Motorola at 90-nanometers

[arn]

Reuters (http://story.news.yahoo.com/news?tmpl=story&ncid=581&e=1&cid=581&u=/nm/20020830/tc_nm/tech_malaysia_motorola_dc) reports that Motorola, STMicroelectronics and Philips have partnered to produce 90-nm chips:

U.S. based Motorola said on Friday its partnership with two other semiconductor makers will probably be the first to produce a new generation of microchips, beating Intel by at least six months

No details on how this will influence PowerPC production.

[barkmonster]

Well that's a good sign for the future. I mean I'm sure they could pump the clockspeed up without any negative performance hits (pipeline stage increases etc...). Maybe that huge fan in the new G4s are for some new moto cpu, maybe the people complaining about no (non existent) PPC7470 being in the new towers will get their new cpu early next year and then apple are saving the IBM PowerPC chip for a rollout of a 64bit powermac a year from now.

Of course all that speculation aside, it's good to see motorla advancing again instead of sitting on their hands like they've done for a few years.

[User X]

We can only hope this will lead to faster Power PC chips. Somthing has got to give.

[Rocketman]

90 nm should produce both higher speeds and lower temperatures at current speeds. But there seems to be no indication of timeframe so I would conservatively assume > 1 year.

Rocketman

[dongmin]

Belden also said the alliance would start production of a high-performance chip by the fourth quarter of this year and a lower performing version in the third quarter of 2003.

Interesting...but it sounds like these chips are mainly for portable devices like phones, etc. I doubt a partnership with Phillips and STM will have much benefit for Macs.

It'd be interesting if Apple decided to go with IBM for desktops, where heat and power is not a concern, and Motorola for laptops and devices.

My feeling is that Motorola is abandoning development for chips designed to run on higher-end PCs. They seem interested solely in the embedded market (network equip., phones, consumer devices, etc.).

[drastik]

I'm not sure we are ever going to see one of these in a Mac. Personally, I think that the Power4's are the way to go. Motorolla has been jerking apple around for two years now. Even if they make a great chip, drop them like a rock, I say.

[ddtlm]

I'm sure G4's will remain Apple's low-end chip for a long time because the Power4-lite will probably be very expensive (and because Apple wants there to be a performance division between top and bottom end).

There is no question that Moto will some day make 90nm G4's (or equivilent). Not only are they better than what we have now because they are cooler running and can be clocked higher, but they are also cheaper to make! A 90nm P4 from Intel is probably not a whole lot more expensive to make than a 180nm G4 from Moto, despite the P4 being so complex and the G4 so simple.

Anyway, if Moto if gona beat Intel by 6 months, that means they should be rolling off chips about, oh, now. Is that happening? No. Intel's 90nm P4 is scheduled for first half of 2003 easily, and it will pack a full 1mb of L2 cache on die. Ouch!

Hopefully the Power4-lite is moving along...

[BRussell]

Originally posted by drastik
I'm not sure we are ever going to see one of these in a Mac. Personally, I think that the Power4's are the way to go. Motorolla has been jerking apple around for two years now. Even if they make a great chip, drop them like a rock, I say. OK, they use the Power4-lite in the PowerMacs. But then what about the other lines of Macs? They're surely not going to put a Power4 in the iMac/eMac, let alone the PowerBook, for some time. Do they stick with the c. 1Ghz Apollo for the next several years until the Power4 variant is small enough?

I think the key to Apple's future chip fortunes is diversification. IBM's Power4-variant in the PowerMacs, Motorola's .09µ G4 in the iMac/eMac/PowerBook, IBM's Sahara and subsequent variations on it in the iBook. Then they can see how the various lines emerge over 2003-2004, and make a choice then if necessary.

[AmigaMac]

Well with Amiga's reentry into the hardware space and adopting IBM's offerings should help push things along in the PowerPC front. Of course Nintendo's presence is also helpful. So I think things can only get better!

[Timothy]

All talk...

We don't even have any 1.25 ghz chips yet, and probably won't for 2-3 months yet. So, I have very little hope for seeing any "new" chip anytime in the next 6-9 months.

Pathetic...

[chubakka]

Last time I checked it was 4-6 weeks...

[ddtlm]

Timothy:

Extremism appears to be part of the whole Apple-user thing. First, people (not incluing me) run off and believe unrealistic rumors about dual 1.6's and DDR FSB's. Next, people (not including me) run off and deny that Apple is ever going to ship thier 1.25.

Ehem. It is announced, and the weeks-till-shipping are counting down as time passes (it was recently pointed out to me that they started with an ETA of 6-8 weeks and did drop to 4-6 weeks at the proper time).

[gregorypierce]

Sadly I agree with Timothy. While its nice that they made their announcement before Intel, unless they can produce tangible VOLUMES of chips using the process - it is a meaningless milestone.

[Rocketman]

Originally posted by chubakka
Last time I checked it was 4-6 weeks...

Inconsiderate people ordered a bunch of them thus delaying your access to them.

:)

[chubakka]

I think if the chips were intended for Apple hardware they would have been sure to mention it in thier press release. Also, if there was some spped breakthrough the would have hyped that too.

[alex_ant]

Motorola? On the cutting edge of microchip design? Maybe the article meant 90nm chocolate chips.

[BRussell]

Originally posted by chubakka
I think if the chips were intended for Apple hardware they would have been sure to mention it in thier press release. Also, if there was some spped breakthrough the would have hyped that too. No way. Only Steve and his ego announce things for Apple - you should know that. Not in a million years would they talk about Apple in this context.

Of course, that doesn't mean it will be used by Apple. But not mentioning Apple certainly doesn't mean Apple won't use it.

[ddtlm]

chubakka:

Has IBM or Moto ever announced that a chip they are making is for an Apple product before Apple does? For example, IBM announced the 750FX long before it turned up in an iBook, and as far as I know they never mentioned Apple as a customer.

At this point denying that the Power4-lite will end up in a Mac is every bit as unrealistic and extreme as everyone who proclaimed G5's last spring and everyone who talks about 7470's to be announced at any moment.

We have to look at the two most important facts: (1) Apple needs a chip like the Power4-lite, and (2) IBM is designing/producing it. IBM will not turn down a customer... the ball is in Apple's court. Who really things they would choose to stay with G4's?

[chubakka]

I just think that we know WAY less about these chips then we do about the new IBM PowerPC chip. Which is extremely likely to be used by Apple.

Also... Motorola's priorities are clear... they are primarily a supplier of chips for mobile devices... that's where the bulk of thier business is now. This announcement only bolsters this position.

[AmigaMac]

""and (2) IBM is designing/producing it. IBM will not turn down a customer... the ball is in Apple's court. Who really things they would choose to stay with G4's?""

Especially since IBM wants to make a profit out of that new chip plant they opened up a month ago! It looks like IBM wants to be agressive in the microprocessor arena again!

[ddtlm]

chubakka:

Perhaps you are confused, because the Power4-lite is the new IBM PowerPC chip, as far as I know. We know the Power4 exists and performs well, but is far to large, expensive and hot to ever work in a desktop machine. The Power4-lite, however, was declared by IBM to be designed for desktops. That sounds a lot like it will be suitable for Apple to use some day.

Apple needs the Power4-lite, and IBM would love customers.

[chubakka]

IBM makes powerpc chips for thier own use too... I'm not disputing that it's a Power PC chip IBM is calling it that... but not all powerpc chips are used by apple... however, I think that the new chip will be ... but we don't KNOW that.

[ddtlm]

chubakka:

Of course we don't know it, nor have I claimed that we do. However I think that there is enough evidence to conclude that Apple will most likely be using a Power4-lite some day.

[chubakka]

I agree... sheeesh... I said that in both post... it's etremely likely to be used by apple.

But we don't know jack about the motorola chip other than that thier part of a consortium that's going to design and manufacture it.

[ddtlm]

Oh I see now... BTW, Moto has not announced any chips here, just a process tech. And actually they didn't even do that, they just said they were gona work with some other companies on it.

A few posts back when you said "these chips" I thought you were referring to the Power4-lite as I was, since there is no Moto chip to speak of. Of course what you were saying is very confusing when looked at from my point of view, but I think it's all cleared up now.

Glad to see we agree happily ever after.

[DavPeanut]

I don't understand why heat even matters (to a point). You can get heatsinks that are so cold, they make water freeze on contact. Just so you know, water takes about an two hours for an icecube to freeze a 0 degrees (the normal freezer tempurature)

[jelloshotsrule]

you gonna put a big heatsink in a laptop???

[ddtlm]

DavPeanut:

Hmm, you seem to have a serious flaw in your thinking. Heat sinks do not do anything but conduct heat from the hot side to the cool side.

In order to cool below air temperature, we need refrigeration or a peltier cooler, but in both cases more heat is actually generated than cold. The big deal is that the heat is outside and the cold is inside... think of a "heat pump." To prove this to yourself, try cooling your house by leaving your fridge open 24x7. :) Or try to cool your house with the A/C unit's cooling fins inside. :)

[biotech]

This announcement sounds all very nice. I just checked, and it was in April 2001 that Mot announced their 0.13 micron process, about a year and a half ago. As far as I am aware, the latest 7455 chips used in the powermacs are still 0.18. The mythical 7470 we were all waiting for was supposed to be fabricated at 0.13 still havn't arrived. Are any of the chips used in macs fabbed at 0.13? Are ANY Mot chips at 0.13? At this rate, I am not holding my breath for 0.09 G4 any time soon.

[ddtlm]

biotech:

Didn't realize Moto had announced it so long ago... yeah I'm getting annoyed at this low-tech 180nm crap, time to get to 130nm before Intel makes it to 90nm!

I would not be especially surprised if the 1.25's were 130nm chips, but I am betting they are not. 130nm could give Apple headroom for a while.

[Johnny7896]

Moto or IBM
IBM or Moto

I hope it's IBM. Using Moto for low end would also work.
I hope it's not a Cyrix based X86 chip from IBM (Ha Ha).

Apple may move to IBM in a year or two.
Moto may produce a .09 G5 this year.
Who knows?
Steve knows.

Interesting how the article said the chips would be ready by the end of this year. The G4 has been milked way past it's life cycle. It's old technology. Buttom line....we need something new. Soon. I'm holding old for the new chip whatever it is. Not an overclocked G4. Duals are nice but you gain more from an upgraded processor.

[blakespot]

Originally posted by AmigaMac
Well with Amiga's reentry into the hardware space and adopting IBM's offerings should help push things along in the PowerPC front. Of course Nintendo's presence is also helpful. So I think things can only get better!
I wouldn't hold my breath on Amiga hardware driving PowerPC production, AmigaMac...


blakespot

[ddtlm]

Johnny7896:

It annoys me to no end that mud-slingers are calling the 1.25's overclocked in the absence of any evidence.

[BRussell]

Originally posted by biotech
Are any of the chips used in macs fabbed at 0.13?The chip in the iBook.

Made by guess who. IBM.

[solvs]

Originally posted by ddtlm

It annoys me to no end that mud-slingers are calling the 1.25's overclocked in the absence of any evidence.

Thank you.

I was just about to say that. For all we know it could be a 1.25 GHz 7470, or 7460, or 7500. Even if it is a 745?, that doesn't mean it's an OC'ed 1 GHz chip, anymore than a 1 GHZ is an OC'ed 933 or 867. A chip is rated at what it's rated. Maybe what people mean is that they don't want another 7455, but that's not how it's coming out. And as ARN pointed out, there is no official 7470. Yet.

But if the 1.25s are OC'ed, then there's gonna be some trouble.

Literally. Do you know how bad you can damage a chip like the G4 if you OC it? Even with the cooling. It's been done, but you're gonna have a lot of problems and failures. I don't think Apple would do that, desperate as they are. Do you?

I don't.

Besides, there are plenty of other things to complain about.

[thedude]

From what I understand, (for right or wrong) the 1.25 chips are OC'ed 1ghz chips. This brings up an interesting thought. The timing of all this development on IBM's part is almost too perfect. Firstly, Mot doesn't deliver on the apollo for macworld, then doesn't deliver for the new PM announcement and Apple is forced to OC their current chips. Meanwhile, IBM announces a Power4 offshoot. Hmmm....It looks like Mot doesn't want to spend money and resources developing for Apple if they're just going to bail and go with the Power4. I sure hope that this chip is everything that people think it is. Many mac faithful that I know are starting to grumble about performance, and are adding Wintel machines to their collection.

Show me the Hardware!

[BRussell]

Originally posted by thedude
From what I understand, (for right or wrong) the 1.25 chips are OC'ed 1ghz chips. This brings up an interesting thought. The timing of all this development on IBM's part is almost too perfect. Firstly, Mot doesn't deliver on the apollo for macworld, then doesn't deliver for the new PM announcement and Apple is forced to OC their current chips.Well, the Apollo was delivered in the dual 1Ghz chips in January 2002, so I'm not sure what to make of the rest of your post. There is no way Apple is overclocking chips. Companies get sued and lose for that sort of thing.

[Kethoticus]

I can tell ya how: It won't.

When Apple has to deliver overclocked chips as upgrades to the line, something is seriously wrong.

Motorola sucks. They're hearts are not into it. Or maybe it's management crap I've read about. Maybe it's departmental egos and other personality nonsense getting in the way. I don't know. Perhaps they're miffed at Apple for canning cloning. Who knows.

But what I am certain of is that Apple's future does not include Motorola, that is unless Apple plans on delivering machines that will eventually be 5-10x slower than its Wintel competition.

[djmcave]

Why would IBM bother to put Acti-vec like capacities on its chips...
Only Apple uses it.....

DJMCAVE

[Wry Cooter]

Originally posted by ddtlm
Timothy:

Extremism appears to be part of the whole Apple-user thing. First, people (not incluing me) run off and believe unrealistic rumors about dual 1.6's and DDR FSB's.



You don't have to be an extremist to notice Motorolas long, long history of too little too late, merely observant.

Like that two by four that just hit you in the head.

[barkmonster]

On the subject of overclocking, Apple's recent Powermac updates have been disappointing and predictable because of overclocking attempts on the previous range.

First the 733 Mhz G4 was OC'd to 867 Mhz, that was the next speed bump from Apple.

Then the 867 Mhz G4 was OC'd to 1 Ghz and Apple brought out the dual Ghz when the next range came out.

Earlier this year someone overclocked a 1 Ghz G4 to 1.2 Ghz, now Apple have presumably got enough chips that work at 1.25 Ghz because less of them are failing at that multiplier setting.

These are speed increases we've seen since the PowerPC 7450 chips appeared in powermacs :

867 Mhz (133Mhz x 6.5)
1 Ghz (133Mhz x 7.5)
1.2 Ghz (133Mhz x 9) - Overclocking attempt
1.25 Ghz (167Mhz x 7.5)

If there really is a predictable pattern here, maybe this will be the next powermac early next year :

1.5 Ghz (167Mhz x 9)

maybe even

1.75 Ghz (167 x 10.5)

if we're lucky.

I'm not suggesting that Apple overclock at all, infact like all companies, they effectively underclock, If they have a bunch of 1Ghz chips that fail at their rated speed they'll just lower the clock speed till they run stable and they use them in lower range models while saving money on the cost of the chips. That's why overclocking is possible in the first place, the odd chip will slip through that works at a higher than rated clock speed and someone will be able to overclock it with or without elaborate cooling depending on how brave they are.

A clockspeed increase of 133 - 250 Mhz every revision isn't going to be much use as far as being a worthy upgrade, If you're already using a 1.25 Ghz G4, why bother with a 1.5 Ghz chip anyway ? At best it's only going to be 20% faster, at least when they jumped from 500 - 733 Mhz it was a theoretical 46% speed increase. The faster a chip gets, the slower the progress of it's potential speed increases become. Look at the Pentium 4, if you'd bought a 2.2 Ghz P4 in January the new 2.8 Ghz P4 might not be too appealing for the 27% extra speed you might get out of it.

The dual cpu G4s seem a really sweet deal for anyone who bought the 800Mhz entry level model earlier in the year but generally speaking the mac is heading down the same road as the PC as far as clock speeds go, it will hit a point where no one will want to upgrade their Mac or PC for a whole year because what they already have is fast enough and the new chips arn't going to justify the upgrade cost. That's why motorola's advancements in fabrication could be a positive thing, they could ramp up the clock speed quite conciderably.

[blakespot]

Originally posted by djmcave
Why would IBM bother to put Acti-vec like capacities on its chips...
Only Apple uses it.....

DJMCAVE
They already have. Their Power4-derived desktop CPU's (that many feel will be Apple's salvation) have a set of vector extensions almost identical to AltiVec.



blakespot

[Telomar]

Originally posted by ddtlm
We have to look at the two most important facts: (1) Apple needs a chip like the Power4-lite, and (2) IBM is designing/producing it. IBM will not turn down a customer... the ball is in Apple's court. Who really things they would choose to stay with G4's?

The point has been made before and needs to be made again. Even if IBM's chip is destined for Apple desktops it isn't likely for the mobile market.

There is really no reason to suspect Motorola won't introduce an improved architecture with the new process and a replacement of the MPX bus with an on-chip memory controller and RIO. Really that and the addition of a second FPU and the G4 is a very nice chip especially in the low end and mobile markets.

This doesn't have to be an IBM or Motorola decision and I would be very very surprised if Apple was thinking of it that way.

[Pauls]

Can there be some sort of definition here?.. multiplying a particular bus speed to reach a particular cpu clock speed is not overclocking, just multiplying. Overclocking is speeding up the chip's (in this case cpu) clock over what it has been reliably tested at. The domain of overclockers are home enthusiasts, not chip and PC manufacturers. Chip manufacturers invariably rate a model below what they can push it with the best quality chips capable of much better ratings.

It would seem to me that the price and delayed availability of the 1.25Ghz G4 are the best of the crop of the 7450 and so can be run reliably at a better rating but are a much lower yield. Apple is trying to milk the last of the .18 micron chips, perhaps to clear inventory, prove the viability of their buying power to make the desktop powerpc worth developing.

Increasing the number of cpu's per model is a good way for apple to sell more overall cpu's and possibly increase their generational lifespan.

[BongHits]

Originally posted by blakespot

They already have. Their Power4-derived desktop CPU's (that many feel will be Apple's salvation) have a set of vector extensions almost identical to AltiVec.



blakespot
dude...haven't we all known this for a month or more??? im pretty sure the post your replying to SAID exactly what u said.. What's the name of the vector processing unit btw?

[solvs]

Originally posted by Pauls
Can there be some sort of definition here?.. multiplying a particular bus speed to reach a particular cpu clock speed is not overclocking, just multiplying. Overclocking is speeding up the chip's (in this case cpu) clock over what it has been reliably tested at. The domain of overclockers are home enthusiasts, not chip and PC manufacturers. Chip manufacturers invariably rate a model below what they can push it with the best quality chips capable of much better ratings.


Again, thank you.

Overclocking and upclocking (or multiplying, or what ever you want to call it) are 2 completely different things.

They may have taken the 1,000 MHz with the 133.33 FSB (7.5 multipier) and clocked it at 1250 MHz with a 166.67 (also 7.5 multipier), but that doesn't make it OC'ed. It's rated to run at what it's rated to run. The G4 had been rated to run a 166.67 FSB for awhile.

For all we know the 1 GHz chip was underclocked because Apple didn't have a 166.67 FSB yet.

Though, that's not very likely.

[BongHits]

Originally posted by Telomar


The point has been made before and needs to be made again. Even if IBM's chip is destined for Apple desktops it isn't likely for the mobile market.

There is really no reason to suspect Motorola won't introduce an improved architecture with the new process and a replacement of the MPX bus with an on-chip memory controller and RIO. Really that and the addition of a second FPU and the G4 is a very nice chip especially in the low end and mobile markets.

This doesn't have to be an IBM or Motorola decision and I would be very very surprised if Apple was thinking of it that way.
doesn't ibm already make a fantastic mobile processor (the 750fx or whatever)?? I have heard they're developing basically the same thing with their version of Altivec slapped on their. These chips (from what i've heard on the boards) can run a bus speed of 200 mhz and can be overclocked via software (apple's way of making it the consumer model)...and can run mhz for mhz with a G4 (not the case with previous g3's) so my question is...what do we need from motorala with the intro of a power4 lite?

[skunk]

Well, I suppose it would be nice for a change to watch them go down the spout without feeling they're taking us all with them....:)

[Telomar]

Originally posted by BongHits

doesn't ibm already make a fantastic mobile processor (the 750fx or whatever)?? I have heard they're developing basically the same thing with their version of Altivec slapped on their. These chips (from what i've heard on the boards) can run a bus speed of 200 mhz and can be overclocked via software (apple's way of making it the consumer model)...and can run mhz for mhz with a G4 (not the case with previous g3's) so my question is...what do we need from motorala with the intro of a power4 lite?

Actually the G3 can't keep up with the G4 except by being on a fabrication technology a generation ahead of the G4's. The G4 core is actually superior to the G3 core though given the generational differences. It's successor also won't appear until late next year at the earliest. Although you could easily fit that into a roadmap and just use current G4s until then.

Without a lot more details at the executive level it's impossible to say for certain which would be the best option and that's the point, this would be an executive decision not a technical one. Unless Apple is going to pick up some worthy benefit by switching entirely to IBM chips they are better off diversifying their supply and keeping the G4 evolving. They can drop Motorola later if they want since IBM is planning to make their chips anyway.

[Luckster]

Well, for all you IBM fans out there, they are already building a fab to make sub 0.10 micron chips... and it will be online before the end of this year(http://www.eetimes.com/semi/news/OEG20001011S0021 (http://) ). This marks a serious commitment to chip development for IBM. Thus, if the IBM chips do again make their way into Apple machines, the preformance should be excellent (desktop or laptop). Also, not that while IBM makes one of the worlds most powerful chips (the Power4) which is one of the largest (680 million transistors) and most power hungry (500 watts), the also make some of the worlds best power conserving chips (G3). So, I would not write off an IBM-based Apple laptop. Furthermore, Wintel companies are using P4 desktop chips (albeit not entirely reliable as a laptop preformer) to boost their laptop preformace. This seriously reduces battery life, but it does provide extra power. The P4, for reference uses around twice the power of a G4. There is therefore no indication that if a chip came out that was plentiful enough and could be squeezed into a laptop that Apple would not do so.
Of note though, is that Motorola has been producing chip for Apple from the beginning, so this would mark a serious departure from the norm and is not necessarily likely. I would expect to see moderate preformance increases from Motorola, becuase their primary target is network equipment and the demand for such has fallen lately.

[nixd2001]

Originally posted by Pauls
Can there be some sort of definition here?.. multiplying a particular bus speed to reach a particular cpu clock speed is not overclocking, just multiplying. Overclocking is speeding up the chip's (in this case cpu) clock over what it has been reliably tested at. The domain of overclockers are home enthusiasts, not chip and PC manufacturers. Chip manufacturers invariably rate a model below what they can push it with the best quality chips capable of much better ratings.

I think this distinction is well worth drawing. If Apple buy a G4 chip from Mot rated at 1GHz, then Mot are saying they warrant this will work correctly at that clock rate within their defined temperature range (typically up to 70 C). If Apple then push that chip to 1.25 or 1.5GHz and it fails (either totally or erratically), Mot will not care in the slightest because they never claimed it could do it. This would leave Apple with "overclocked" chips performing unreliably and with no comeback from their supplier, but with an obligation to their customer to make it work. Apple would then gain a tremendous badly reputation - "fast and crash" or something like that. Apple simply wouldn't do this.

If Apple ship a chip running at 1.25GHz, then it's because Mot have supplied it on the basis it will run reliably at that clock rate within its temperature range. Anything else is [B]commercial suicide[B].

[But to contradict myself slightly: if Apple used one of those wonderful "fridge coolers", and hence could be certain the chip never got abov, say 30C, Mot might well warrant a higher clock rate - the back of chip data sheets sometimes has a nice frequency vs temp chart showing what is warranted and what is not. But given the new DPPM only has heat sinks and fans, there isn't (IMHO) the certainty of keeping the temp down enough to know they will work.]