Apple, IBM and Magma - Powermacs sooner
Call my crazy or whatever 
but I find it interesting that IBM and Magma DA are working together at the East Fishkill, NY plant (home to the PPC970):
<a href="http://www-3.ibm.com/chips/news/2002/1028_magma.html"; target="_blank">Magma and IBM collaborate to accelerate chip designs from concept to silicon</a>
My favorite read from the article is, "...customers can benefit with faster time-to-market and enhanced product designs."
And I wonder what big customer would benefit from that?
I vaguely remembered that ex-Apple employee Lung Chu joined Magma DA, which is headquartered in Cupertino, earlier this year.
<img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
I am guessing new Powermacs sooner than Fall 2003.
but I find it interesting that IBM and Magma DA are working together at the East Fishkill, NY plant (home to the PPC970):<a href="http://www-3.ibm.com/chips/news/2002/1028_magma.html"; target="_blank">Magma and IBM collaborate to accelerate chip designs from concept to silicon</a>
My favorite read from the article is, "...customers can benefit with faster time-to-market and enhanced product designs."
And I wonder what big customer would benefit from that?
I vaguely remembered that ex-Apple employee Lung Chu joined Magma DA, which is headquartered in Cupertino, earlier this year.
<img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
I am guessing new Powermacs sooner than Fall 2003.
Comments
Is that Rush, 2112, in your sig? I really dug that album back in high school.
<strong>Interesting!
Is that Rush, 2112, in your sig? I really dug that album back in high school.</strong><hr></blockquote>
indeed.
EDIT: to stay on topic...It seems like this development is too late to have much impact on the PPC 970's development, but then perhaps this was in the works since Apple and IBM began courting.
Something tells me that Apple's got a PPC 970-ready mobo in the bowls of their skunkworks and that it's been in testing with 970 samples for some time....long enough to work out many of the bugs. I wouldn't be surprised if Apple announced 970 based towers as soon as IBM cranks up the fab to full capacity. Maybe early next year, perhaps a MWSF announcement, shipping in March? It's been done before.
[ 11-04-2002: Message edited by: Junkyard Dawg ]</p>
"We claim:
1. A method for operating a microprocessor to reduce power consumption, the microprocessor including a functional unit formed of a plurality of stages, the method comprising: evaluating instructions to be executed to determine the operation type of each of said instructions; producing activity indicators based upon the operation types of said instructions; controlling the supply of current to each of said plurality of stages such that only selected stages of said plurality of stages will draw current from a power supply, the controlling being based upon activity indicators associated with each of said stages; advancing said instructions within the microprocessor; and executing said instructions that are within each of said selected stages.
2. A method for operating a microprocessor as recited in claim 1 wherein said microprocessor is a very long instruction word processor."
BTW, check <a href="http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=/netahtml%2 FPTO%2Fsearch-bool.html&r=0&f=S&l=50&TERM1=Apple&FIELD1=AS&co1=A ND&TERM2=&FIELD2=&d=PG01" target="_blank">this</a> for all kinds of neat Apple laptop applications.
However, I think I'll resist a "*CONFIRMED: Apple to use IA 64 in PowerMacs!!!" thread.
<a href="http://www.insanely-great.com/news.php?id=1210"; target="_blank">IGM: Rumor MWSF Powerbook G4 Gets VBT...</a>
<strong>As I understand it, Power/PowerPC are not VLIW ISAs. Curiouser and curiouser.
However, I think I'll resist a "*CONFIRMED: Apple to use IA 64 in PowerMacs!!!" thread.
Well, if I'm not mistaken, the POWER architecture does break instructions into IOPs, which doesn't make it a VLIW processor by any means, but it does share that trait with VLIW.
So if the technique works with POWER (and the 970) then it may work equally well with VLIW. Patenting it with VLIW doesn't mean that they are going to do it, it just means that Apple is willing to tell the world *how* to do it in exchange for the world asking Apple for permission to actually do it for the next decade or so (which Apple may be willing to do for a small fee.) All Apple did was throw a wall up in front of Intel and AMD. My guess is that Apple/IBM/Mot share patents in this area.
Basically it means that you probably *won't* see this technique used in IA64 unless Apple gets a check.
Intel's IA-64 is actually called "EPIC" which is a specific variation on the VLIW theme that I won't go into here.
The Apple patent above is basically a way to control the power consumption and heat generation of a processor by only powering the parts currently in use. Since they explicitly describe it for both VLIW and non-VLIW I would suspect that they want to use this for both PowerPCs and other things. Remember a couple of months back when Moki was talking about DSPs in a new chipset from Apple? Those DSPs (or media processors or GPUs or whatever) could leverage the technology described to keep power consumption under control. This would be especially important in a motherboard chipset since the thing is always in use, even if the DSP portion is not.
I don't think this has anything to do with the IBM-designed PPC 970...
In any event, sure, Apple has had prototypes of 970-based machines for a while... that's how it works. Prototypes always exist long before products do, the same way novels exist in unfinished form long before you can buy 'em at Borders.
This seems to click in nicely with the basic PowerPC philosophy, that the design should enable easy, custom specific, adding or deletion of subprocessing units. In other words, it is a way to develop new versions of PPC processors easily, for instance adding extra SIMD units - or specific DSP units. On-board bus processing also springs to mind.
They also clearly state in the article that it can be used to speed up tasks "such as..." embedded processors - it doesn't leave out desktop versions of CPUs.
engpjp
PS - What was the title of that PPC specification - Red Book?
[ 11-05-2002: Message edited by: engpjp ]</p>
<strong>... a functional unit formed of a plurality of stages...only selected stages of said plurality of stages will draw current.
</strong><hr></blockquote>
I read it to say that the power limitation is based on shutting down individual PARTS of the CPU when they are inactive, not on lowering the frequency.
Programmer?
engpjp
<strong>
I read it to say that the power limitation is based on shutting down individual PARTS of the CPU when they are inactive, not on lowering the frequency.
Programmer?
</strong><hr></blockquote>
I'd agree with that.
The PowerPC philosophy has never really been about modular chip designs that customers can extend for their own purposes... that is a relatively new direction Moto and IBM have adopted for their embedded products, some of which are PowerPC-based. This Magma product seems specifically aimed at that audience and while Apple might go in that direction in the future they haven't done it in the past. Adding the SIMD unit isn't an appropriate kind of add-on because it is too tightly integrated into the processor. Alternative bus interfaces or I/O devices are typically the kind of customization these things are aimed at which is why the embedded space is the most likely market for it.
The embedded DSPs I mention above are in the motherboard chipset, not the processor, by the way. At least that's how Moki outlined it, and this makes sense since the goal would most likely be to put more processing power on the memory (not CPU) side of the MPX bus.
<strong>My read of the article is that it has to do with helping customers get their custom-designed chips to market using IBM's new fab facility in a more efficient manner.
I don't think this has anything to do with the IBM-designed PPC 970...
In any event, sure, Apple has had prototypes of 970-based machines for a while... that's how it works. Prototypes always exist long before products do, the same way novels exist in unfinished form long before you can buy 'em at Borders.</strong><hr></blockquote>
AND, generally, the longer a novel remains in an unfinished state, the closer it becomes to being done.
So if PPC 970s have been in testing for some time at Apple, then surely Apple has a fully completed 970-based mobo set to go. I know this is a bit optimistic but it's fun to think about the 970 showing up early next year in a dual configuration....it would totally lay waste to anything Wintel brought out for competition.
No, it just means that Apple has had prototype 970's for a few months. New chip interconnect is no small matter.
That said IBM has not begun ramping up for fab either (at least if reports from Fishkill are to be believed), as FK is still running far below capacity, the old fab at FK is being nursed along far beyond its shutdown date.
Don't get too hopeful.
[ 11-05-2002: Message edited by: cowerd ]</p>
<strong>
The embedded DSPs I mention above are in the motherboard chipset, not the processor, by the way. At least that's how Moki outlined it, and this makes sense since the goal would most likely be to put more processing power on the memory (not CPU) side of the MPX bus.</strong><hr></blockquote>
Some years ago, Atari released their last computer model, the Falcon, built on Motorola 68030. Its strength was not the CPU but the fact that it had a DSP built-in which could perform certain tasks, such as compression, decoding, image transformation and sound operations, faster than anything else on the market - except the Mac AVs and specialized hardware. Part of the chipset was a unit that acted as a kind of "switchboard", routing data from processor to processor. The CPU, the DSP, the GPU, and the SPU (plus I/O units) were thus decentered, rather than having the CPU being the command/control center. To what extent would a similar architecture be possible with PPC970, a DSP and one of the new GPUs?
engpjp
<strong>That said IBM has not begun ramping up for fab either (at least if reports from Fishkill are to be believed), as FK is still running far below capacity, the old fab at FK is being nursed along far beyond its shutdown date.]</strong><hr></blockquote>
From IBM's 10/16 8k:
"The new 300 millimeter plant is in test production and on schedule. We have
orders that would fully load the facility well into midyear of 2003. From what
we can tell about our competition, we may be the only manufacturer capable of
130 and 90 nanometer technologies."