IBM PPC 970MP A Reality?

[Rod Rod]

Of course, Apple could throw us a bit of a curve by not introducing a twin 970MP configuration (the "quad" system that has been discussed). In that case they might try to position the single-processor 970MP (dual core) as a substitute for today's dual G5 and keep a single 970FX at the low end. They might even be forced to do that if the initial yields on the 970MP are very low which would force the per-chip cost much higher. I hope, however, that this won't happen.

Yup. Apple has done that in the past, specifically when the Digital Audio Power Mac G4 came out. They cited short supplies of the processor. One good thing was that the pricing structure remained the same across the line, top to bottom. So, if a single dual-core 1.8 or 2.0GHz G5 becomes the low end, it would (according to this pattern) cost $1499.

 

[sw1tcher]

Hiroshige:

There's no mystery about the Freescale's enhanced G4:

http://www.freescale.com/webapp/sps/site/overview.jsp?nodeId=018rH3bTdG7249

I guess the only question is if Apple intends to use it. I'd be pretty surprised if they ignored the MPC8641 and went with some newer 64-bit Moto processor (which would presumably be some time away).

Wow. That MPC8641 looks nice. Having a MPC8641D in the next PowerBook revision would be awesome.

But I doubt it'll ever happen anytime soon. (Next Tuesday? )

 

[Cochrane]

Well, there was this laboratory that burned down from a G4 if I remember it correctly...

No, the cause was the power inlet the PowerMac was plugged into. Unless you are referring to a different story than I am.

And now for something completely different: According to the PDF-File, each core of the 970MP will have 1MB L2-Cache (specifically, it says the 970MP will "double L2 caches (1MB each)"), which is twice the amount of a normal G5 (always counted per core, the complete 970MP would then have four times of what you know today).

With all the stuff about first the 970GX on the IBM page about AltiVec, then the CHUD Tools and now this, it looks like Apple and IBM are working hard to give us computers with these processors quite soon. I hope they will manage to.

 

[isgoed]

jesus man! why are so many people flipping out about this? we all know it's coming, and it's an IBM doc. If Steve slippe dup and mentioned it on CNBC, that's a quirly little rumor, but come one, we all know this is written in stone. The question is when? And it sounds like this thing's too hot to not be refrigerated, let alone put in anything smaller than the powermac (or powertank, the new MP model).

We are flipping out because this is a rumour site. We are supposed to flip out over it! You even took the effort to reply a 95th post, Now how fanatic is that!?

 

[fpnc]

Wow. That MPC8641 looks nice. Having a MPC8641D in the next PowerBook revision would be awesome.

But I doubt it'll ever happen anytime soon. (Next Tuesday? )

The dual-core e600 chips from Freescale (like the MPC8641) are rumored to start sampling in the second half of 2005. So, I think it is likely that they couldn't appear in a shipping product until early 2006 (at best). However, the single-core, very-low-power MPC7448 (also an e600 core) was supposed to sample in the first half of 2005 (I've even seen some rumors that it sampled late last year).

I've been saying for months that the MPC7448 is the likely next candidate for the Mac mini and iBook and possibly even the PowerBook (in the latter case, first in the PowerBook, later in the mini and iBook). I suspect that the MPC7448 would perform about as well as a similarly clocked G5 under a pretty good range of applications, but it would use much less power than the current G5s and the MPC7448 is supposed to scale to at least 2GHz (with a faster FSB -- still only 200MHz compared to the current 167MHz -- but with 1MB of L2 cache -- twice the size of the current G4). The MPC7448 may not be used in the next PowerBook, but I'm pretty certain that it will be used in the Mac mini and iBook. Beyond that, whether Apple decides to use a Freescale dual-core chip for the PowerBook or something from IBM is hard to know.

 

[Cochrane]

By the way: There is also what looks like a reference to the 7448 in the CHUD Tools.
You can find it in "/System/Library/PrivateFrameworks/CHUD.framework/Versions/A/Frameworks/CHUDCore.framework/Versions/A/Resources/com.apple.CHUD.shortcuts.plist" if you have the CHUD Tools installed.

 

[Platform]

This means dual-core processors. You can have a dual processor machine and only take the space of a single one.

Not PowerBook G5's, because they'll run even hotter than normal G5's.

But iMac Dual G5's, or eMac Dual G5's, or PowerMac Quad G5's...

I sure hope that the next iMac's could have a dual core CPU

And could you make it soon as well apple

 

[visor]

How nice - new Powermacs April 1st

Obviously - introducing Powermacs and Tiger at once - nice call.

 

[visor]

hmhm mp core

so the next low end Powermac will feature a single MP processor - as such be as good as the current high ends. nice. it would also lack the need for 2 processor coolers and... basically everything else wich has to be pretty much dedundant now... makes a cheaper system, wouldnt you think? might even be quieter. one fan instead of two - drops another 3 DBa for proc cooling

 

[miloblithe]

Those freescale processors do look very nice for the next few rounds of revisions for the powerbooks, iBooks, Minis, and [eMacs]. I would hope that the powerbook transition to the 8641 would skip the 7448, but even that would be a nice transitional boost in the bigger L2 cache and 200Mhz system bus. The 8641 looks like it would be fantastic for a portable. G5 be damned! On the other hand, this thing isn't 64bit, is it?

 

[cr2sh]

Wow... quad powermacs and two button scroll mice... all of our apple dreams are coming true this morning.

 

[BillHarrison]

This is great stuff - and its curious that they haven't pulled it yet. I wonder if it will stay online now or not.

The only thing that concerns me is that they're talking about the fact that there is a lot of heat generated by these things.....I wonder if there are any issues with them because of that.

D

There has not been a processor aside from the PowerPC G3/4 family in the last 5 years that would not cook itself within 30 seconds of booting without a properly attatched heat sink. So, yes, there is alot of heat generated by these. But, almost ALL processors have on die diodes for this reason, including AMD and Intel Processors. This is simply an informational release, not to be taken as any thing more than that. Go ahead, boot a current G5 without a heat sink.... Trust me, you will be very sorry within about 1 minute

That aside, having 2 cores is going to equal more heat to dissipate, however it will be spread over a larger physical die, so this will help alleviate some of that.

In other words, this means nothing, other than them touting an on chip temp diode, which is far from unusual.

 

[iGary]

So, for us processor doofus'

Correct me if my recap is wrong.

1. A developer tool (not inside Tiger, right?) showed up that has dual dual processor support.

2. This support document showed up confirming the existence of a multi processor G5 chip.

In real terms, when can we expect this thing to hit market (be realistic).

 

[Evangelion]

At the same clock speed, I think it is unlikely that a single 970MP would be faster than a dual processor 970FX (as in the current dual G5 Power Macs). The problem is that the dual cores on the single 970MP will have to share the same Front Side Bus (FSB or system bus), while the dual processors in the current Power Mac G5s have independent FSB connections to the system controller and i/o.

the smart thing for IBM/Apple to do, would be to add an on-die mem-controller like Opteron/Athlon64 does. Memory-access is THE thing that stresses the FSB. If you have on-die mem-controller, the memory has a dedicated bus to the CPU(s), freeing the FSB to other tasks (HD's, PCI-Express etc.). Hell, Apple could freeze the FSB at 1GHz and it would still be fast enough!

And apart from having more effective bandwidth, on-die mem-controller would seriously cut down the latency. Right now when the G5 wants to talk to the RAM, it goes like this:

G5 talks to Northbridge, Northbridge talks to RAM, RAM talks to Northbridge, Northbridge talks to G5

Instead, you could have this: G5 talks to RAM, RAM talks to G5.

Apart from mem-controller, they could add Symmetric Multithreading from IBM's POWER-series. Two threads per core, two cores per CPU, two CPU's per system. 8 threads in total. You know it makes sense.

yes, multithreading is not as effective as real multiprocessing is. But it would increase the amount of transistor by a modest amount, while offering tangible performance-benefits. If Multiprocessing gives at most about 70-80% benefit over single-CPU-system, I would guesstimate that multithreading would give about 20-40% benefit. Nothing to sneeze at.

 

[kristo]

So, for us processor doofus'

Correct me if my recap is wrong.

1. A developer tool (not inside Tiger, right?) showed up that has dual dual processor support.

2. This support document showed up confirming the existence of a multi processor G5 chip.

In real terms, when can we expect this thing to hit market (be realistic).

This is a rumors site, no time to be "realistic"

This doc exists, so obviously the chip exists, which means apple has it in a lab somewhere working out how to fangdangle the thing into a powermac. The other thing is that the doco got put up on 03/13/05, meaning that IBM has had this chip for a long time (these kind of docs always get written a while after the fact) which as I said, means apple has had this chip in their hands for a pretty long time.

I think its not unreasonable to assume apple will use it soon.

I mean, think about it, IBM had to design it, most likely talking to apple techies in the process ("so, this chip, dual core, works for you?" apple "sure, hit us with em") then they fab some, give some to apple, apple gives feedback, everyone tests them, apple probably also sends some chips to some developers as well, chucks the info in chud so they can debug so that when the new systems hit, there is some software to utilize them.

Just a thought

 

[Fredstar]

How long was the 970, or standard G5 chip, confirmed by ibm themselves before it was actually seen in the Powermacs?
This could give us some kinda of time scale till they are properly released in the pmacs, i can't see them actually being released anytime soon...but if they were released at WWDC that would be kewl.

 

[kristo]

Apart from mem-controller, they could add Symmetric Multithreading from IBM's POWER-series. Two threads per core, two cores per CPU, two CPU's per system. 8 threads in total. You know it makes sense.

yes, multithreading is not as effective as real multiprocessing is. But it would increase the amount of transistor by a modest amount, while offering tangible performance-benefits. If Multiprocessing gives at most about 70-80% benefit over single-CPU-system, I would guesstimate that multithreading would give about 20-40% benefit. Nothing to sneeze at.

What I want is NUMA.

 

[Evangelion]

What I want is NUMA.

you would get that with integrated mem-controllers

 

[gkhaldi]

Dual core technology ready

I was just looking at this, can somebody say IBM just let the cat out of the bag big time.

It's a little harder to hide this than the 970GX scandal.

Michiro was absent for a long time (months and months) and drops this on ARS Technica.

Edit: In case IBM sees the need to close the page.

I know for a fact that the 970DP is up and running. IBM will more then likely start using this processor in a future lineup of "OpenPower", not based on the Power 5 architecture but on the 970. They need this to compete in the open space.

I understood from a "fellow" that they have crancked up the speed to 3.5 Ghz and got the thermal issues under control (read smaller packaging). They shipped a number of processors to outside parties for "integration" in November.

I expect that Apple to announce in the fall the roll-out of a 2 x Dual core G5 3.XX Ghz with support up to 64 GB of memory.

 

[relimw]

I sure hope that the next iMac's could have a dual core CPU

I hate to lump you in with the 'Powerbook G5s next Tuesday!' crowd, but I doubt you'll be seeing the dual cores G5 in the non-pro line for at least the next two revisions. I'd expect the powerbook to get the motorola dual core G4 in the next two revisions with speed bumps, and maybe graphic bumps.

The dual core, dual processor G5 does have uses beyond what most of you use the mac for. Remember, Apple has been trying to make major pushes into the scientific/research arena. Unix-level programs can fairly easily make use of as many processors as you can throw at it. Also, Xgrid, and distcc (distributed compiler) should see nice gains from the dual-dualies.

 

[relimw]

How nice - new Powermacs April 1st
Obviously - introducing Powermacs and Tiger at once - nice call.

Must be an April fools joke

 

[Rincewind42]

the smart thing for IBM/Apple to do, would be to add an on-die mem-controller like Opteron/Athlon64 does. Memory-access is THE thing that stresses the FSB. If you have on-die mem-controller, the memory has a dedicated bus to the CPU(s), freeing the FSB to other tasks (HD's, PCI-Express etc.). Hell, Apple could freeze the FSB at 1GHz and it would still be fast enough!

To do what you want would either require all cores on a single chip or a transition to a NUMA architecture. If you put all the cores on a single chip, your fine for p=2, but it gets hairier as you scale up. NUMA however I imagine would require some OS support as neither CPU sees all of RAM (so you would need to scale up the FSB so that the separate memory controllers can talk to each other quickly). I suspect that IBM is exploring memory controller on die, but might not have seen a win yet for the small systems that it is building with these chips. There is a huge win for this on larger systems, but these are larger than what we are discussing on this thread (i.e. systems larger than 2-way at the bus level).

 

[cr2sh]

I know for a fact that the 970DP is up and running. IBM will more then likely start using this processor in a future lineup of "OpenPower", not based on the Power 5 architecture but on the 970. They need this to compete in the open space.

I understood from a "fellow" that they have crancked up the speed to 3.5 Ghz and got the thermal issues under control (read smaller packaging). They shipped a number of processors to outside parties for "integration" in November.

I expect that Apple to announce in the fall the roll-out of a 2 x Dual core G5 3.XX Ghz with support up to 64 GB of memory.

Wow... for a first post those are some mighty tall words. I'm not so sure of what you're saying but I'm pretty sure your timeline is off... we're well past an update now, shipping 4months off. I think we'll have an announcement soon.

Big news really brings 'em out of the woodwork, eh?

 

[Fredstar]

Omg imagine dual dual-core processors running at over 3ghz and maybe even addressing like 16gb memory, similar to the dual-core Opteron's (i think)...not that anyone here other than high end science users etc would ever need more than 8gb.
Apple will again have the fastest computer in the world, and not by just a bit, by a considerable margin. If they can announce them at WWDC to ship around Q3 then that would be amazing.

 

[Sun Baked]

Wow... for a first post those are some mighty tall words. I like what your saying but I think your timeline is off... we're well past an update now, shipping 4months off. We'll have an announcement soon.

Big news really brings 'em out of the woodwork, eh?

As people have said, news of the chip is 8 months old already -- with an original ship date of Jan 2005 declared last year.

If we had to wait for fall for these things to arrive, there would be a revolt.

Of course you would expect IBM went back and massaged the chip with a new generation (970GX/MP) to solve some of the voltage & thermal issues along with tweaks to improve the yield -- or we're still in trouble.