7457 RM canceled by Motorola, 970 on track - Page 3

I interpreted the "desktop" as referring to Apple desktops.

I think IF this is the plan IBM & Apple have, it can make sense:

IBM gets the back office (where they where always likely to win anyway)

Apple gets the front office.

IBM wins no matter what because they're not making any money as it is on desktops, so if they sell tons of extra chips to Apple as a result of a partnership, hey big bonus ...

Apple wins because IBM becomes their trojan horse into areas they would never have otherwise gotten into, and IBM might finally start taking WebObjects and Objective-C (and CORE Foundation) seriously (as they should) ... or, in the worse case, some of their boffo hi-end tools will migrate to Apple.

Either way, it's a great dance because there won't be (much) stepping on toes.

pretty cool huh?

[edit grammar: fricketty-hell]

They could be referring to Mac desktop computers, without saying so. They simply stated the target markets for a 970, and not IBM's intention to use the 970 in all such markets. However, I might not be remembering exactly what IBM said.

Remember, IBM sell workstations too. In fact you could pick up a powerful 200MHz based 604e system off of IBM for only $11,000! Bargain.

The great thing about the 970 etc is that IBM desperately needs them itself. And the future versions.

At last, a microprocessor vendor who wants to beat the competition.

Now that Intel are banging on the door of IBM's turf, it's in IBM's interest to get themselves and PPC out of bed.

Lemon Bon Bon

Yes, but IBM's desperation has been a fleeting thing. Obviously they were't desperate enough to update that 200Mhz 604 machine, or develop the G3 into something competitive in the desktop market, or even make any truly relevant desktop/server/workstation CPU other than the insanely expensive POWER chips.

IBM might just be content to sell the same old 1.8Ghz PPC970 in 3-4 years time. It happened before, it can happen again. How big a market is there really for desktop non-windows IBM towers/workstations??? NOT servers or big iron, to which IBM can always throw power chips rather than PPC's.

I don't think it's that big. Apple almost certainly will buy/sell more desktop PPC than IBM. The seeds of discontent are already planted even if no one wants to see it.

Some other arrangements must be made. Apple ALWAYS has to have contingency plans at the ready, that's what hapens to you when you have less than 3% of the market. A second supplier has to be at the ready. Mebbe that's Moto, mebbe that's Apple acquiring Moto designs and patents and shopping custom PPC designs around, mebbe that's even Apple taking a greater role in contributing to the design of the PPC (in the form of engineers and money) that IBM would fab and also use in some of their products.

It's is not enough that a company is interested now, Apple has to have a way to light a fire under their ass, or make significant contributions if/when development lags.

How so El Bee-Bee?

I know this is off topic but if Apple releases all single processor machines I hope they have an option on each model to add another processor.

... ah, but this state of affairs was before the 'kelp' of the computer age (or Grapefruit, whichever geologically very new lifeform you prefer) came onto the stage: THE BLADE.

schwing! (ai!)

Blades seem to be taking over many a niche, more than a fad, less than a breakthru (well, a little), but definitely more than just another form factor ... where-ever they got the idea from (perhaps dial-up modem racks of the mid 90's?) it seems to fit in perfecty with many things, especially as a bridge point between both IBM's and Apple's plans ...

In short, blades aren't going to go away, they're definitely back office (IBM country) and the competition keeps making them since everybody seems to want them; thus, it behooves (I love that word) ... yes, it behooves IBM to keep making chips to handle that niche ... ok, mebbe 'behooves' is a little strong, but let's just say that IBM has an advantage with the 970 in this ever expanding niche with a chip that is very powerful, low power consumption and runs (relatively) cool ...

The next battle for the back office shall be fought with blades, and IBM plans to make their own from the ground up (with a litte Xeon thrown in for ass covering) ...

And into this fray comes Apple, who needs a chip that pretty much perfectly matches the blade spec ... admittedly, asking an embedded chip company (moto) to cough up energy to make a desktop is a bit of a leap, but asking a behemoth (laviathan, hey, it's your call) to schwinge the design of a BLADE chip, somethng that sits nicely between a heat-honkin' x86 desktop and a coy embedded thing, a tad more in the desktop direction ain't asking much ... which is why Apple shouldn't worry too much about IBM orphaning the 9X0 series, especially considering the POWER5 derivative coming down the pipe.


Anything's possible - David Hume and all that - but we're put on this earth to choose between realistic options, so while it's still somewhat possible that IBM might roll their eyes in a mighty WTF? in a year or so and orphan the 970, considering the blade scenario (and the upcoming POWER5 derivative at 4x the speed of the 970 - apparently) this seems rather unlikely: but most importantly ... you don't get to go on the same ride twice, and the ride that brought us a 200Mhz 604e workstation just doesn't seem to be around anymore.

Summing up: mitigated realism says we shouldn't be too concerned about getting run over by it.

(that's enough sentiments for today)

Does it make sense for Apple to use blades in the new PowerMacs? I could see room for possibly four blades, before heat becomes a big issue. Need more power? Just add blades. I had rejected this idea a while back, but reading your post made me reconsider.

Regarding cooling. A technique not mentions in most discussion on cooling is a continuously variable speed blower. Its speed can be gradually changed to accommodate cooling needs At low speeds it is almost silent, and a slow, gradual increases in speed is less noticeable than sudden jumps in speed. The larger a blower or fan, the less noise it makes for any given airflow. That is because it delivers the same flow of air at lower RPM. Such a blower could be directed right across the blade stall.

NeXT had a setup like this, actually, and frogdesign came up with a blade setup for Apple in the mid-80s.


This is pretty much what the 120mm "leaf blower" in MDD PowerMacs does now.

Unfortunately, it's a poor quality fan, thus yielding the 'leaf blower' nomenclature.

That, and the other three in the box... all low quality...

But that's the whole friggin point.

With the G3, IBM made the chip... but they had _no_ interest in _using_ the chip.

With the 604, IBM made the chip, and it was useful, but it was superceded by the Power3/4's....

But this is a new era. There's been a lot more development on parallel processing/load distribution/etc. The 970 is going into more than one line of IBM's. Sure, some of those lines might not need frequent upgrading. But the _blades_ are a line that are relatively easy to pop new CPU speeds into, so that's at least one line IBM would want up-to-speed. The entire point of slamming maximum computing power in minimum space is DAFT if you're using 3-year-old-tech.

Well, I don't know what IBM's intentions are for the rest of the Intel line, but it seems pretty clear that they're out to squash the Itanic. How would you guys interpret this little snippet:


That was taken from here: IBM plots road ahead with Power5

Now If Itanic dies, that leaves Intel strapped with the old monkey of x86 competing with AMD once again - that can't be good for them. Just how long does Intel expect to be saddled (hindered?) with such an architecture? It's obvious that Itanic was supposed to mark a turning point *away* from x86. With all the uncertainty, it really does look as if the Wintelon world will be stuck with x86 for a very, very, very looooong time. If that's the case then I see PowerPC advancing even further as time progresses. Say what you will, but it appears that the Wintelon camp has painted itself into a corner. Anyone disagree?

--
Ed

Apparently, current machines ship with Papst fans, which I am given to understand are quite high quality. The "silencer" kit Apple ships to owners of older MDDs also uses Papst fans. So they figured that much out.

Nevertheless, Apple has already shipped machines using variable-rate fans. At least the "leaf blower" is usually idle...

I get it, but I'm just throwing the other side out there for consideration. Despite a slow trend towards clusters, and distributed computing, there is no guarantee that IBM won't present difficulties and it would be wise for Apple to have "options" Which they like, right? Mebbe Apple will have to be more closely involved in PPC design in the future. I can see the PPC getting lost between embedded and big iron applications. IBm can always look elsewhere for the desktop, so that is always the CPU line with the greatest susceptibility to stagnation and disinterest, and coincedentally the only CPU Apple should be interested in.

However, I agree that there is enough about the 970 to make us very optimistic about it's ability to subtely alter the computing paradigms of desktop future -- more like to a blade or multi CPU environment with it's ability to pack lotsa chip to chip to memory throughput and lotsa power into a very nice small, efficient package. The CPU itself could evolve two ways, get faster and faster, or get smaller and cooler at the same speeds, and even more suitable to 2, 4, 8... way machines.

It is the best option for Apple for now, but unexpected turns ought to be expected by now, anything can happen.

I guess you're all right to be happy though, present concerns are too great to worry about the future ATM. Let's give big blue a chance.

With his reference to the Itanium, I assumed he meant the one that his engineers had, not the Itanium in general.

IBM still (has to) use the 604e in their PowerPC workstations. The POWER chips run are too expensive, and are server not workstation oriented... RS/6000

All the more reason for IBM to develop the 970.

Barto

Um, why would that matter and what would be the difference? I mean surely they aren't that obsessed with producing a machine that could *off* some particular Itanic box they might be tinkering with.... Nah, I'd have to say they meant it in general, since Itanic is their competition and they know that Intel is betting the house on it.

One thing seems quite certain and that's the notion that x86 is going to be the mainstay in the Wintelon world for *at least* the next 10 years, given the fact that Yamhill (x86) is Intel's answer to AMD if AMD's x86 chips take off and Itanic withers...

I guess what I'm trying to say is that it looks as if it's going to be a "support and compatibility nightmare" for the Wintelon camp. Just consider all the *forks* with regard to processor architecture, Operating System compatibility, developer support, application compatibility, driver support and compatibility ... the list goes on. In short, I just don't see developers supporting multiple drivers and applications software for two (or more) VERY different hardware architectures so they can function under the SAME Operating System (i.e., Windows).

Imagine how many versions of Windows there might need to be... Imagine if Autodesk (AutoCAD) or Adobe et. al. had to bring their wares to the same OS that just happens to be running on completely different hardware. Imagine the confusion for users. Imagine.... You get the picture.

--
Ed

Custom 64bit Intel chip for Apple w/ Altavec. Options.

Me wonders. If the 970 brings Apple back to performance parity...will the Power 5 9xx derivative give Apple the option to go uber Workstation while a .09 970+ takes on and kicks Prescott?

...and a .09 G3 plus 200mhz bus + simd unit to help Moto out the door...?

IBM seems to be on an accelerated cpu schedule, surely wise as yesteryears apathy to PPC would be unwise with Intel eyeing up your Server market... Seems IBM have motivation these days...



Lemon Bon Bon

Your points are completely valid and understandable; unfortunately, I have been using similar arguments since before the 386 came out. The x86 architecture has been panned and declared at the end of its life since the move from 8 to 16-bit. Somehow, the chips just keep getting bigger and faster. Intel has gobs of money, the best microprocessor engineers in the world, and the best fab plants. Their philosophy, "we made x86 twice as fast, but, we need $2B to make a new plant" continues to work for them (much to my chagrin).

As long as there are fab and materials advances to be made, Intel will continue to make the x86 chips the speed demons they have become.

Software (OS) is another issue in that they are bound to Windows; however, Linux (and Solaris may jump in there) does give them an alternative to deploying 64-bit.

As long as IBM maintains its interest (which as Matsu points out is not the gamble you want to bet your savings on), then, we can give Intel a run for its money. If Intel says to IBM, hey, we have a Pentium IV chip that is faster than your 970 and we'll sell it to you at 25% of the 970 cost, IBM would most likely allow the 970 to drop into maintenance mode ala the G3. People love to pan Motorola on the G4, but, what has IBM done with the G3? Sahara? Big deal. The G3 has gone through fewer advancements than the G4 in the past 4 years.

Oh, I remember those arguments well. There was no way x86 could scale past 200MHz...

However, the difference is that the prior arguments assumed that x86 would stall out on its own and some other ISA would supercede it, ignoring the tremendous momentum of legacy compatibility and the fact that x86 was (and remains) a cash cow.

Now, however, it's Intel who are deliberately trying to phase out x86. This is not some handwaving appeal to "market forces" or design elegance. The dominant chip maker has spent years and billions of dollars trying to supplant x86. They have a sort-of backup plan if Itanium fails (Yamhill), but it leaves Intel stuck in 32-bit mode, and that much less able to compete with the likes of IBM. It also opens a big door for AMD.

Ironically, they're also under pressure from Motorola(!) to not focus so much on speed-at-all-costs designs, because of the success of Apple's laptops. If Apple continues to press that advantage with the 970, they'll be able to press into territory that Intel has no answer for.

They'd better be able to, because Mot cancelling the 7457-RM tells me that we won't see much more of the G4.

Well, the circumstances are different now that its IBM's ass on the line.

I guess this next year will prove us all right or wrong.

The 970 will be the beginning of the end of the G4 but it is IBM's 9xx roadmap that will change everything.

As for IBM's handling of the G3. I've a quiet side bet that in the next year the G3 will show Moto' how the G4 should have been done.

Hey, just my gut...

Lemon Bon Bon

Not likely. It is to IBM's advantage to produce their own chips. If IBM decided to go with Intel chips, there would be little difference between their hardware and that offered by anyone else. This is no way to dominate the market, which IBM likely wishes to do. So, even if the 970 were to cost IBM a little more than a chip from Intel, it gives IBM opportunity to have well differentiated products, which have potential to beat the competition.

Pursuing the mini-me versions of their Power series chips give IBM the hope of taking the lead in CPU performance. If they were stuck with Intel, IBM would not be able to rise above the crowd of other hardware makers. There would be no hope of being any better or any different from all their competitors in the lower end server market.

There is a lot more than cost-of-CPU at stake here.

I disagree. IBM culture has changed dramatically since 93/94 when they lost billions upon billions. They are a business above all else. They want a profit. If the best profit for them is from Intel chips (due to low cost and high demand), they will ditch the 970 out of their Blades and follow the cash. I am actually concerned about IBM's use of the 970. I hope their salesman can get their 970 Blades into customer sites; otherwise, IBM can (and has) lose interest quickly. Already, we are seeing a half-hearted attempt by IBM. They are promoting their "fastest in class" Itanium blades all the while saying they plan to introduce 970 blades. They have three blades to choose from (before long): Xeon, Itanium, and 970. They are hedging their bets before the 970 is out of the door.

If they were in this for dominance, they wouldn't have pulled out of the PC market. They lost money. They pulled out. Business decision.

I hope that the 970 and its future follow-ons prove to be highly successful for IBM because I want them continually advanced and in my Macs. I believe that as soon as there isn't a solid business (read revenue) reason for the 9x0 series, IBM would stop using it. That is not to say they wouldn't still make it, but, like the G3, they may just keep it going and sell it to other contracted users. IBM will have an interest in the 9x0 series for at least 3-5 years, minimum, but, if they can't sell them to their own customers, they may give up and go with what their customers want.

To summarize, all I am saying is that IBM is not going to make 9x0 chips just for the good of Mac customers. If IBM customers don't adapt the 9x0 series and help make it profitable for them, they most likely (and from a business stanpoint should) lose interest in further developing the 9x0 series. Of course, this point is moot if Apple can take 15% of the computer market and then, the profits from Apple's purchasing of IBM chips would probably justify IBM's continued investment.

Can Linux help us out here? That is to say, can the speed of Linux running on a 970 and compiled with the latest and greatest GCC, out-perform whatever blade based answer for the 970 Intel has, which is also running Linux?

If so, then we're likley to see 970's around for a while.

... and while we're at it, how the hell could Intel sell a comparable blade chip for 25% the price of what IBM can make 970's for, Hmmmmmmmmmm?

I don't think that they can. It was merely a ridiculous example that IBM--for economic reasons--might lose interest in the 9x0 series. I certainly hope and want IBM to beat out Intel, but, the fact is that Intel is the 800lb gorilla. They have all of the engineering and monetary leverage. Everytime AMD competes directly against them, they just squeeze a little more.

The example was along the lines if Intel wanted to pressure IBM into really pushing its chips by giving them a big incentive... Intel could decide to sell its chips to IBM for a loss if they thought it would help them.

I imagine that Linux on the 970 really rocks. I don't know about IBM's customers, but my fear is that they may have resources tied up in Intel Linux (such as Oracle, etc.). IBM would have to get DB2 running great on it since there isn't (at least the last time I saw) a Linux PPC version of Oracle.

The flip side of the coin of IBM's interesting in PowerPC is that their big iron is POWER and will remain POWER. The POWER4, 5, 6 lineage offers advantages over the IA-64 and commands large enough margins that IBM's business model is solid and likely to remain so. Given their continued use of POWER, having PowerPC versions which are compatible and can be used in workstations, desktops, laptops, blades, and massively parallel systems (including the Cel architecture) is very attractive. IA-64 does not offer this kind of scalability and can't / won't for quite some time due to the nature of the EPIC design. PowerPC, on the other hand, scales well from tiny embedded chips to the big iron POWERs.

A bit of prognostication: Personally think the future looks bright for the PowerPC once again, after the dark winter of the last couple of years. Apple will no doubt hedge their bet, and wisely so, but will use PowerPC for the foreseeable future (just maybe not exclusively). IBM is with POWER through to at least POWER6 and whatever PowerPC derivative is based on it (990?). If they happen to get the PS3 and next GameCube contracts too then the volumes of PowerPC processors they sell will be insane, making the PowerPC the jewel of IBM's crown.

One problem I see in Apple licensing 10.3 to IBM for use with their x86 machines is Microsoft. They won't recompile their Office code for "OS X for Intel". Think about it. This summer they'll barely have a native Outlook Exchange client. This would seriously hamper OS X's adobption by big business. The only out I see is a secret Apple branded Office app suite that is MS compatible and just as powerful.

Hmm.

Wrong, they pulled out because a smart decision... What exactly differentiated their systems from the competition? That's right, NOTHING. It was just another x86 crate running Windows (Yaaaaaawn....) with chips from Intel. You say IBM is a *business* above all else, but they are a smarter business and will avoid the mistakes others have made. The difference is that IBM is in for the long haul (i.e., they are planning on sticking around for a very loooong time. They got out of the desktop PC business because there was no money in it. Why compete, so they can become another Gateway? Compaq? Packard Bell? The trouble is that everyone competing with Dell really has nothing different to offer, In fact, the only thing that differentiates them (and I've said this in the past) is a sticker on the side of a cheesy plastic case and price - nothing else. Apple gives IBM an inroad back into the desktop market by actually allowing them to offer a platform that differentiates itself from the played-out PC OEM offerings, so I'm not convinced that IBM is stupid enough to follow the short-term/quarterly report mode of doing *business* (profit taking). They're smart enough to know *not* to rely on the competitions offerings because it will only allow the competition to become stronger and IBM weaker... Completely unwise and strictly short-term which equates to *not good*. If IBM rejects Intel's offerings at any cost then Intel looses and looses big. Again, I think IBM knows this, since it will make Intel weaker (i.e., they will be selling less chips at higher prices) The flip side is that IBM becomes stronger. Here is a huge "what if?"

What if Apple decides to license OS X to IBM and have it run on some of the bigger-iron IBM systems? Imagine if IBM started offering turn-key OS X solutions to corporations. I'll lay odds that a TON of business would move off of Windows onto an alternative platform which would likely save them money. Just a thought.

--
Ed

> Outlook Exchange client. This would seriously hamper OS X's adobption by big business.

unless apple makes mail.app exchange compatible.

See above...

And who said anything about IBM licensing it for it's x86 offerings?? I was suggesting a REAL alternative to corporate other than something from Micro$oft... Don't forget, IBM was hosed by Micro$oft once, I think IBM is planning a little get-back... Oh, and I'm not convinced the majority of clients are even using all those nifty bells and whistles within Office to their fullest extent. Until now there was nothing. Now there is OS X and it's advancing-- fast.

--
Ed M.

Because IBM is in business to make money is exactly why IBM will do their best to make the 970 and 980 succeed. IBM took a beating when they had to be just another vendor of Intel powered PCs. Now, if IBM uses Intel chips in all their lower end servers, the same thing will happen to them again. If they go with the Itanium because it is cheaper for their high end servers, it will happen them in that market too.

IBM will pull out all stops to keep the PC debacle from happening again. That is why they need the Power series and smaller, cheaper versions of these CPUs. They will do everything they can to make these the best performing CPUs in the world.

You misinterpret why they sell products with Intel chips. It's just good marketing. IBM becomes a one stop market for your server needs. "You want a Xeon, we got it. You want an Itanium, we got it. You want a 970 or Power4, its here too. We don't mind which one you choose; we have it to sell." This keeps their sales up while PPC servers get better established. This also makes IBM seem pretty neutral regarding their products, the perfect position to subtilely promote their own CPU.

If it's secret, how will people know to buy it?

Sorry, I couldn't resist.

Tears of sadness... (sniff, ahem...ahw...you got me...yes, tears of joy...)

With the deliriously long delivery time on it, I'm glad it's been canned. Who wants a G4 with most of the G5 tech' in it three years after the fact?

At least its spare endless posts by LBB complaining how slow Apple cpus are and endless posts by apologists saying: 'It's fast enough for email and word processing...and hey, it aint too bad, cos you that mega-hurtz myth...well...er...it's true y'know...an'...why don't you go and buy a PC? Cos sales of 'power'Macs are okay at 55,000 and when the economy rebounds...and when we get Quark for X...' (Cue the tapping of Ruby red slippers..., 'There's no place like home, there's no place like home...')

Lamarola. I hope by early 2004, we've forgotten all about them.

That they lasted four years. Yeesh. Must have been one hell of a contract.

Lemon Bon Bon

But this is already happenning now - its called Linux. Why do you think MS sees Linux as a REAL threat, than even OSX, for obvious reasons OSX is currently not supported on Intel CPU's.

Further, Programmer echoed my thoughts succinctly. Remember, when you talk about IBM - you should invariably discuss divisions that due to market realities cannot carte blanche support others. The server division WILL purchase from microelectronics in quantities that the market dictates. ie If the 970 Linux Blade sells well - IBM as whole will be happy. The top down development approach by IBM Micro - helps apple (with their little sideline requests - Altivec).

Im not quite convinced there is enough benefit for IBM's server or other product divisions to support Mac OSX &/or Server, unless IBM executed a reseller agreement like Dell and purchased product directly from Apple. Neat.

I was out this weekend. I was specifically replying to Programmers speculation in the last page. Sorry!