[Eug]

Yowza!

Some comparative numbers (typical power usage):

1.8 GHz G5 0.13 um - 51 Watts
1.2 GHz G5 0.13 um - 19 Watts (likely higher actually)

2.0 GHz G5 0.09 um - 24.5 Watts
1.4 GHz G5 0.09 um - 12.3 Watts

G5 PowerBook 1.6 here we come!

[k2director]

Do you know what the comparable numbers are for the current G4 in the Pbook?

[issa]

Thanks for posting those specs and the link to the source, Eug. Looks like things might get interesting before long. As well as my 1GHz TiBook continues to serve my needs, it would be fun to see a new chip and hopefully appealing new enclosure to tempt me to reach deep down into my depression-state pockets.

[Minty Fresh]

*keeps fingers crossed* Hoping for the G5 PB's soon...

[csoledad]

I want a Powerbook G5 as much as the next guy, maybe more.

However I recentely learned the memory/bus controller developed by Apple (still on the old process) consumes 30 watts typical.

That is one of the reasons the 1u G5 xserve still makes so much fan noise and sucks so much air.

I hope the 30 watt figure is wrong, but if it is true the problems are more than G5 power and heat.

[gator]

Not sure if this is the right reference page for G4 processor information:

http://e-www.motorola.com/webapp/sps...018rH3bTdG8653

[issa]

csoledad,
Thanks for the insight. Your concerns are understandable. Guess we're at a disadvantage not having seen Apple's specs on the memory bus/controller itself.

The Xserve is designed to keep dual 2GHz CPUs with two separate front-end memory buses, three HDDs, three PCI-X card slots and more running full bore 24/7. I take it these are some of the reasons you refer to regarding fan noise and air sucking. No surprise it needs lots of cooling. Now, even if the memory bus/controller in the first generation of G5 desktops and the Xserve is as power hungry as you've heard, the challenge to Apple then becomes to wrap the performance we'd expect from a G5 PowerBook that runs a single CPU and single 2.5" HDD, (including front-end bus speed, proper RAM support, etc.), into a scaled-down controller that can keep total power consumption at a level similar to what's in the current G4 PB lineup. Not to suggest that it's easy or that we'll see it tomorrow. Still, PowerBooks accounted for something like 48% of unit sales last quarter; this is great motivation. Here's hoping the designers have their aluminum beanies on and are finishing up schematics for the controller and other subsystems needed to encourage us to break into our piggy banks.

[Eug]

Originally posted by k2director:
Do you know what the comparable numbers are for the current G4 in the Pbook?

Here is what they are, but the numbers don't seem to be comparable.

MPC7455 (TiBook, G4 Xserve)
1 GHz - 15 W typical, 22 W max
1.33 GHz - 30 W typical, 30 W max

MPC7447 (most recent version of AluBook)
1 GHz - 15.8 W typical, 22 W max
1.267 GHz - 18.3 W typical, 25.6 W max

The reason I say you can't compare the numbers is because IBM lists the G5 2.0 970FX at 24.5 W typical, but Apple lists the G5 Xserve CPU at 55 W max.

Thus IBM's "typical" is about 45% of Apple's "max", whereas Motorola's "typical" is about 75% of their "max".

Nonetheless, one might guess the power utilization of the G5 1.4 is probably somewhere around that of the 1 GHz G4, and I suspect a G5 1.6 would be in the range of the 1.25 GHz G4 in the PowerBook.

csoledad,
That 30 W figure sounds about right for the system controller, but that is on the 0.13 um process, and one wonders how quickly it will go 0.09 um. My guess/hope is rather quickly. Indeed, it's likely that the G5 Xserve controllers have been in production for quite some time, and the timing would have necessitated 0.13 um. But it also stands to reason that the controller at 0.09 will simply be a drop-in replacement, and they can go 0.09 anytime without haven't to maintain a separate 0.13 um process for the system controller.

[dennis88]

Here is what to come:

Pb 12: 1,25ghz G5.
Pb 15: 1,25ghz G5.
Pb 15: 1,5ghz G5.
Pb 17: 1,6ghz G5.

And the 15" and 17" will likely have ati mobility 9600 pro 128mb (not the current non pro) or the mobility 9700 that will come.
The 12" will have the ati mobility 9600 with 64mb ram (the same as in the current 15" and 17")

They will be launched in early april, and they will start shipping on the day.

[KraziKid]

Originally posted by dennis88:
Here is what to come:

Pb 12: 1,25ghz G5.
Pb 15: 1,25ghz G5.
Pb 15: 1,5ghz G5.
Pb 17: 1,6ghz G5.

And the 15" and 17" will likely have ati mobility 9600 pro 128mb (not the current non pro) or the mobility 9700 that will come.
The 12" will have the ati mobility 9600 with 64mb ram (the same as in the current 15" and 17")

They will be launched in early april, and they will start shipping on the day.

That's a very hopefull prediction. If you've read the rest of the thread, you would have seen that Apple's bus controller is still at 30W. This is not suitable for a portable. I still hold my prediction of earliest at the end of 2004, and most likely 2005.

[Pierre B.]

Originally posted by dennis88:
... or the mobility 9700 that will come.

Is there already a mobility version of this one for PCs?

[dennis88]

Originally posted by Pierre B.:
Is there already a mobility version of this one for PCs?

No, but it will come.
It's the next ati mobility chip.

[Pierre B.]

Originally posted by dennis88:
No, but it will come.
It's the next ati mobility chip.

Some hint on that? Is it in the rumor realm, logical evolution, official announcement?

[ae86_16v]

Originally posted by KraziKid:
That's a very hopefull prediction. If you've read the rest of the thread, you would have seen that Apple's bus controller is still at 30W. This is not suitable for a portable. I still hold my prediction of earliest at the end of 2004, and most likely 2005.

Yeah, I think that is fairly right. I remember that IBM said a while ago that they'll most likely make a Portable version of the chip.

[dennis88]

Originally posted by Pierre B.:
Some hint on that? Is it in the rumor realm, logical evolution, official announcement?

Just wait and see.
It will come.

[Eug]

Originally posted by KraziKid:
That's a very hopefull prediction. If you've read the rest of the thread, you would have seen that Apple's bus controller is still at 30W. This is not suitable for a portable. I still hold my prediction of earliest at the end of 2004, and most likely 2005.

A die shrink of the system controller doesn't have to wait that long. For all we know, they could be doing that now.

A die shrink would go a long way to reduce power utilization.

[Spheric Harlot]

Originally posted by Eug:
Thus IBM's "typical" is about 45% of Apple's "max", whereas Motorola's "typical" is about 75% of their "max".

Makes sense, though, if average usage hasn't increased. If you're doing the same stuff on average on a much more powerful processor, your typical usage percentage is obviously going to be much lower.

-s*

[Pismo2001]

Originally posted by Eug:
A die shrink of the system controller doesn't have to wait that long. For all we know, they could be doing that now.

A die shrink would go a long way to reduce power utilization.

Not that much, actually. Because a great deal of the power dissipation in such a controller is due to the drivers for all the high frequency busses.

On a dual G5, there are 2 32 bits busses to the processors (and 2 coming from the processors), 2 address (16 bits each ) and data busses (64 bits each) to the RAM modules, and a few other things like AGP/HT/whatever. Each bus also has a few control signals (quite a few clocks, byte enables, and other although the low frequency
ones don't really count).

Now the power dissipated on the busses depend on the characteristic impedance of the lines, on their capacitive loading (especially the RAM since their may be up to 4 modules on each bus) and other physical parameters, and a significant fraction of that power comes through the supply pins of the
bridge chip.

However a portable version of the bridge might have reduced power supply requirements due to:
a) only one processor interface,
b) why not a single DDR bank ?,
c) maximum 2 DDR modules and,
d) very short distance between the bridge and the processor on one side and the bridge and memory on the other.

The short distances (obviously not the case in the tower since the length of the processor busses is at least about half the distance between the processors) allow you to consider the busses as capacitive loads, not transmission lines. At 1 GHz, the power dissipation starts to diminish at about 5 centimeters (2 inches) but you still need some form of impedance matching to limit reflections for lines longer than about half this value.

I don't mean that the smaller geometry process won't be able to reduce the power dissipation, it would be wrong, but on a chip which mostly shuffles data around between a lot of high speed busses, power dissipation is dominated by the bus drivers these days, not by the internal logic.

Actually, a few years ago some people claimed that these chips have so many connections and so few logic transistors that the die is pad limited, i.e., that a process shrink won't drastically reduce the die size. The idea at the time was to use that free area as cache memory (would be L3 cache on a G5). This would be a very neat idea if the cache can be made large (8 or 16MB) and it might even further reduce power dissipation by lowering the number of accesses to main memory.

Ok, I'm dreaming ;-)

[SEkker]

Steve Job's history at product release is largely market driven. If their PB factories are at full capacity, which is what the recent Apple business numbers indicate, then releasing a G5 PB now will not increase the company profits in the short run (and would cost money in the long run). And at the moment, the AlPBs -- even with a G4 -- look competitively priced, with only a minor hit in performance, over a Centrino notebook computer.

I expect the PB G5s to be announced ~1 month before the next iteration of the Centrino, so Jobs can lay stake to the 'fastest notebook computer on the planet' title. Anyone know the latest from Intel (all of the dates I have are from forcasts 6 months ago, an eternity in the semiconductor field).

A dual G4 upgrade, if the current motherboard design allows, would bridge the time nicely from a business perspective.

Please don't get me wrong -- a G5 PB for release on leap day 2004 would be terrific news for those purchasing PBs in 2004, but I think it would not make too much business sense at this time.

Besides which, Steve Jobs will clearly want to show the PB G5 himself, and he seems to be completely occupied with making the Apple-HP alliance work. First iPods, next -- Ives' branded HP notebook computers? Who knows. If it increases Apple's profits, all Apple customers win.

[binghd]

Originally posted by dennis88:
Here is what to come:

Pb 12: 1,25ghz G5.
Pb 15: 1,25ghz G5.
Pb 15: 1,5ghz G5.
Pb 17: 1,6ghz G5.

And the 15" and 17" will likely have ati mobility 9600 pro 128mb (not the current non pro) or the mobility 9700 that will come.
The 12" will have the ati mobility 9600 with 64mb ram (the same as in the current 15" and 17")

They will be launched in early april, and they will start shipping on the day.

The next version of the ATi Mobility probably won't be 9700. If you look at their history, it'll be named after ther lower-end of whatever their next desktop chip release is. But their mobile chips are designed specifically for laptops, not just a low-power version of their desktop chips.

[ae86_16v]

Originally posted by SEkker:
Steve Job's history at product release is largely market driven. If their PB factories are at full capacity, which is what the recent Apple business numbers indicate, then releasing a G5 PB now will not increase the company profits in the short run (and would cost money in the long run). And at the moment, the AlPBs -- even with a G4 -- look competitively priced, with only a minor hit in performance, over a Centrino notebook computer.

I expect the PB G5s to be announced ~1 month before the next iteration of the Centrino, so Jobs can lay stake to the 'fastest notebook computer on the planet' title. Anyone know the latest from Intel (all of the dates I have are from forcasts 6 months ago, an eternity in the semiconductor field).

If it increases Apple's profits, all Apple customers win.

That makes sense. Intel just updated the Centrino's WiFi component to be Dual Band (802.11b/g).

http://news.com.com/2100-7351-5141629.html

The next Centrino chip code name Dothan (2nd Generation) was suppose to come out February of this year, but has now been delayed till Second Quarter. Dothan will be manufactured on a 90nm process.

http://news.com.com/2100-1006-5141239.html

Sonoma (3rd Generation Centrino) is also expected at the end of the year. All we know right now is that it'll be Tri-Band WiFi (802.11a/b/g).

http://news.com.com/2100-1006-5078270.html?tag=nl

[Drakino]

Originally posted by csoledad:
I want a Powerbook G5 as much as the next guy, maybe more.

However I recentely learned the memory/bus controller developed by Apple (still on the old process) consumes 30 watts typical.

I hope the 30 watt figure is wrong, but if it is true the problems are more than G5 power and heat.

Well, it would be wrong for a Powerbook G5 controller. The controller in the desktops/servers is built around being a dual processor controller. Because of the dedicated links to the system controller that the G5s have, this dual processor controller will be much more complex then a single G5 CPU controller. Easially shave 20%, if not more for the reduction in heat from getting rid of a processor bus. Then, add a 0.09 micron process, and you come within specifications for a laptop.

I still don't see a G5 Powerbook until late 2004, possibly early 2005 though. Apple Portables are doing really well now, and will continue to do so. The G5 jsut isn't quite needed yet to drive sales.

[Eug Wanker]

Originally posted by Drakino:
Well, it would be wrong for a Powerbook G5 controller. The controller in the desktops/servers is built around being a dual processor controller. Because of the dedicated links to the system controller that the G5s have, this dual processor controller will be much more complex then a single G5 CPU controller. Easially shave 20%, if not more for the reduction in heat from getting rid of a processor bus. Then, add a 0.09 micron process, and you come within specifications for a laptop.

Pismo2001 also comments about the memory controller's potential power savings.

I still don't see a G5 Powerbook until late 2004, possibly early 2005 though. Apple Portables are doing really well now, and will continue to do so. The G5 jsut isn't quite needed yet to drive sales.

Well, the competition is ramping up in Q2, with new Pentium M's coming online, and I suspect that PowerBook sales may slowly decline slightly, until the G5s appear.

I know for one, I will not buy another PowerBook until it goes G5, and I have the cash for a new one right now.

I'm predicting mid-2004 for a G5 PowerBook.

Mid-2004 is technically possible, and it's perfect timing for a refresh, with the last one having been in September. And if not mid-2004, sometime in 2004 for sure. Definitely not as long as 2005.

[ae86_16v]

Originally posted by Eug Wanker:

Mid-2004 is technically possible, and it's perfect timing for a refresh, with the last one having been in September. And if not mid-2004, sometime in 2004 for sure. Definitely not as long as 2005.

Well, I don't think we mean that there isn't going to be another refresh. What I meant was that I think there are going to be at least another G4 Speedbump before we see a G5.

[Eug Wanker]

Originally posted by ae86_16v:
Well, I don't think we mean that there isn't going to be another refresh. What I meant was that I think there are going to be at least another G4 Speedbump before we see a G5.

I disagree with that too actually. I'm thinking that the next refresh, whenever that may be, will be with the G5. However, I could be totally wrong of course.

[mishap]

Originally posted by Eug Wanker:
I disagree with that too actually. I'm thinking that the next refresh, whenever that may be, will be with the G5. However, I could be totally wrong of course.

What could they do to the current g4 powerbooks? speedbump them .05 mhz?

[Patcarla]

Originally posted by ae86_16v:
Well, I don't think we mean that there isn't going to be another refresh. What I meant was that I think there are going to be at least another G4 Speedbump before we see a G5.

Why? and out of curiosity, what G4 speedbump do you expect?

[ae86_16v]

Originally posted by mishap:
What could they do to the current g4 powerbooks? speedbump them .05 mhz?

Originally posted by Patcarla:
Why? and out of curiosity, what G4 speedbump do you expect?

Okay, I am just guessing here. Here is a shot, reintroduction of L3 Cache back into PowerBook line (i.e. use 7457 processors). And bump them up to 1.42GHz or 1.6GHz, but I think I remember seeing that Moto's 7457 maxes out at 1.33GHz. Higher bus speed, say 200MHz, faster HD across the line, faster SuperDrives, maybe a 9600 Pro.

But the main reason I believe that the G5 will not come until 2005 is because not too many months ago, I remember IBM saying that the current version of the G5 isn't portable friendly. They they'll have to make a new chip for the PowerBooks. I take that they have to develop a new chip from scratch. Now, I don't remember who said that or exactly where (too lazy to do the search), but I think the key of it. Keep in mind that the PPC970 was derived from the Power4 which was a Server chip. So the 970 in a sense was/is already stripped down.

But this 12.3w @ 1.4GHz is very encouraging.

Originally posted by Eug Wanker:
I disagree with that too actually. I'm thinking that the next refresh, whenever that may be, will be with the G5. However, I could be totally wrong of course.

All in good fun.

[Patcarla]

Yep you may be right..
But today is 20th Mac birthday!! I read somewhere that they're planning something big for the monday 26th..One can dream of a powerbook G5 even though an iMac G5 is more likely..

[Simon]

I'm with ae86_16v rather than with Eug on this one.

I believe many of us have just got tired of the G4 and are kind of fed up with the minor speed increases in the PB line. And of course many want to show off big numbers to their Centrino friends.

Problem is, it's not about what we want or wish for, but what IBM can deliver for Apple's PB boundry conditions. And to me that points more to another G4 rev until the PB's get a G5.

There's a reason that every Apple rep asked on the subject has tried to be very careful and not get people's expectations up.

Indeed however, the wattage of the 970fx looks very nice.

[CIA]

I seem to recall that Apple and IBM announced awhile back that they were co-developing a version of the power4 or 5 specifically for low power (Read: Powerbooks) use.

Regardless of business tactics, the sooner apple goes 64bit across the whole line of Macs the better. Then we may finally see OS11 and it's true, full blown 64bit clean OS goodness.
What better reason to finally jump names from 10-11? 10 becomes "Classic" and 11 is a 64bit powerhouse with all the trimmings!

[all2ofme]

Do we have any evidence of any low-powered G4s faster than the ones in the current PBs?

If they do exist then I think we're more likely to see Apple put them in a PB soonish as a stop-gap before the G5 finds its way in there.

If not, I'm with Eug. And I'll be buying one when he does, too

[Eug]

Originally posted by all2ofme:
Do we have any evidence of any low-powered G4s faster than the ones in the current PBs?

No we don't. It is true however, that Apple has released products before with G4s that Motorola had no documentation for on their website.

The concern I have here is the speed and the heat of the G4. Yes, you read that right... In some ways, I'm more concerned with the heat of the G4 than the heat of the G5. Motorola's G4 7447 chips at 1.33 GHz are already pretty hot, at 30 W max. IIRC, these are the hottest chips that Apple has ever used in a laptop. It may not be feasible for Apple to use another 7447 if it's just a scaled up part. It'd have to be a new lower power version of it that can handle Apple's CPU and bus MHz needs (possible), or else a completely new chip category for them (such as the fabled but vapourware 7457-RM with integrated memory controller and hopefully better power characteristics). OTOH, the power characteristics of the G5 1.6 are probably in the same ballpark as the G4 1.33, so it seems like a logical upgrade.

Also, there would have to be other compelling reasons to stay with the G4 like scalability up to say 1.8+ GHz (over 2 revisions) with reasonable heat dissipation specs. If not, then it makes sense to go directly to the G5 1.6.

Now the concerns about the G5's system controller are valid, but we have to remember that it's gonna get a die shink to 90 nm, and of course it's not as if we need dual 1 GHz busses in our single-CPU laptops running at 1.6 GHz. Similarly, in terms of the memory controller, we won't be running either dual channel DDR400. The likely controller will simply only need to handle single channel DDR333. Certainly, it's not as fast as the desktops, but it's already twice speed of CPU-memory access on the G4 AluBooks, using the exact same memory.

It seems unfathomable that Apple and IBM have been working together for so many years without the G5 in mind for a laptop use early on. I'd guess that the vast majority of these design considerations have already been addressed at this point, and it's just a matter of time before we see the G5 PowerBook (even if it's after the G5 90 nm Power Macs and G5 iMac).

If not, I'm with Eug. And I'll be buying one when he does, too

Well, I REALLY want a rev. A G5 PowerBook, but the practical side of wonders if I should wait until 2005 for a Rev. B...

[all2ofme]

Hmm, well it definitely sounds like you've done your homework. When it boils down to it, it seems to me that the main reason most aren't expecting G5s PBs any time soon is that it would be too good to be true given past experiences with Apple taking their time with new processors (or that their suppliers have held them up each time)!

I may be reading too much into it, but Smalldog normally seem to have a decent stock level of most of Apple's gear and they have next to nothing of anything in the current range at the moment.

No new 15" PBs whatsoever, a single 17" and only seven 12" machines. That's a total of eight Powerbooks. Not clever stock levels if there's nothing coming. Of course it could just mean that the current machines have been unexpectedly popular...

The same is true for the G5s, though - two 1.6GHz machines, four 1.8 duals and one dual 2GHz. Total of seven.

I've got my fingers crossed. I'm planning a trip to the US soon as it is, and being able to bring back a new PB would be great news.

[workerbee]

Originally posted by Eug:
Well, I REALLY want a rev. A G5 PowerBook, but the practical side of wonders if I should wait until 2005 for a Rev. B...

I wonder how we'll be able to resist the G5's siren songs, once it will be available.
I almost bought a 1.33GHz 17" yesterday, using this exact logic ("Never buy a Rev A"), but the more I read people phantasizing about tasty 1.6GHz G5s with real fast RAM in PowerBooks, the more I waver.

[djohnson]

I am hoping to get a Rev A. G5 PowerBook. I know all about getting rev a stuff... I will be buying out of necessity though...

[Simon]

Originally posted by CIA:
Regardless of business tactics, the sooner apple goes 64bit across the whole line of Macs the better. Then we may finally see OS11 and it's true, full blown 64bit clean OS goodness.
What better reason to finally jump names from 10-11? 10 becomes "Classic" and 11 is a 64bit powerhouse with all the trimmings!

You will probably have to wait quite a while for this to happen, even if we are all-G5 by the end of 2004. Apple tends to be very conservative with such things. Just think of the fact that in System 9 there was still non-PPC 68k code or the fact that Apple still sells systems that boot OS 9.

Of course they're a company that doesn't want to lose any customers. So leaving people behind is not something they really like to do.

I suppose for quite some time (3 years maybe) what we will get is an OS that has 64bit calls that replace 32bit ones where it really gives a good benefit, a fat OS so to say (like PPC and 68k). But rather not an OS that leaves everybody with G3s or even G4s behind.

[Crusoe]

I think we've seen the last G4s in Powerbooks.
There's no speed bump to go to.

G5s are next.
I would imagine that they could release a 1.6 with the system bus running at 400Mhz (instead of 800Mhz) This would be an enourmous improvement over the G4s 167 and keep wattage and heat down.

So perhaps

12" at 1.4 with 350Mhz Bus
15" at 1.4 and 1.6 with 400Mhz Bus
17" at 1.8 and a 450Mhz Bus

I don't know if Bus Controllers can be scaled back, but imagine so. Also do the G5 chips have some sort of speedstep technology to have them run in power saver mode?
Scaling back 1/3 or 1/4 to 933Mhz, 1.2Ghz and 1.33Ghz or something similar would be a requirement not an option.

but what RAM would they use?
PC2700 for the 12" and PC3200 SODIMMS for the rest.

Changes that should be made to general design.
Ambient Light sensor move to the front of the grill instead of near the screen.
Higher capacity batteries.
Full size keys on 15 and 17" models.
Grey power adapter and cords.
Optical in and out.
Perhaps dump the PC Card port.

When?
September at the earliest. This is a huge engineering feat. Trying to satisfy the powerbook lovers who want fast, quiet, cool, with good battery life at a reasonable price is gonna be tough.

Complaints folks will have fan too noisy, too hot, battery life stinks, too expensive and what no dual processor in the 17". Oh and the ever famous, it hisses/makes electrical/mechanical sound. If it hisses it's because the OS is named after a cat, if it makes an electrical sound perhaps there's some sort of electronics inside, if it make a mechanical sound perhaps some sort of mechanics was needed to be able to insert and eject CDs and perhaps spin the hard drive.

Price?
12" Combo $2199
12" SD $2399
15" Combo $2599
15" SD $2999
17" SD $3399
if we're lucky

I'm hoping to salvage about $1400 out of my current ALBook 15.2 when the time comes. I'll hop on board with a Rev A without fear or hesitation.

[Simon]

Originally posted by Crusoe:
Full size keys on 15 and 17" models.

Already the case, even on the 12" PowerBook.

[Crusoe]

Originally posted by Simon:
Already the case, even on the 12" PowerBook.

Arrow and Fn Keys?

[Crusoe]

Originally posted by Simon:
Already the case, even on the 12" PowerBook.

Arrow and Fn Keys are half sized.

[Simon]

Originally posted by Crusoe:
Arrow and Fn Keys are half sized.

They're also like uh half-used...

[Crusoe]

Originally posted by Simon:
Already the case, even on the 12" PowerBook.

What about the Fn and Arrow keys?

[uv23]

Originally posted by Crusoe:

Optical in and out.

Definitely. I'd love to see a high quality audio in/out in the G5 PBs.

[Drakino]

Originally posted by Crusoe:
G5s are next.
I would imagine that they could release a 1.6 with the system bus running at 400Mhz (instead of 800Mhz) This would be an enourmous improvement over the G4s 167 and keep wattage and heat down.

Bad idea. Slowing down the bus on a G5 is devistating to them. Remember how Apple touted the G4 as being better then a P4 because of the very short pipeline? The first G4 processors had a 4 stage pipeline, while the newer ones have 7. Now, the G5, has up to 23, depending on where the data goes. Now, if you remember, their big argument was that a longer pipeline causes problems when the wrong data gets sent since the entire pipeline has to be cleared.

The part about that presentation that always bugged me is that they never talked about how bus speed factors into this. The faster the bus, the quicker you can get that data back in the right order and into the processor. So, we have pipeline that is more then 3 times as big as the current G4s, so the bus speed has to be at least 3 times as fast as the G4 to be as fast clock per clock cycle. Thus, a G5 would have to have a bus of 500mHz at 1.4 gHz to keep up with a G4 at 1.4 gHz.

The P4 had this problem in the earlier days when it came out. At low speeds, like the 1.3gHz P4, the P3 at those speeds (or even slower speeds) creamed it. This was due to several factors, one being the 400mHz bus not being a big enough jump over the P3, and also the slower memory usually tucked into those P4 systems (PC133 was common in the low end P4 systems).

The bad thing though, is I could see Apple pulling another backstep in peformance, similar to when the TiBooks first shipped. So, I guess I should restate my estimates. I don't see a Powerbook that I can justify upgrading to over my 1.25 G4 until early 2005. The G5 Powerbooks, if launched in a neutered state, should be better by then.

[Crusoe]

Agree it's not the best solution but if the controllers need 30w, that's too much heat and power consumption. 450Mhz bus on a 1.8 isn't optimal but neither is 167 on a 1.33. Maybe they can trottle it back 25% instead of 50% put the 1.8s on a 600Mhz bus. It's a tough balancing act for the engineers I'm sure.

[Eug]

Slowing down the bus on a G5 is devistating to them.

Remember, the memory would be single channel DDR333 in a PowerBook, not dual channel DDR400 like in the Power Macs. So one has room to slow down the FSB since memory speed is the bottleneck.

[ae86_16v]

Originally posted by Eug:

It seems unfathomable that Apple and IBM have been working together for so many years without the G5 in mind for a laptop use early on. I'd guess that the vast majority of these design considerations have already been addressed at this point, and it's just a matter of time before we see the G5 PowerBook (even if it's after the G5 90 nm Power Macs and G5 iMac).

Eug. . . like I pointed to earlier. The PPC970 is a reality simple derivative from the Power4. So I don't know about that. . . again I think they'll have to manufacture a new Laptop version as I pointed to earlier.

[Eug]

Originally posted by ae86_16v:
Eug. . . like I pointed to earlier. The PPC970 is a reality simple derivative from the Power4. So I don't know about that. . . again I think they'll have to manufacture a new Laptop version as I pointed to earlier.

Not really. Yes, it's a stripped down POWER4, but there have been substantial design considerations taken into account specifically to maximize power efficiency. Remember that IBM seems to be positioning this chip at the (high-end) embedded market. You don't get far into the embedded market without an energy-efficient design. Plus, the addition of an Altivec unit is nothing to sneeze at. IBM has traditionally shunned Altivec in their chips, but it seems IBM has designed this chip with Apple in mind specifically as well. And part of Apple's main business is laptops.

[Drakino]

Originally posted by Eug:
Remember, the memory would be single channel DDR333 in a PowerBook, not dual channel DDR400 like in the Power Macs. So one has room to slow down the FSB since memory speed is the bottleneck.

Well, DDR400 memory for laptops does exist, and should be more commonplace here soon. Very good point though, will Apple go to a dual channel interface for memory in laptops (possibly building in 512mb RAM on the motherboards, and putting in 2 DIMM slots), or reduce it to a single channel? (Ironic, since DIMM means Dual Inline over the old need to pair SIMMs)

The 15 and 17 inch Powerbooks have room to do this. The 12, might, but the fact that it has one DIMM slot now is not a good sign. I wonder if that might be part of the holdup, the 12 inch model... I bet we see a G5 in the 15 and 17 first, with the 12 inch bumped up a notch on the G4 scale one last time. All depends on how busy the hardware teams are, and how quickly they want to release something after design. The 12 inch Powerbook was actually done by the iBook team in their downtime, while the Powerbook team worked on the 17 inch Powerbook. With the G4 iBook out, the major changes to the iBook are a ways out.

(All speculation on my part, none of this comes from any source, except the 12 inch Powerbook design team comment)