[Eug]

Original Japanese article & Babelfish translation:

"IBM in the lecture, "Design and Implementation of the POWER5 Microprocessor", adopts 130nm process, accumulates 2 hundred million 7600 ten thousand transistors, announces concerning the design of the POWER5 processor which is driven above 1.5GHz. In addition, in the lecture, "PowerPC 970 in 130nm and 90nm Technologies", it announces concerning PowerPC 970 which introduces 90nm SOI technology."

So I wonder if this means we won't see 90 nm chips until mid-2004 or later. It makes you wonder if the 3 GHz G5 is 130 nm or 90 nm. And I guess I won't get my G5 PowerBook until about a year from now.

[Landos Mustache]

Originally posted by Eug:
I won't get my G5 PowerBook until about a year from now.

I think that is what all the rumors have been sayin' for a while now anyway.

[Eug]

Originally posted by Landos Mustache:
I think that is what all the rumors have been sayin' for a while now anyway.

Esp. since that's what Teh Steve� said too.

[toshigen]

How many nanometers are the G4 processors being used in current Powerbooks... just out of curiousity?

How much power savings/heat dissapation could you expect from going from 130nm down to 90nm?

[robo from AI]

Then again, the G5's in Apple's machines appeared much earlier, and at higher speeds, than what people expected from IBM's announcements, which turned out to be _exteremely_ conservative...

So there is hope for a 90nm process G5 somewhat earlier than one would normally expect following a Feb '04 info session.

[Eug]

Then again, the G5's in Apple's machines appeared much earlier, and at higher speeds, than what people expected from IBM's announcements, which turned out to be _exteremely_ conservative...

So there is hope for a 90nm process G5 somewhat earlier than one would normally expect following a Feb '04 info session.

You are right, although I should point out that the details of the G5 were presented in October 2002, and we didn't get the Apple G5 announcement until mid-2003. That was a whole 7 months later.

But yeah, I think it's reasonable predict a 90 nm G5 Power Mac at mid-2004 (which would suggest the 3 GHz G5 will be 90 nm) but it's also possible it could come later. Interestingly though, this is in line with the timing as laid out in eWeek's article about mid-2004 IBM blades using an updated chip, and we all know the 130 nm PPC 970 IBM blades came out a lot later than the G5 Power Mac.

How many nanometers are the G4 processors being used in current Powerbooks... just out of curiousity?

Current PowerBooks are 130 nm. Current iBooks are 180 nm, like the previous generation PowerBooks.

How much power savings/heat dissapation could you expect from going from 130nm down to 90nm?

I dunno. Smaller dies means less power, but people tell me that at this size there is also potentially significant leakage, negating some of the benefit.

It will also be interesting to see if they introduce more power management features in the 90 nm version and/or introduce a separate ultra-low voltage version specifically for laptops and the embedded market.

[PoisonTooth]

Originally posted by Landos Mustache:
I think that is what all the rumors have been sayin' for a while now anyway.

And I wouldn't buy Rev A of the G5 PB unless someone paid me. No way.

[Eug]

Originally posted by PoisonTooth:
And I wouldn't buy Rev A of the G5 PB unless someone paid me. No way.

Yeah, that worries me too. I just don't know if I could hold off updating my TiBook SD until 2005 for a rev. B G5 PB.

Mind you, I could theoretically wait until near the end of 2005 and still have warranty left (AppleCare), which would be a bonus for anyone buying the TiBook.

It'd be nice though getting a rev. B 2.0 GHz G5 PowerBook on 90 nm with 1 MB L2 cache, in spring 2005. Maybe by then we'll have faster laptop memory too.

[Eug]

AppleInsider claims that although 2.5 GHz 130 nm parts exist, all Apple machines with G5s over 2 GHz will have 90 nm parts.

Also, the claim is that going from 130 to 90 nm for a 2.5 GHz part changes the power utilization from 96 to 62 Watts.

They also talk about POWER5 derivatives.

[thunderous_funker]

Even at 90nm, wouldn't it still be a bit too hot for an air-cooled chasis? At least the kind of super slim, relatively cool chasis we've come to love.

Maybe I'm the only one, but the powerbook would loose much of it's appeal if it started resembling the hefty, bulky, space-heater p4 laptops of the last couple years.

And the power consumption would mean pathetic battery life with current technology, unless I've mistaken the comparitive facts and figures.

Seems to me, a powerbook G5 will be waiting on liquid cooled chasis and revolutionary battery technology longer than it will be waiting for IBM chips.

Am I way off?

[xylon]

Originally posted by thunderous_funker:
Even at 90nm, wouldn't it still be a bit too hot for an air-cooled chasis? At least the kind of super slim, relatively cool chasis we've come to love.

Maybe I'm the only one, but the powerbook would loose much of it's appeal if it started resembling the hefty, bulky, space-heater p4 laptops of the last couple years.

And the power consumption would mean pathetic battery life with current technology, unless I've mistaken the comparitive facts and figures.

Seems to me, a powerbook G5 will be waiting on liquid cooled chasis and revolutionary battery technology longer than it will be waiting for IBM chips.

Am I way off?

I believe you've got the numbers backwards. A chip made with 90 nm would consume less power and thus generate less heat. I am no expert though and I could most definitely be wrong, but from the threads I've read the benefits of moving from a 130 nm to a 90 nm process are less power consumption and lower heat dissapation among other things.

I believe also that someone pointed out in the PowerBook forum that with the current power consumption and heat dissipation numbers tentatively associated with a 90 nm chip, it is possible to put the G5 chip into a PowerBook without adverse effects, but the chip would be slowed down some.

Originally posted by Eug:
I dunno. Smaller dies means less power, but people tell me that at this size there is also potentially significant leakage, negating some of the benefit.

I remember reading somewhere that someone had invented a sort of polymer coat that severly diminshed the leakage or electron hopping. Pretty cool stuff and if it's true, it might negate that problem. Can't remember where I read that though, I need to take better notes about the things I read...

[Eug]

Originally posted by thunderous_funker:
Even at 90nm, wouldn't it still be a bit too hot for an air-cooled chasis? At least the kind of super slim, relatively cool chasis we've come to love.

Maybe I'm the only one, but the powerbook would loose much of it's appeal if it started resembling the hefty, bulky, space-heater p4 laptops of the last couple years.

And the power consumption would mean pathetic battery life with current technology, unless I've mistaken the comparitive facts and figures.

Seems to me, a powerbook G5 will be waiting on liquid cooled chasis and revolutionary battery technology longer than it will be waiting for IBM chips.

Am I way off?

You're assuming that it's going to be the exact same design as the destop chips. I'm hoping for this however:

introduce a separate ultra-low voltage version specifically for laptops and the embedded market.

In fact, as speeds increase, something like this will be necessary.

[thunderous_funker]

I'm having trouble finding the specs but here is what I can tell:

The current G5 at 1.8Ghz uses 42W. The current G4 uses 15W.

If the 2.5Ghz G5 at 90nm uses 62W, where does that leave us for a laptop chip?

To get inside a powerbook chasis, you'd have to abandon the speed gains of the 90nm process. And I'm not even sure if a 90nm running at 1.8 Ghz (or lower) would get anywhere near the sub 20W range.

So even if we imagine that the big brains at IBM figure out how to dumb down a 90nm G5 to get somewhere below 40W (which I'm not sure is possible), we'd still be talking about something that would eat through a current technology laptop battery in short order.

So unless there is some drastic revolution in laptop bettery technology, I just don't see how a G5 is going to make it. For it to even approach the realm of feasability at a considerable distance, they'd have to dumb the chip down to below 1.8 Ghz so you'd have to question the advantage of using it in the first place.

I can't find the heat dissipation numbers anywhere, but even at 90nm and very low speeds, I imagine the G5 is going to require liquid cooling before it is even remotely feasible for a laptop.

I'd be happy to be wrong. Elvis in Heaven knows I'd love to see a G5 laptop (provided it isn't a massive space heater), but I just don't see how they are going to manage it.

There are rumors of liquid cooling technology that might handle the heat, but it seems to me it would take a Sterling Engine to give that thing more than a few minutes of battery life.

[xylon]

Originally posted by thunderous_funker:
I'm having trouble finding the specs but here is what I can tell:

The current G5 at 1.8Ghz uses 42W. The current G4 uses 15W.

If the 2.5Ghz G5 at 90nm uses 62W, where does that leave us for a laptop chip?

Huh, I guess I misread the numbers. I've been under the impression that the 90 nm process decreases power consumption. Great, these numbers definitely dismiss the possibility of a current G5 laptop (not that we haven't known that for a while).

I do wonder how much slowing the clock speed of a chip reduces power consumption though. I mean if a G5 clocked at 1.25 Ghz was outputting the same heat as a the current G4s that'd be kind of interesting. But then, I guess, the questions is whether a 1.25 Ghz G5 would be inherently better than a 1.33 Ghz G4.

So many questions.

[Eug]

The G4 1.25 GHz is 18 W typical, and 26 W max. The G4 1.33 is even higher.

The G5 1.2 GHz at 130 nm is supposedly 19 W typical (although you can't always compare the numbers). I dunno what the max is.

The G5 2.5 90 nm number of 62 W (if true) might be the max power I'm guessing esp. since they say the 2.5 130 nm would be 96 Watts. (96 W typical would be horrendously high. )

With a say sub 2 GHz 90 nm part with additional tweaking and a lowered voltage it would be fine. Low voltage chips cost significantly more, but it's worth it for a laptop.

[thunderous_funker]

Originally posted by xylon:
Huh, I guess I misread the numbers. I've been under the impression that the 90 nm process decreases power consumption. ...

You're not confused. You had it right. It does lower the power consumption and heat dissipation--at the same clock rate.

By how much, I'm not sure. Could a 90nm G5 running at 1.8Ghz approach 15W (the status quo)? I'm skeptical. I doubt it would drop below 30W.

Can some of the smart people around here extrapolate the data we know and make a decent guess at it? Like I said, I'd love to be wrong.

[Eug]

By how much, I'm not sure. Could a 90nm G5 running at 1.8Ghz approach 15W (the status quo)? I'm skeptical. I doubt it would drop below 30W.

As I said in the previous post, 15 Watts is not the status quo. It's closer to 30 W (max), at least in the top-end 17" AluBook.

I'm guessing (with no understanding of chip design) that they could easily make a G5 1.5-1.6 usable in a laptop for sure, and could go higher with further tweaks to the design.

[thunderous_funker]

Originally posted by Eug:
The G4 1.25 GHz is 18 W typical, and 26 W max. The G4 1.33 is even higher.

The G5 1.2 GHz at 130 nm is supposedly 19 W typical (although you can't always compare the numbers). I dunno what the max is.

The G5 2.5 90 nm number of 62 W (if true) might be the max power I'm guessing esp. since they say the 2.5 130 nm would be 96 Watts. (96 W typical would be horrendously high. )

With a say sub 2 GHz 90 nm part with additional tweaking and a lowered voltage it would be fine. Low voltage chips cost significantly more, but it's worth it for a laptop.

Crossed posts with you...

Ah, so those numbers are max rather than mean power usage? That would make a lot more sense. Are we sure that is the case though?

I'm having trouble finding any reliable data on the subject.

[Eug]

Originally posted by thunderous_funker:
I'm having trouble finding any reliable data on the subject.

That's because IBM hasn't released any info on it yet.

The G5 2.5 90 nm numbers are from AppleInsider, which claims to have inside info.

[Big Mac]

I think it's important to look at the rest of the industry (meaning Intel), because where it has gone with the Centrino crap changes the competitive landscape for laptops, at least as far as GHz are concerned. Look at the PC laptop offerings there are: You still have a couple of lap-desktop, twenty minute battery life, 2.xGHz P4, and then you have the Centrinos which are running at what, 1.3 or 1.4GHz? The funny thing is, the Centrinos cost more! This is very good for Apple, because at least we don't have to contend with the MHz Myth in the laptop department.

Yet, it also raises the interesting question of where Apple will take the PB from here. Eventually there will be a G5 in the PB-that much is obvious. Anyone who believes there won't be just needs to look at the history of laptop chip improvements. Not too long ago people and pundits alike swore that there would never be a G4 based PB for the same reason people are now swearing the same about the G5. (Heck, the same argument was made about the 604 and even the 68040, if I remember correctly.) As the chip and fabrication technology matures, each new generation eventually makes its way into the PB. Those who thought it was/is going to happen over night were/are foolish, but the opposite sentiment-that it will never happen-is silly too.

IBM will improve the chip and process in order to derive a variation suitable for laptop use. I'm sure Apple has already negotiated that in general terms with IBM. It will eventually happen, but it's going to take some time. I don't know what Apple's going to do until then; that stretch of time maybe a painful period. Now that the iBook has gone G4, Apple is in an arguably worse position-it has to rely more or less completely on Motorola for laptop chips. That is a scary prospect, but thankfully we're not stuck with Motorola for Power Mac chips anymore. I suppose the lesson is you can't escape opportunity cost.

Finally, Apple will follow Wintel's lead and try to beat them in clock speed without trading too much battery life. We're certainly not going to see 2.5GHz G5s in PBs anytime soon, because it would be too costly. But again, we don't need that much to be competitive with Wintel. I expect Apple to match or slightly exceed them in clock speed, and that shouldn't be too difficult. After all, we're talking about a) IBM (which has already exceeded our expectations) and b) the PPC. The expectation was that the PPC could provide higher performance and higher efficiency compared to the x86. And if the original marketing of the PPC is at all applicable to contemporary circumstances, then IBM should be able to outdo Intel in the laptop chip department.

[Eug Wanker]

You still have a couple of lap-desktop, twenty minute battery life, 2.xGHz P4, and then you have the Centrinos which are running at what, 1.3 or 1.4GHz?

The Centrino runs at up to 1.7 GHz, and is very efficient. It has a thermal design power of about 25 Watts. TDP can't be compared to PPC power ratings accurately, but probably is somewhere around what would be called the max power rating of the PPC chips.

Centrino is quite a nice chip, and probably will be much faster than 1.7 GHz by the time fall 2004 rolls around.

[kupan787]

If I remember correctly, when IBM first introed the 970, they said at 1.2GHz on the 130nm process would yield 19W. So my guess is a 1.6 (or maybe 1.8) on 90 nm could easily fit in a laptop.

Remember, we aren't expecting 2.5GHz 970 laptops here. The current top of the line is 1.33, so the next gen high need only be up to 1.6 (or 1.8 if it could work out).

[MikeJW]

So maybe to clear out the remaining G5's apple could offer an entry level 1.8 GHz G5 ($2000), a mid range 2.0 (with Radeon 9600 for $2500) and a dual 2.0 (with radeon 9800 for $3000)

Then on Valentine's day, New G5's with entry level 2.0, mid range 2.5 and top of the line dual 2.5. All of this on 0.13

Then in about 6 months, with the mighty 0.09 micron G5s.
Entry level 3.0 GHz 1 MB L2,
Mid range dual 2.5 1 MB L2
Top of the range, dual 2.0 dual cores 1.5MB L2 shared

Then move to PPC 980

Anyone care to sound off.

[Eug Wanker]

IBM plans to boost the 970's clock speed to 2.4GHz in mid-2004.

I hope they're only talking about the blades (which will ship at only up to 1.6 GHz initially). If 2.4 GHz is the max overall for the G5, that would be a downer.

[FlatLyna]

If a G5 @ 1.2GHz or thereabouts used 19W, similar to a G4, I wonder about the posibility of a desktop replacement laptop model that ran at say 1.2GHz on battery and maybe 2GHz when running of the power adapter.
In that case I guess the challenge would be the cooling system required. A cleverly designed docking station could perhaps also combine with a cooling solution too.

[Superchicken]

The 17 inch is pretty much a desknote anyway nobody's gona bring that big boy to class, I think it's fairly feasible if they wanted they could put a 1.6Ghz in that perhaps room for a second battery, and you'd have a very competitive speedwise notebook with the PC world. Then you throw in a 1.3Ghz into the 15 inch PB, and a 1.1Ghz into the 12 inch and you'd be fairly good to go. Seems like this would make a fair bit of sense...