IBM to Apple: Eat This, New PPC Chips!

[boing]

http://www.theregister.co.uk/2005/07/08/ibm_powerpc/

IBM to Apple: eat these chips
By Andrew Orlowski in San Francisco
Published Friday 8th July 2005 02:45 GMT

As a parting shot to Apple, a piqued IBM today announced two new lines of its PowerPC 970 processor, better known as the G5. And they're impressive beasts.

A new, low-power 970FX consumes between 13W and 16W at frequencies of 1.2GHz, 1.4GHz and 1.6GHz. That's more than the 10W that the Freescale MPC7448 found in today's 1.5Ghz PowerBooks consumes, but around half the maximum power consumption of Intel's Pentium M, which powers today's Centrino laptops. IBM is also unveiled the dual-core 970MP codenamed 'Antares', at clock frequencies of 1.4GHz to 2.5GHz. Each core has 1MB of L cache, and one core can be turned off to save power.

Contractual commitments and supporting remarks from IBM since the WWDC announcement suggest that these processors will form the final PPC-based Macs that Apple produces, despite the ill-tempers raised by Apple's switch to Intel.

Clearly responding to the belief that only Intel can afford to invest and manufacture in volume, IBM showed a slide that the pointed to the investment in its semiconductor platform. Between them, IBM, AMD, Sony, Samsung, Infineon and Toshiba have invested $17bn worth of capital expenditure on silicon, compared to $12.6bn by Intel. The PPC and Cell will form the heart of games consoles from Microsoft, Nintendo and Sony over the next 12 months.

Exactly ten years ago, Apple's board agreed to sell the company to IBM, which was trying to create a volume business for its PowerPC chips. The deal was done and dusted, and only fell apart at the signing ceremony. Now IBM looks like it will finally have a volume business - but Apple won't be part of it. ®

[Magic-Merl]

Why did Apple decide to move to Intel.

The PPC/Cell specs were enough to make the mouths of consumers drool.  The very same consumers that would have bought a non-Microsoft OS to run on it.

Shame.

OS4 or even Linux may have a niche market here. You never know?

[bloodline]

Magic-Merl wrote:
Why did Apple decide to move to Intel.

The PPC/Cell specs were enough to make the mouths of consumers drool.  The very same consumers that would have bought a non-Microsoft OS to run on it.

The PPC roadmap does not project as good a performance/watt ratio as the Pentium-M roadmap.

Apple, want to sell high powered computers that are cost effective. They also want to make sure their CPU supplier has the volume and capacity to meet their needs.

Also by moving to the x86, they have the advantage of the Intel/AMD competition.

[Magic-Merl]

But the PPC market, IMO, will potentially, grow due to Cell/PPC combo.  Granted, at face value, the PPC seems to be nothing more than a Co-Processor or controller for the Cell but the fact remains that PPC will still be manufactured and a future roadmap does still exist.  Unless I am mistaken (please feel free to jump in at this point) the Cell will perform beyond the Intel roadmap [as far as performance is concerned] and should it continually grow will leave the market wide open.

[T_Bone]

Heh, they might just as well announce a 10Ghz G5, without a computer that uses it being available, who's going to know if they're making it up or not? ;-)

[bloodline]

Magic-Merl wrote:
But the PPC market, IMO, will potentially, grow due to Cell/PPC combo.  Granted, at face value, the PPC seems to be nothing more than a Co-Processor or controller for the Cell but the fact remains that PPC will still be manufactured and a future roadmap does still exist.  Unless I am mistaken (please feel free to jump in at this point) the Cell will perform beyond the Intel roadmap [as far as performance is concerned] and should it continually grow will leave the market wide open.

The Cell is not relevant to this discussion, it's not a desktop CPU.
The PPC will exist, and will continue to be developed for the high performance embedded market by Freescale. IBM have stated that there isn't any money to develop the 970, at least at the pace intel can push their chips.

[Sparky]

T_Bone wrote:
Heh, they might just as well announce a 10Ghz G5, without a computer that uses it being available, who's going to know if they're making it up or not? ;-)

Well 6Ghz ones do exist in IBM R&D labs ... had a presentation the other month from a US IBM R&D guy, being an IBMer has it's advantages  :-D

[melgross]

I still feel that the PPC architecture is cleaner and better. But IBM isn't really interested in developing it at the pace that Apple needs. If Apple invests $350 million as did Sony, fine. But why should they do that? It would only bring the cost of the cpu up that much further.

The Cell - ha! The amount of work that it would take to get that to be competitive would be enormous. And it might not work.

Whatever works in the lab means nothing in the real world. IBM sampled 2.8GHz G5's two years ago, where are they?

Apple had no choice. They did what was best for the future.

The Amgia community should be so lucky as to have someone like Jobs who can make a decision and run with it.

Sparky wrote:

T_Bone wrote:
Heh, they might just as well announce a 10Ghz G5, without a computer that uses it being available, who's going to know if they're making it up or not? ;-)

Well 6Ghz ones do exist in IBM R&D labs ... had a presentation the other month from a US IBM R&D guy, being an IBMer has it's advantages  :-D

[DonnyEMU]

The reality of all of this is Jobs knows the future is "laptop" portable machines with low power, high performance (found in the intel roadmap). The G5 laptops aren't as capable as they could be right now..

Where the move to Intel matters is right there, in the laptop market.. Desktops sales don't even match laptop computer sales anymore.

IBM's volume business is really NOT PC sales, it's more server solutions (linux ones). Not even the SAME business so this is another apple/oranges thing..

The cell is focused on being something other than a desktop CPU, and the game machines coming out have specialized processors for gaming. They may not be the best implementations if one is to consider a general purpose computer..

[melgross]

So, when is the portable Amiga coming out? :-)

[T_Bone]

melgross wrote:
So, when is the portable Amiga coming out? :-)

The good news is, that they're just as widely available as the desktops are at the moment.  :-)

[Hammer]

Refer to
http://www-6.ibm.com/jp/press/20050707003.html
Or
http://www-1.ibm.com/press/PressServletForm.wss?MenuChoice=pressreleases&TemplateName=ShowPressReleaseTemplate&SelectString=t1.docunid=7773&TableName=DataheadApplicationClass&SESSIONKEY=any&WindowTitle=Press+Release&STATUS=publish

Notice "under typical workloads" in IBM's offical statements.

[Hammer]

Magic-Merl wrote:
But the PPC market, IMO, will potentially, grow due to Cell/PPC combo.  Granted, at face value, the PPC seems to be nothing more than a Co-Processor or controller for the Cell but the fact remains that PPC will still be manufactured and a future roadmap does still exist.  Unless I am mistaken (please feel free to jump in at this point) the Cell will perform beyond the Intel roadmap [as far as performance is concerned]

Not with Intel's own "Special Purpose HW Engine" (SIMD/DSP) co-processor array. Intel already has experience  a 16 core SoC with XDR infrastructure (e.g. Xscale based IXP2800).

[Hammer]

The PPC/Cell specs were enough to make the mouths of consumers drool

Cell's POWER(PPE) core is nothing like PowerPC970 in terms of OOO processing.

[minator]

Notice "under typical workloads" in IBM's offical statements.

According to Processor Report (the free version of Microprocessor report), the thermal envelope is 20W.

Cell's POWER(PPE) core is nothing like PowerPC970 in terms of OOO processing.

They've traded clock speed for complexity, Cell should clock up to 4GHz without problems (it's held lower in the PS3 because they don't want it becoming too noisy).  At 4GHz the PPE alone should kick a 2.7GHz 970's ass.

The Cell is not relevant to this discussion, it's not a desktop CPU.

What is?
Cell isn't being marketed as a desktop processor - yet.
It is general purpose though, even the SPEs are general purpose.  The difference is they are optimised for certain types of problems, x86 are optimised for different types of problems.

The 970 wont be developed much further, it'll probably get a die shrink next year but not much beyond that.

That only* leaves x86 on the desktop, you really think it'll stay that way for long?

OK technically, Genesi and some RiscOS machines are on the desktop but they're not exactly big compared to the x86 market...