IBM's POWER Event

[Mr Maui]

He stated profit margins, not gross margins. Apple has lost money on sales, minus expenses, for about half of the last 12 fiscal quarters. The company made a profit in some of those quarters off of interest and investments.

Lost money on what exactly? And you're agreeing that their overall net profits are 2%? Please offer some numbers to support these statements. Apple has $6 billion in cash in their portfolio. I find it very hard to believe you get this from 2% net margins. That equates to 300 billion in sales?

 

[thogs_cave]

"POWER5 will be the "brain" of a new line of powerful computer systems that will be introduced in 2004."

Hmm, one wonders what new line this will be...cough, Apple.

Not the POWER5. I've heard some noises that will be in a new desktop workstation with killer specs and a very surprising price.

I can't believe I find my self rooting for IBM - in 1984, I *hated* them. Does this mean I'll cheer for Microsoft someday? No, wait - I *used* to like them, so it's O.K....

 

[joeconvert]

PS3 Indeed

more likely it is for some other device (computer) sony, toshiba, and IBM are developing a new type of process called "cell" where the processors can be linked to others in home electronics, other PS3s, etc, to draw more power. only xBox and Nintendo are using the PPC for their systems so far.

hope that clears things up,
technocoy

yes, it is. Anyone who follows console development knows this. Sony has already previously discussed it.

 

[squatch]

The 970fx should easily reach 3 Ghz. The die shrink and IBM adding up to a 20-30% boost in speed with strained silicon gives the 970fx the potential to reach speeds of at least 3 GHz.

Just because IBM does not mention speeds beyond 2.5 GHz for the 970fx does not mean that Apple will not get 3 GHz 970fx chips. Apple may have exclusive rights to use higher chip speeds than what IBM officially lists.

True, but I was staying on the conservative side of things. Just looking at the IBM pdf and the typical power consumption rate of a 2.0 Ghz 970fx at 24.5W and a current 1.8 Ghz 970 at 51W, theoretically a single 2.5 Ghz 970fx would be around 50W. I can't imagine them going much beyond 2.7 Ghz safely without encountering overheating problems.

But hey, Apple is the overclocking experts (especially when it came to Motorola chips), so I wouldn't be surprised!

 

[DeepIn2U]

look closely

I think most of us are missing something important here. Check it, "licensee". I guess with this, from what I gather, IBM is licensing the technology and the RISC & other technology to other vendors (so to speak) and manufacturers (I'm hoping) to produce PowerPC based products/or Power-based products.

Remember the consortium of Apple, IBM, and motorola (small case name done on purpose for lack of ambition!!) that formed the PowerPC technology??
If such a contract does exist can anyone find out when it expires for a particular (motorola), or all parties??

 

[Phinius]

True, but I was staying on the conservative side of things. Just looking at the IBM pdf and the typical power consumption rate of a 2.0 Ghz 970fx at 24.5W and a current 1.8 Ghz 970 at 51W, theoretically a single 2.5 Ghz 970fx would be around 50W. I can't imagine them going much beyond 2.7 Ghz safely without encountering overheating problems.

The 970fx at 3 Ghz should not use much more watts than the 2 Ghz 970 that Apple now uses. IBM has implemented some power saving features on the 970fx.

The upcoming 9XX IBM PowerPC processor, that will likely be based largely on the Power5, will raise the power use per Ghz over the 970fx. That's due to a likely larger L2 cache and a feature similiar to Intel's Hyperthreading causing the chip to utilize on average more of the chips resources at any one time.

But don't expect much higher frequency rate on the next 9XX chip after the 970fx, IBM focused on using the available resources more efficiently rather boosting the clock rate.

 

[gbojim]

Lost money on what exactly? And you're agreeing that their overall net profits are 2%? Please offer some numbers to support these statements. Apple has $6 billion in cash in their portfolio. I find it very hard to believe you get this from 2% net margins. That equates to 300 billion in sales?

You have to use the numbers correctly if you're going to argue about this. The only info that matters is what is in Apple's 10k filing - you can get a copy in the investor relations page on their site. Look for the Consolidated Statement of Operations.

If you want to know if Apple made money on product, look at gross margin which is sales less cost of sales (not overall cost). In FY2003, that is 27% gross margin from $6.207B in sales less $4.499B cost.

If you include operating costs which is primarily overhead (R&D, admin, etc.) and subtract that from sales as well, you end up with operating income. This does not include other income or expenses such as taxes, investment interest, etc.

But including operating costs can be misleading. In FY2003, Apple suffered an operating loss of $1M. However, that included a one time restructuring charge of $26M. Without that charge, Apple would have had a $25M operating income (profit). It is usually best to ignore one time charges to make a fair assessment.

Also, investment gains contribute to the company as do operating costs. Many organizations use investment gains to gauge the amount of R&D that can be funded.

Bottom line is in FY2003 Apple made $69M net income on $6.207B in sales which translates to 1.1% net profit.

 

[gekko513]

2:1 Dual: 2.0, 2.2, 2.4 GHz
3:1 Dual: 1.8, 2.1, 2.4 GHz

I'm not sure about 3:1 processors at 1.8 and 2.1 GHz becuase that would mean a system bus of 600 and 700 MHz. Doesn't the PowerMac use dual memory controllers which gives memory throughput of DDR400 x 2 = 800MHz? Maybe this isn't as significant as I think, but if it works the way it looks, this would make the bus a bottleneck.

 

[Analog Kid]

I would not be surprised to see Apple using a 3 GHz 970FX, with a 1.5 GHz bus, in the next few months. The effective bus speed would actually be less than 1.5 GHz due to its inherent design, so Apple could potentially keep that bus speed busy by using 667 MHz DDR2 memory that is just recently went into production. Because this memory is now being produced in very low volumes, (due to the X86 chip manufacturers not yet using it) there will be a considerable price premium if Apple does go ahead and use it. Considering the price of the topend PowerMac computers, the additional costs of 667 MHz DDR2 memory would not add substantially to the overall price of the computer. Although the costs for adding memory will come as a sticker shock to many potential purchasers.

I'm expecting to see DDR2 in the next Powerbooks and probably the next desktops. The power savings seems to play into Apple's strategy quite well...

Yeah, there will probably be a bit of sticker shock at first but that will be short term. The main reason for the higher prices is volumes. Apple has done this before with less central technologies-- AirPort for example.

DRAM manufacturers are pushing this pretty hard-- this is where they want to go. Apple tends to move most quickly on trends that seem inevitable. PC makers wait until the market is established. To some extent, putting the technology into a Mac helps to establish the market.

 

[Analog Kid]

Someone mentioned that if more of the G5's end up getting out there because of Microsoft or Sony or whoever, the price will go down. This probably true. They were however skeptical about Apple passing on the savings. Since Apple's profit margins are 2% he does have a valid point, they certainly could use the money. (Compare this to Microsoft's 25%. Ouch.) However, since people do take price into account when buying a computer, if Apple can keep good margins and bring the cost down, I doubt they will hesitate. They are at roughly 5% of the market, anything that can work in their favor I expect they would consider.

Lower prices on the processor means tens of dollars. This isn't going to make an appreciable difference to either the market price of the machine or to the margins Apple sees.

I'd guess the main cost difference in an Apple computer is in their custom system chips and their enclosures. Neither can benefit from wider adoption because they can't be more widely adopted...

 

[Analog Kid]

I think most of us are missing something important here. Check it, "licensee". I guess with this, from what I gather, IBM is licensing the technology and the RISC & other technology to other vendors (so to speak) and manufacturers (I'm hoping) to produce PowerPC based products/or Power-based products.

From what I read in IBM's press release this is their key point. They're opening the architecture and trying to push it into the mainstream. They're trying to address everything from the embedded market to high end servers with the same instruction set and companies that use it will have a "scalable" option that shows more breadth than Intel or ARM.

Mostly it amounts to a public statement that they're not dumping their own CPUs for Itanium or Opteron. A policy statement to quell possible confusion from they tendency to keep a finger in every pie.

There's very little in this event for us-- other than the fact that IBM is standing firmly behind POWER and the architecture isn't going away anytime soon... That in itself is good news, but doesn't mean anything for Apple users other than that they won't be switching to Intel/AMD.

The reconfigurable processing blurb might be an oblique reference to Cell or their Xilinx partnerships, but most likely it was just a blue sky reference from R&D. This was the "we're still working on cool stuff" plug to wrap up a talk about how advanced their current line is.

All-in-all a positive event for IBM. They've got an alternative architecture that is seeing broad acceptance, and they're giving Intel a run for their money in all markets from Itanium to XScale. Not much in there for us though...

 

[AngryAngel]

For the person that asked: no, the PPC 400 Series in not the G3- which is the 700 series (750, etc.). There is not a 700 series 'system on a chip' (yet?). If what you read was true about Sony buying a licence for a 'system on a chip', it makes sense that it will go into a PDA, phone or other consumer device.

An IBM person has stated that the 970 is capable of several bus multipliers:

"Peter Sandon: The processor design itself supports several ratios. The one that Apple announced was a 2:1 ratio. And the processor supports at least 3, 4 and 6 as other ratios."

From <a href="http://www.arstechnica.com/cpu/03q2/ppc970-interview/ppc970-interview-2.html">Ars Technica</a>

Apple will not license their OS ever again- so Sony will not be making Mac-compatibles.

 

[QuiteSure]

If both Sony and MS are using an IBM chip in their new gaming consoles, can Nintendo be far behind?

And if so, can we look forward to a new Zelda game??

THAT would be exciting!

 

[TMay]

perhaps Apple will eat its own dogfood

For the person that asked: no, the PPC 400 Series in not the G3- which is the 700 series (750, etc.). There is not a 700 series 'system on a chip' (yet?). If what you read was true about Sony buying a licence for a 'system on a chip', it makes sense that it will go into a PDA, phone or other consumer device.

I've often wondered why Apple hasn't taken the 400 series and run with it for Xcode compatible consumer products (OSX Lite). I think that Developers would eat this up, but maybe its not that important in basically closed devices.

 

[Analog Kid]

I've often wondered why Apple hasn't taken the 400 series and run with it for Xcode compatible consumer products (OSX Lite). I think that Developers would eat this up, but maybe its not that important in basically closed devices.

Making OS X available for consumer products is just begging for someone to make a very large desktop consumer product... Not good.

They could do this for Darwin though. Linux has been finding its way into consumer products, no reason another open source OS couldn't. It would mean a change of focus though-- Apple still sees itself as a hardware company even if some of its greatest assets are in software...

To address the second part of your question: embedded developers consider development tools to be a critical decision criteria. They expect support with the tools though, so Apple would have to set up a support infrastructure for embedded development...

 

[szark]

If both Sony and MS are using an IBM chip in their new gaming consoles, can Nintendo be far behind?

Actually, they're very far ahead as the current GameCube uses a PowerPC processor.

 

[Phinius]

Upcoming G4 chip compared to 970fx

Motorola recently stated in a internal document that the company plans to double the frequency of the G4 about every 18 months. The 7447 G4 was introduced late in January of 2003 with a maximum listed frequency of 1 GHz. That puts a doubling of frequency to 2 GHz, and it should arrive by the end of July 2004, if the plan stays on track.

Motorola has also mentioned plans to make a dual-core G4 at .10-microns, that will have a onboard controller and DDR-2 capability. That chip was evidently given the ok for production last year. A Motorola document mentioned that it would average 25 watts of power use running at 1.5 GHz. Compare that to IBM's stated 50 watts power use for a 2.5 GHz 970fx processor. It could well be that this upcoming dual-core G4 running at 2 GHz might use about 35 watts. That should put it's performance for small enclosures well above what the 970fx can manage.

I would not be surprised to see Apple using this dual-core G4 in all iBooks AND possibly the PowerBooks, along with the iMac and eMac. The advantages are a much higher performance per power use compared to the 970fx and it could possibly be cheaper for Apple to impliment in a notebook computer. It would not need a more expensive dual-channel memory board as is used in the G5 Power Macs and it's on board controller would also reduce costs, while uping performance. Perhaps there could also be a new xServe produced with two of these chips, which would be four processors in a 1U server. Although the desire of Mac users for a G5 PowerBook, and Apple wanting to differentiate the product lines, would more point toward the possbility of a G5 PowerBook, rather than a dual-processor G4 PowerBook.

Apple has stated that a dual-2 Ghz G5 PowerMac performs about 60% faster than a dual-1.42 GHz G4. Steve Jobs also stated IBM promised a 3 GHz 970 in a years time, which would effectively be a 50% frequency boost over the maximum 2 GHz available now. A 2 GHz dual-G4 would be about a 41% jump in frequency over the 1.42 GHz G4. Considering that this dual-G4 will have a on-board memory controller, with much faster bus and memory, it should narrow the G4 performance gap with the 970. In a small enclosed space it should far exceed what's possible in speed with the 970.

 

[Analog Kid]

Motorola recently stated in a internal document that the company plans to double the frequency of the G4 about every 18 months. The 7447 G4 was introduced late in January of 2003 with a maximum listed frequency of 1 GHz. That puts a doubling of frequency to 2 GHz, and it should arrive by the end of July 2004, if the plan stays on track.

Motorola has also mentioned plans to make a dual-core G4 at .10-microns, that will have a onboard controller and DDR-2 capability. That chip was evidently given the ok for production last year. A Motorola document mentioned that it would average 25 watts of power use running at 1.5 GHz. Compare that to IBM's stated 50 watts power use for a 2.5 GHz 970fx processor. It could well be that this upcoming dual-core G4 running at 2 GHz might use about 35 watts. That should put it's performance for small enclosures well above what the 970fx can manage.

I would not be surprised to see Apple using this dual-core G4 in all iBooks AND possibly the PowerBooks, along with the iMac and eMac. The advantages are a much higher performance per power use compared to the 970fx and it could possibly be cheaper for Apple to impliment in a notebook computer. It would not need a more expensive dual-channel memory board as is used in the G5 Power Macs and it's on board controller would also reduce costs, while uping performance. Perhaps there could also be a new xServe produced with two of these chips, which would be four processors in a 1U server. Although the desire of Mac users for a G5 PowerBook, and Apple wanting to differentiate the product lines, would more point toward the possbility of a G5 PowerBook, rather than a dual-processor G4 PowerBook.

Apple has stated that a dual-2 Ghz G5 PowerMac performs about 60% faster than a dual-1.42 GHz G4. Steve Jobs also stated IBM promised a 3 GHz 970 in a years time, which would effectively be a 50% frequency boost over the maximum 2 GHz available now. A 2 GHz dual-G4 would be about a 41% jump in frequency over the 1.42 GHz G4. Considering that this dual-G4 will have a on-board memory controller, with much faster bus and memory, it should narrow the G4 performance gap with the 970. In a small enclosed space it should far exceed what's possible in speed with the 970.

I guess this is a bit off topic, but since the thread seems to have died otherwise...

I'd like to see this dual-core G4 come out if it exists-- I think it would give Apple a lot more flexibility in their products and would make a great laptop processor.

Then there's the whole Motorola wildcard... Doesn't the PPC fall under the Freescale spin off?

The only mention I've seen of the dual core G4 was a few mentions on The Register that never seemed to amount to anything. Have you seen it mentioned elsewhere? Has Mot said anything about it officially? I'd love to find more details...

Twice the CPUs, twice the Altivec, half the power... It sounds nice. I've got mixed feelings about Mot as a supplier though and it's hard to tell if this would amount to the laptop of the future or a sand trap.

I am worried about Apple trying to shoehorn a G5 into a Powerbook. I'm not convinced it's the right thing to do but without a chip such as the one you're describing I don't know if they have a choice.

Judging from the posts I've seen in these forums, there's at least a vocal minority that seems to want a G5PB even if that means performance at or below what a G4 could offer and reduced battery life. It's not usually put that way, but when looking at the compromises people are suggesting it is what it amounts to... I'm curious to see how Apple works this-- do they give people what they think they want, or what they need.

 

[Phinius]

Moto and IBM chip updates...

Doesn't the PPC fall under the Freescale spin off?

The only mention I've seen of the dual core G4 was a few mentions on The Register that never seemed to amount to anything. Have you seen it mentioned elsewhere? Has Mot said anything about it officially? I'd love to find more details..

Motorola made a presentation that mentions the dual-core G4 at their Smart Networks Developer Forum held in June 2003 at Disneyland Paris (Motorola also posted PDF file on their website from the presention listing the dual-core G4, which the company has subsequently removed from their website. Perhaps partially because of an Apple objection to it being displayed).

Here's the link (unfortunately registration is required to view the article):

http://www.pcpro.co.uk/news/news_story.php?id=42928

Notice that in the presentation Motorola mentions 'process and architecture improvements' that will bring the PowerPC above 3GHz. That would likely come in the next chip model right after the dual-core since another process would only move the existing G4 architecture from 2 GHz to about 2.6 GHz. A jump to 3 GHz is simply to great a leap from a process shrink without some significant chip architecture improvements. So it seems that speed improvements to the current G4 architecture will come to an end when the dual-core comes out (again probably around July of this year).

Concerning the use of a G5 in the Powerbook...

Apple could use the 970fx in a PowerBook up to a maximum of about 2 GHz, since it uses about 25 watts according to IBM. However, the next version of 9XX chip ,which will be based on the Power5, will almost certainly use more watts than the 970fx at the same frequency. That's due to IBM adding simultaneous multitasking, which is similar to Intel's Hyperthreading. That leaves Apple with using the 970fx in the PowerBook in the next few months and waiting about a year after that for a process shrink to arrive in order to increase the speed of the PowerBooks, or else IBM is working on a PowerPC chip that would suit Apple's notebook use more ideally. Another PowerPC chip in the works by IBM for Apple is purported to be the case according to a recent report from another rumour site.

Motorola is expected to make the Freescale spin off of the chip division in the summer, and yes, the PPC G4 would be part of that. However, that should not stop or delay the dual-core G4 from being produced.

 

[Analog Kid]

Motorola made a presentation that mentions the dual-core G4 at their Smart Networks Developer Forum held in June 2003 at Disneyland Paris (Motorola also posted PDF file on their website from the presention listing the dual-core G4, which the company has subsequently removed from their website. Perhaps partially because of an Apple objection to it being displayed).

Here's the link (unfortunately registration is required to view the article):

http://www.pcpro.co.uk/news/news_story.php?id=42928

Notice that in the presentation Motorola mentions 'process and architecture improvements' that will bring the PowerPC above 3GHz. That would likely come in the next chip model right after the dual-core since another process would only move the existing G4 architecture from 2 GHz to about 2.6 GHz. A jump to 3 GHz is simply to great a leap from a process shrink without some significant chip architecture improvements. So it seems that speed improvements to the current G4 architecture will come to an end when the dual-core comes out (again probably around July of this year).

Concerning the use of a G5 in the Powerbook...

Apple could use the 970fx in a PowerBook up to a maximum of about 2 GHz, since it uses about 25 watts according to IBM. However, the next version of 9XX chip ,which will be based on the Power5, will almost certainly use more watts than the 970fx at the same frequency. That's due to IBM adding simultaneous multitasking, which is similar to Intel's Hyperthreading. That leaves Apple with using the 970fx in the PowerBook in the next few months and waiting about a year after that for a process shrink to arrive in order to increase the speed of the PowerBooks, or else IBM is working on a PowerPC chip that would suit Apple's notebook use more ideally. Another PowerPC chip in the works by IBM for Apple is purported to be the case according to a recent report from another rumour site.

Motorola is expected to make the Freescale spin off of the chip division in the summer, and yes, the PPC G4 would be part of that. However, that should not stop or delay the dual-core G4 from being produced.

Well, hopefully is wasn't taken down because the program was scratched...

Thanks for the link-- a bit short on details but it certainly helps confirm that this chip has been talked about.

Just working from memory here, but I think the 25W is about twice what the current G4s pull. Add the power of the rest of the system to that and I think we're looking at a pretty hot machine...

I'd really like to see the portable and desktop lines run on different architectures optimized for their respective purposes, and a new G4 might be just the ticket for that.

 

[Phinius]

Dual-processor would use quite a bit more power

Just working from memory here, but I think the 25W is about twice what the current G4s pull. Add the power of the rest of the system to that and I think we're looking at a pretty hot machine...

I'd really like to see the portable and desktop lines run on different architectures optimized for their respective purposes, and a new G4 might be just the ticket for that.

Your right, the 1.33 GHz 7447 G4 that is used in the PowerBook burns an average of 18.3 watts. Which means that the upcoming dual-processor G4, using 25 watts, would be about 37% more than the highest average use of any PowerBook now. It would be more impressive for potential buyers to use a single G4 at 2 GHz in a PowerBook or iBook and leave the dual-processors for desktop computers. That way the portables would have a high GHz number and yet a low power use.

Unless IBM comes up with a processor design geared toward notebook use, it looks like the embedded processors, such as the G4, will be much more practical for portable use than the G5 series. Yet since the demand for a G5 Powerbook is overwhelming it's very likely that Apple will put them in the PowerBooks very shortly and leave the G4s for the iBooks. That could limit the topend frequency that the iBooks would use even though they would have a different processor design. Having a 1.5 GHz 970 in the topend PowerBooks and a 2 GHz G4 in the iBooks would probably not be something Apple would try.

 

[aswitcher]

SNIP

Yet since the demand for a G5 Powerbook is overwhelming it's very likely that Apple will put them in the PowerBooks very shortly and leave the G4s for the iBooks. That could limit the topend frequency that the iBooks would use even though they would have a different processor design. Having a 1.5 GHz 970 in the topend PowerBooks and a 2 GHz G4 in the iBooks would probably not be something Apple would try.

I tend to agree...