at those heat sinks").

I need sleep.

[mac15]

damn thats cool, some rather nice news and wow .65nm is so damn cool

[Shadow_Raptor]

First up QCassidy352 the IBM 970 is based on (or derived) from the IBM Power 4 server processor. The 980 is based on the Power 5 server processor. In both cases the Power 4/5 processors had dual cores while the 970/980 have only one core. They are a slightly less powerful processor with some changes made to suit their needs as a desktop prcoessor in Apple Macs.

The G4 was of course made by Motorola and so has nothing to do with the IBM Power 4 server processor.

And C14ru5 eLiza error correction is for when the branch prediction fails. The processor tries to guess what it will be asked to do - and begins to do that before it gets told to. If it gets it right it continues - if it gets it wrong it dumps the whole branch and starts from strach - very efficient if its right (which it is 80-90% of the time) but costly if it isn't.

As for everyone worrying about the 20-25Ghz 45nm process rest assured that this problem has been expected for some time now. It is true that the theoritical limit for current processors is up around 30Ghz. After that everything gets too small, too fast, too power hungry and too hot. Of course was to over come this will eventuate - think quad Power Macs - but I don't think we will see desktop computers going Quantum any time soon. I say this because Quantum computers are only fast at what they do - that is to say not very much.

[iGav]

Either way... I think IBM is very good for Apple!!

[Analog Kid]

Originally posted by QCassidy352
One question though- can someone explain "power 4" and "power 5" as compared to 970 and 980? The 970 is a power 4 derivative, while the 980 is a single core power 5? Is the G4 also a power 4 derivative? Is the 990 also a power 5 derivative? Thanks for the help.

IBM, Mot and Apple co-developed the PowerPC architecture so all of the chips share the same basic instruction set. This is why applications don't need to be recompiled when Apple changes manufacturers and why the Apple-Intel rumors were bunk from the start ("uh, now that you guys just finished porting your applications to OS X, we need you to port to an entirely different instruction set-- mind your byte ordering!!").

IBM and Mot cross-licensed the Alti-vec instruction set. Alti-vec is what made the G4 the G4 and has since been carried into the G5.

Power4/5 are IBM brand names, so the G4 (a Mot chip) isn't derived from either. Again, they all use the same instruction set (the G4, G3 etc use a 32bit extension to the original 64 bit PowerPC specification).

The 970 is a "stripped down" Power4. The Power4 is dual core, the 970 is single core. I think the Power4 also had a larger cache and an integrated memory controller which the 970 doesn't. The only addition I'm aware of is that the 970 added Alti-vec which the Power4 had no use for.

The Power4 is also a "beefier" more robust technology. Fatter oxides on the transistors and such for the never-fail requirements of a server. This actually slows the Power4 down a bit so slimming down the process allows the 970 to run at a faster clock rate than the Power4 could.

Ars Technica had a really detailed set of articles on the 970 if you're interested in that level of geekiness.

The 980 will make similar modifications to the Power5, except this time they are being designed in parallel.

All indications are that Apple and IBM will continue following this methodology and Apple will continue to use the IBM chips until Mot pulls their collective thumb out of their collective.. uh... ear... and realizes that cell phones aren't going to keep the company afloat.

This is the first mention I've seen of the other two chips, but I'd guess the 990 is a Power6 derivative.

If you're not interested in the fundamental architecture, I think the important thing to carry away from this is that IBM has an interest in developing world class processors for their own servers and workstations (which net much higher profit margins than just the chip would) so as long as the relationship between Apple and IBM remains warm and Apple doesn't start eating into IBM server sales we can expect leading edge processors in our Macs.

The deal is good for IBM because the added volume going to Apple keeps the IBM fab lines running more efficiently and helps defray the R&D costs. It's really a win-win.

Mot never had this incentive. Apple isn't going to build their Mac's around a cell phone processor (and Mot doesn't even use Mot processors for their phones, which should tell you something) and the biggest application for Mot PowerPCs is routers and networking equipment which is a very different problem with very different design goals. Beyond that it's mostly "embedded" applications like automotive.

IBM has an insane R&D budget and they span everything from practical manufacturing and OS development to basic physics research that won't show up in a product until, well, until the 9900

[jbrown]

Originally posted by Ja Di ksw
Not to be rude, but what does someone need 25 Ghz for? Honestly, that is just insane, especially if they are dual processors. I know people making music or animation or whatever would like them, but for your average Joe, do we really need that much? I guess that's what the iMac's and the like are for, though by then they will have 10 or 20 Ghz or whatever.
.

I remember pushing the envelope at 8 mb RAM

[skunk]

Does 25GHz mean I'll be able to type faster?