A Mobile 64-bit PowerPC?

[MacRumors]

According to an anonymous report...

In light of the ongoing difficulties in incorporating desktop-class Processors (such as the PowerPC 970) into the PowerBook, Apple and IBM have laid the foundation for an architectural revamp of the PowerPC to produce an ultra low power 64-bit mobile processor.

According to this report, the upcoming mobile PowerPC will be part of a 300 series of processors from IBM. This new mobile processor is not due to debut until 2005.

 

[Dippo]

I would take this with a big grain of salt.

I am sure by 2005, Apple would be able to put the 90nm G5's into the powerbooks.

 

[Ambrose Chapel]

plus steve has already said they want the PB G5 by next year. maybe that would be a stop gap until this new tech is available?

 

[arn]

I reworded the post.

This isn't meant to imply that they are going to wait until 2005 before creating a new PowerBook with a 64-bit processor... but they are going to have the same problem with all upcoming desktop-class processors, and so are planning ahead. (according to this report)

This would presumably parallel the mobile Intel processors.

arn

 

[DillHarris]

This would be cool... Get the G5 into the Powerbook for now, but have true mobile 64-bit processor new from the ground up... Good performance, wicked battery life...

 

[mrsebastian]

drool = )..... i want my 17"pb g5!

 

[SiliconAddict]

Originally posted by mrsebastian
drool = )..... i want my 17"pb g5!

Ditto *sighs*

 

[johnnyjibbs]

Could this possibly be in response to struggling to put the G5 in a PowerBook? Maybe, they had to rethink this one.

Or maybe they would put G4-comparable G5s into the PB next year (i.e. low clocked, low power, 90nm), and then put this truly remarkable mobile chip in the 2005 models.

I don't think Apple has actually committed to an absolute timeframe for a PowerBook G5, other than that they would really like to have one before the end of next year...

 

[dho]

for now, it looks like the pbg5 will have an amazing 15 minutes of battery life

 

[hvfsl]

Originally posted by dho


for now, it looks like the pbg5 will have an amazing 15 minutes of battery life

There are some P4 laptops that have as much as 30mins battery life. I expect Apple will easily be able to get 3hours + out of the G5 since they use small batteries compared to PC laptops.

 

[phrancpharmD]

extended battery life

Originally posted by hvfsl
There are some P4 laptops that have as much as 30mins battery life. I expect Apple will easily be able to get 3hours + out of the G5 since they use small batteries compared to PC laptops.

Maybe this will help propel the fledgeling fuel cell technology I've read about the past few months? If I recall correctly, fuel cells in a more "traditional" battery form factor should last weeks in current laptops before needing a "refill." Maybe this technology would at least allow several hours of use with a PPC 970 or later processor. . . Sorry, I don't have a very good grasp on the tecnhical aspects of this, but if anybody else does that's cool. . .

 

[leet1]

Originally posted by hvfsl
There are some P4 laptops that have as much as 30mins battery life. I expect Apple will easily be able to get 3hours + out of the G5 since they use small batteries compared to PC laptops.

I gotta call bs on this. What laptop with a p4 has a 30 minute battery?

 

[ITR 81]

We had some P4 Panasonic laptops they wouldn't go beyond 2hr and 40 mins working with just normal apps. I don't know what would happen if you ran games on it. We used them in Mil.

 

[leet1]

Originally posted by ITR 81
We had some P4 Panasonic laptops they wouldn't go beyond 2hr and 40 mins working with just normal apps. I don't know what would happen if you ran games on it. We used them in Mil.

I couldn't see it loosing 2 hours just playing a game, thats way overexagerating lol

 

[Golem]

IBook G5


It would be the logical progression.

2004 we have Ibook G4 + a battery hungry Powerbook G5, 2005 would mean new processors acrss the whole book line.

 

[manitoubalck]

While not a ture mobile proc, the Athlloon 64 has been successfully incorperated into a laptop, All be it a heavy one by VoodooPC.

 

[Henriok]

This is logical. Power consuption is becomming a real problem with high end PowerPC processors, while the low end is doing a REALLY good job ceeping consumption down to a minimum. IBM is after all tageting the 440 core to PDAs so we are looking at chips pulling less than 1 W at frequencies around 400 MHz. Motorola is doing a great job with its G4-line. less than 10 W @ 1 GHz. It's really unacceptable for a 3 GHz processor to draw ~100 W so IBM should make every effort to reduce that.

Fabrication process is one thing but it can only take one so far, reducing power consumption with approximately one third. Can they improve fabrication so that they can use a low volt core, perhaps at 0.9 V instead of 1.3 V? I think complete redesign is the way to go and stip the processor from elements that might be unnecessary. Are two FPUs and a large L2 cache really that important? Is there a way to power down ALUs when they are unused? Can they run different areas of the processor at different voltages at the same time? If they improve compilers, can they skip one of the load/store unit? If they integrate the northbridge, they they might save power when looking at the chipset design in whole? AltiVec? Is it really that important?

And so forth.
Compared to other chip designs, and theeir evolution over the years, I really cant se any reason why IBM should be able to cut power consumption in the 970 design by at least halv by just being clever. 970 is a new design and we are looking at the first implementation. IBM have a continuing evaluation process going on and are constantly looking for areas that they can improve upon. Their East Fishkill facility is a relatively young plant too so there's probably a lot that can be made there too.

I wouldn't be surprised that Microsoft is concerned with power consuption of the future Xbox 2-chip. I guess that the abyssmal record of the x86 processors was one reason for them not to choose that path, but IBM's G5 line is not that impressive either. Not compared to the flavours of G3s.

In short: This is a natural move from IBM. It's a job that must be done.

 

[cb911]

aahhh... it's times like this... like my sig says, i always have my grain of salt with me.

it would be really good if they get it a low-power version going...

 

[manitoubalck]

Originally posted by Henriok
If they integrate the northbridge, they they might save power when looking at the chipset design in whole?

Might have copyright issuse with for using AMD's true 'Hyer transport' which allows processors to address each other without going though the north bridge. It also allows each proc to have it's own dedicated RAM, which it address' without the use of a north bridge.

 

[dgbatchelor]

A little off-topic, but...

"Faster" seems to always be better for a laptop... But it is the overall performance of the laptop which is important -- not just the raw processing power of a given processing core...

The natural progression has been from 8-bit, to 16-bit, to 32-bit, to 64-bit processing cores...

Has anyone given thoughts to a future PB architecture with several (read this as 8 or more) low-power 32-bit processing cores which would act like a small local SMP "cluster"?

Would we likely see better overall performance for the same battery life? Energy savings techniques could put one or more of the processing cores to sleep if they were not needed at the time... Would this also give lower overall heat generation?

Clearly, there are issues to solve regarding bus arbitration...

I know that I'm over simplifying, but could you imagine a time when the user could purchase extra CPU power on the equivalent form factor of a DIMM and install/replace 8 (or more) processing cores at a time?

Think of the approach to be an analogy of RAID...

...that would let the consumer decide how much processing power to install ("Did you want a 2-pack of processor modules... each with eight low power 2GHz G4 cores...?")

Another BTO option?

I know that this solution wouldn't likely result in the equivalent of a supercomputer cluster of 1100 DP G5 nodes... (But we can dream...)

Thoughts?

-- dgbatchelor --

"If you have built castles in the air, your work need not be lost; there is where they should be. Now put foundations under them." -- Henry David Thoreau

 

[wizard]

While you bring up some good points at the same time you seem to fall intp the abyss of regressiveness.

First where is it said that the 970 burns 100 watts of power???

I agree that the Motorola processor is good on a watt/performance basis, it is unfortunate though that Apple can't seem to get Motorola to produce faster revisions. All they really need to do is to increas the clock speed of the G4 75% and they would have a viable processor for another year. I'd like to see a show of hands as to the likely hood of Motorola accomplishing that.

As to your comments on cache and FPU units hell yeah we want them. On a laptop the larger the on die cache the better. As far as the other execution units, I want it both ways. That is better power disapation and all the features of the latest processor generation. Lets not short change ourselves at least in the case of the PowerBook.

As far as Alt-Vec goes, that is really a silly question isn't it. Every mainstream processor that has come on the market recently has had some facility to process vectors. I could see a reduced capability vector unit for chips destined for certain markets such as the IBook but not for mainstream laptops. Now if they find ways to improve it fine, but lets not castrate it for the sake of power reductions.

Thanks
Dave

Originally posted by Henriok
This is logical. Power consuption is becomming a real problem with high end PowerPC processors, while the low end is doing a REALLY good job ceeping consumption down to a minimum. IBM is after all tageting the 440 core to PDAs so we are looking at chips pulling less than 1 W at frequencies around 400 MHz. Motorola is doing a great job with its G4-line. less than 10 W @ 1 GHz. It's really unacceptable for a 3 GHz processor to draw ~100 W so IBM should make every effort to reduce that.

Fabrication process is one thing but it can only take one so far, reducing power consumption with approximately one third. Can they improve fabrication so that they can use a low volt core, perhaps at 0.9 V instead of 1.3 V? I think complete redesign is the way to go and stip the processor from elements that might be unnecessary. Are two FPUs and a large L2 cache really that important? Is there a way to power down ALUs when they are unused? Can they run different areas of the processor at different voltages at the same time? If they improve compilers, can they skip one of the load/store unit? If they integrate the northbridge, they they might save power when looking at the chipset design in whole? AltiVec? Is it really that important?

And so forth.
Compared to other chip designs, and theeir evolution over the years, I really cant se any reason why IBM should be able to cut power consumption in the 970 design by at least halv by just being clever. 970 is a new design and we are looking at the first implementation. IBM have a continuing evaluation process going on and are constantly looking for areas that they can improve upon. Their East Fishkill facility is a relatively young plant too so there's probably a lot that can be made there too.

I wouldn't be surprised that Microsoft is concerned with power consuption of the future Xbox 2-chip. I guess that the abyssmal record of the x86 processors was one reason for them not to choose that path, but IBM's G5 line is not that impressive either. Not compared to the flavours of G3s.

In short: This is a natural move from IBM. It's a job that must be done.

 

[jade]

well i hope pbg5 have a much faster frontside bus. The current crop of centrinos have 400mhz, and version 2 will have 533mhz.

 

[MacRAND]

Mobile multi-chip 32-bit PowerPC • Cluster Computer Age

dgbatchelor I agree

While I drool at the thought of a G5 dual 2.5 GHz box, and I look forward to multi-chip G5 xServers
I really question the rational behind putting a currently HOT G5 chip into a PowerBook, until things cool off a bit. Why bend yourself out of shape with extreme radiators, just wait till the size of the chip and current needed to run it match for a laptop.

In the meantime, I'd love to see a G4 dual 1GHz chip configeration in a 15" or 17" PowerBook. Smokin'! Or, would it also be too HOT!
Yes, G4 multi-chip PB, even before a slow G5 1.5 GHz PowerBook, and I'll bet that this is the G5 PB direction, and NOT 1.8 or 2.0 GHz. Problem is, two G4 chips may run hotter than one G5, can't do the math, but Apple is.

The move to G4 in the iBook was long overdue, I won't miss G3 (which I have), unless they pep it up with radical cache or something. G3 Multi-chips, why not? How about some of these new low-power IBM chips (not G5)?

 

[CoreForce]

Originally posted by mrsebastian
drool = )..... i want my 17"pb g5!

Me too.

Apple should set up a site for taking pre-orders for PB G5, 15'' and 17'' and advertise this to some selected places only.

Just to see what the demand really is.

 

[Henriok]

Originally posted by wizard
First where is it said that the 970 burns 100 watts of power???

I really can't vouch for the accuracy of AppleInsider's sources, they do state that a 130 nm 970 @ 2.5 GHz consumes 97 W. FAR too much.

Lets not short change ourselves at least in the case of the PowerBook.

No I certainly hope they can avoid that.

As far as Alt-Vec goes, that is really a silly question isn't it.

No I don't think so. Assume for a moment that we can get a "PowerPC 960" that is a 970 that run at 3 GHz but does not include AltiVec.. Wouldn't you prefer that one to a 1 GHz G4 in a iBook?

when looking at the die of 970 it seems to me that they could've used the space more cleverly. When looking at what Intel did for Centrino and what they will do for Tejas, rearranging and revising what stuff goes where in the processor can do wonders for performance and power consumption. IBM should look real hard on how they can optimize the use of real estate. A large chip will always draw more power than a smaller.

On the other hand.. IBM seems to be doing some of the things we brought up here. Stuff that will go into the 90 nm version of 970 will be extensive power saving feeatures. At least from the little blurb I can read from the ISSCC'04 agenda.