IBM unveils dual-core PowerPC chips up to 2.5GHz
Presenting at the Power Everywhere Forum 2005 in Japan, IBM today formally introduced a dual-core version of its PowerPC 970 (G5) processor, which could find its way into Apple Power Macs in the coming months.
The 64-bit chips, code-named Antares, contain two processing units per chip, each with their own execution core and Level 1 cache. Additionally, each core includes a storage subsystem with 1MB Level 2 cache, making the chips twice as efficient as IBM's current 970FX PowerPC G5 processors.
Officially labeled the PowerPC 970MP, the chips pack several power conservation features that allow frequency and voltage demands to adjust on the fly. It's also possible to completely cut power to one of chip's cores when high performance computations are not needed.
According to IBM, the 970MP will be made available in speeds ranging from 1.4 to 2.5GHz. It's believed that Apple has been working with prototypes of the chips since 2004 and could use them in a forthcoming update to its Power Mac G5 desktops, the company said.
Along with the dual-core 970MP, IBM today also officially introduced a new family of low-power PowerPC 970FX chips running at speeds up to 1.6GHz with a 512K Level 2 cache. PowerPC 970FX chips are currently used in Apple Power Mac G5 and iMac G5 systems.
Initially, the new 970FX chips will be available in three flavors: 1.2GHz, 1.4GHz and 1.6GHz. IBM says the 1.6GHz version will typically operate at 16 watts while the the two former versions require a cool 13 watts, possibly allowing Apple to release a G5-based PowerBook.
An official press release on the new chips is available in Japanese. As of press time, IBM had yet to update its US website with information on the new chips. A set of slides from the company's presentation are also available via ASCII24, a publication based in Japan.
The 64-bit chips, code-named Antares, contain two processing units per chip, each with their own execution core and Level 1 cache. Additionally, each core includes a storage subsystem with 1MB Level 2 cache, making the chips twice as efficient as IBM's current 970FX PowerPC G5 processors.
Officially labeled the PowerPC 970MP, the chips pack several power conservation features that allow frequency and voltage demands to adjust on the fly. It's also possible to completely cut power to one of chip's cores when high performance computations are not needed.
According to IBM, the 970MP will be made available in speeds ranging from 1.4 to 2.5GHz. It's believed that Apple has been working with prototypes of the chips since 2004 and could use them in a forthcoming update to its Power Mac G5 desktops, the company said.
Along with the dual-core 970MP, IBM today also officially introduced a new family of low-power PowerPC 970FX chips running at speeds up to 1.6GHz with a 512K Level 2 cache. PowerPC 970FX chips are currently used in Apple Power Mac G5 and iMac G5 systems.
Initially, the new 970FX chips will be available in three flavors: 1.2GHz, 1.4GHz and 1.6GHz. IBM says the 1.6GHz version will typically operate at 16 watts while the the two former versions require a cool 13 watts, possibly allowing Apple to release a G5-based PowerBook.
An official press release on the new chips is available in Japanese. As of press time, IBM had yet to update its US website with information on the new chips. A set of slides from the company's presentation are also available via ASCII24, a publication based in Japan.
Comments
Yonah looks to be far more powerful than a
"mobile G5"
And the dual core deal is not as impressive as AMDs offering. Shame.
I'm all for a 1.6 GHZ G5 Powerbook, but, wow. We should have gotten at least 2 GHZ in a mobile chip by now.
I am really glad about the decision now. I was skeptical earlier.
Originally posted by 9secondko
And the dual core deal is not as impressive as AMDs offering. Shame.
On what evidence are you presenting this assertion as a fact?
Originally posted by 9secondko
And the dual core deal is not as impressive as AMDs offering. Shame.
What a pile of crock. That will eat any dual core Athlon.
Too little, too late, too bad.
Hey, I wonder if Freescale's new 90n G4 outperforms the mobile-G5?
(and it's also worth noting that IBM has just demonstrated these chips. It may be half-a-year before they can mass produce them)
The low power G5s will be cherry picked dies, like AMD cherry picks Athlon 64 dies that run cool and brands them as Turion (35W and 25W max). Intel's max is 27W. I don't know what the max is of this new processor, but if 13W is typical then I imagine that it'll be around 26W or so.
So IBM announces 1.2, 1.4 and 1.6? So what! Previously Apple has had faster machines out than IBM's product line would suggest. Maybe IBM are making 1.8GHz 16W G5s for Apple?
The real issue is the power management features on the latest processors. 13W AVERAGE would be rather poor.
The dual-core looks quite interesting however.
My answer is yes for the MP in desktop computers and no for the mobile G5
Of course I could be wrong.
Time will tell
Who want to bet a bottle of Champagne with me
Originally posted by kenaustus
Looks like IBM is trying to show that Apple was wrong. Unfortunately the only major customer they will have for it is Apple, and only for a year or so. Just shows how important it is to be proactive instead of reactive.
Not really. Like Steve Jobs said, there will be many more great powerpc products to come. This is one of them.
Originally posted by 9secondko
No wonder Apple went with Intel.
Yonah looks to be far more powerful than a
"mobile G5"
And the dual core deal is not as impressive as AMDs offering. Shame.
I'm all for a 1.6 GHZ G5 Powerbook, but, wow. We should have gotten at least 2 GHZ in a mobile chip by now.
I am really glad about the decision now. I was skeptical earlier.
ghz does not = preformance. not by a long shot.
i dont feel a need to go any further than that.
I don't think that there is any question that Apple will use these. One question though is whether these chips will be made on the 90nm lines, or whether they will wait for the 65. That affects the timing of the introduction.
If they are 90nm then we might expect them for January, possibly (though doubtfully) for September. 65nm might mean March, at best, unless IBM has made great strides in the new process (which is certainly possible).
Will Apple use these singly, or will we see dual duals as well? Hmm.
If they run cool enough (and are cheap enough), maybe dual dual Powermacs and single duals for the iMacs.
Where do the Powerbooks fit into this? Despite what some here are saying, we won't have an idea until we know more about the FX's. I can see a possibility, but it's too early to tell.
Originally posted by jms698
What took IBM so long? They should have released these chips (especially the low power G5) a year ago. That would have been just right. As it is:
Too little, too late, too bad.
Hey, I wonder if Freescale's new 90n G4 outperforms the mobile-G5?
(and it's also worth noting that IBM has just demonstrated these chips. It may be half-a-year before they can mass produce them)
i personally would have prefered these annoucements a day after steve jobs intel lovin speech :-) Sorry, steve is wrong this time
Originally posted by mike12309
i personally would have prefered these annoucements a day after steve jobs intel lovin speech :-) Sorry, steve is wrong this time
Apple's switch is most likely more complicated then we think.
These processors are still behind in the sense that they are coming out 8 months after the x86 duals are. We expected to see IBM's duals BEFORE the x86 ones. Why this is so, we don't know. But it is telling.
The speculation that they would debut at 3.0GHz is also obviously wrong (I kept on insisting that they would be at 2.5, but people would flame me about that).
Nevertheless, if Apple comes out with a dual dual Express machine within the next 9 months or so, I will buy it. The sooner the better, because I don't like buying the first iteration of ANY new technology.
But, going to Intel is probably still better for the long haul.
I just got my PB so I guess I will just wait for Macintel PB's
Originally posted by mike12309
ghz does not = preformance. not by a long shot.
I dont feel a need to go any further than that.
Um, in the case of the G5, it is already common knowledge that the AMD Athlon 64 keeps up very well and the Pentium M destroys the G4. The G5 does better than G4 at many things, but requires at least a certain amount of clock speed to be effective. I would say 1.8 at minimum (yes, I am aware of the Powermacs low end debut at 1.6 and that was quite a shame in the benchmarks wasn't it?). A current 2.1 GHz Pentium M would probably outperform it. I hate to say this as I have bee anticipating a G5 Powerbook for quite some time, but it is time to face facts. Apple did the right thing going to Intel. This 1.6 G5 would have been nice a year ago. Now it is quite seriously a joke. I would not mind buying one as it will be superior to the current G4, but it will not blow it away like it would if it were at 2 GHZ.
I cannot wait to see a barefeats.com benchmark pitting the 970 FX-low power against ANYTHING. Be it a G4, Pentium M or Current Imac G5.
GHZ DOES matter, contrary to some opinion.
If that were the case, no one would mind the 3 GHZ miss by IBM. In the case of AMD, clock for clock and Ahtlon 64 FX runs neck and neck with 2 G5. The G5s are faster when they are clocked higher and AMD is faster when they are clocked higher. However, the Pentium M runs almost as fast at 700 MHZ less.
Clock it at 3 GHZ and the M would be an almighty jugernaught.
Architecture is king, sure. But, GHZ is what gets the king around.
That V8 engine in your Mustang cobra would not be so fast if it did not rev so high.
I don't think I need to go any further than that.
Too little, too late from IBM. They tried to play Apple and got dropped. Now they are trying to save face. Apple does not need to beg IBM and wait forever while IBM gets cozy with Apple competitors and puts Apple (longstanding and loyal customer) on hold for them (just plain bad business ethics). They can just get BETTER CPUs from someone else. Bold move. And we are all going to benefit from it.
As far as G5 vs. AMD and Intel CPUs, go to anandtech.com, who recently did a thorough comparison.
This may prove not to be the case at all, as the statement says...
".. the 970MP will be made available in speeds ranging from 1.4 to 2.5GHz. It's believed that Apple has been working with prototypes of the chips since 2004 and could use them in a forthcoming update to its Power Mac G5 desktops.."
Apple have previously seen and tested these chips long before they announced a move to Intel, therefore I would trust the Apple techs intuition on the future of Mac processor type and not go down the road of "I told you so".
My $0.02
Originally posted by 9secondko
Um, in the case of the G5, it is already common knowledge that the AMD Athlon 64 keeps up very well and the Pentium M destroys the G4. The G5 does better than G4 at many things, but requires at least a certain amount of clock speed to be effective. I would say 1.8 at minimum (yes, I am aware of the Powermacs low end debut at 1.6 and that was quite a shame in the benchmarks wasn't it?). A current 2.1 GHz Pentium M would probably outperform it. I hate to say this as I have bee anticipating a G5 Powerbook for quite some time, but it is time to face facts. Apple did the right thing going to Intel. This 1.6 G5 would have been nice a year ago. Now it is quite seriously a joke. I would not mind buying one as it will be superior to the current G4, but it will not blow it away like it would if it were at 2 GHZ.
I cannot wait to see a barefeats.com benchmark pitting the 970 FX-low power against ANYTHING. Be it a G4, Pentium M or Current Imac G5.
GHZ DOES matter, contrary to some opinion.
If that were the case, no one would mind the 3 GHZ miss by IBM. In the case of AMD, clock for clock and Ahtlon 64 FX runs neck and neck with 2 G5. The G5s are faster when they are clocked higher and AMD is faster when they are clocked higher. However, the Pentium M runs almost as fast at 700 MHZ less.
Clock it at 3 GHZ and the M would be an almighty jugernaught.
Architecture is king, sure. But, GHZ is what gets the king around.
That V8 engine in your Mustang cobra would not be so fast if it did not rev so high.
I don't think I need to go any further than that.
Too little, too late from IBM. They tried to play Apple and got dropped. Now they are trying to save face. Apple does not need to beg IBM and wait forever while IBM gets cozy with Apple competitors and puts Apple (longstanding and loyal customer) on hold for them (just plain bad business ethics). They can just get BETTER CPUs from someone else. Bold move. And we are all going to benefit from it.
As far as G5 vs. AMD and Intel CPUs, go to anandtech.com, who recently did a thorough comparison.
When the cry that GHZ doesn't = performance goes out, it is correct, mostly.
It just so happens right now that AMD's chips are pretty well matched to the G5's, even slightly faster, clock for clock.
But overall, GHz doesn't say much by itself. As Einstein would say "It's all reletive". The architecture determines what the clock will do, or even what it means.
Going back in time, a 1GHz G4e would perform as well as a P3 at 1.3GHz in integer, and be outfront in floats. Add Altivec, and it way way ahead in certain tasks. Move the P3 to 1.8GHz and the G4e was well behind though Altivec held up under those same tasks. Higher than that, forget it!
The G5 goes further. A 2GHZ unit would equal a 3.06GHz P4, pretty much. The problem was not in the slow ramping up of the clock though. IBM actually did at least as well as Intel did there. 2GHz to 2.7GHz is 135%. 3.06GHz to 3.8GHz is 124%. That's pretty good!
The problem is that Intel keeps increasing the caches. IBM hasn't. Until now. And only, as far as we know, only on the dual.
That's made a big difference.
Originally posted by 9secondko
Clock it at 3 GHZ and the M would be an almighty jugernaught.
it would also be running at 80C like the P4 line (which hits about 70C under load).
im not sure how much of a juggernaught it would be while its on fire or melting the plastic around it.
The M is good because it cuts its Mhz as much as it can to reduce heat and increase battery... and it is good... but it proves your whole "all hail the Mhz" attitude misplaced.
Mhz nowadays is only good to tell different models of the same chip apart, like i can assume a 2.7 P4 EE is better than a 2.2 P4 EE... try to use it to compare different chip cores and models and your gonna find yourself owning a Celeron 4.6Ghz and getting burned.