Slashdot Mirror
IBM Officially Unveils Dual-core PowerPC Chips
PM4RK5 writes "Today at the Power Everywhere Forum in Japan, IBM officially unveiled its rumored dual-core PowerPC line of chips, the 970MP. Code-named Antares, these chips have been rumored to be under development since 2004. It is believed that Apple has been working with prototypes and is likely to use them in forthcoming updates to the PowerMac G5 line. The press release is in Japanese; as of this writing, IBM has not released an English version. Some of the slides from the presentation given by IBM are available.
The processors pack some impressive specs, ranging from 1.4 to 2.5 GHz and including 1MB L2 cache per core; the chips also include the ability to power down the extra core when it is not needed. Alongside the 970MP, IBM also announced its low-power 970FX chips, ranging from 1.2 to 1.6 GHz, with power consumption ranging from 13 to 16 Watts, respectively."
Looks like Steve Jobs is getting all the things he said he couldn't have. Dual core processors, low-power G5's for laptops, everything.
IBMs own server products and embedded processors. IBM's blue gene used the core from earlier PowerPC series.
Because they are most likely going to use Intel chips to replace the old G4 chips long before they replace the G5s. And they are not going to totally over to Intel till 2007. So, their needs to be updates between now and then.
the world has plenty of PPC chip uses besides filling Macs, from network appliances to video games to Unix & Linux servers and mainframes and supercomputers. Still, Apple chips are almost 1% of IBM's $99 billion revenue, that's a big chunk of money.
The Freescale chip was deemed to hot to be used in an Apple laptop. Presumably the 970MP would be too hot as well, but the low power ones would be well suited to a laptop.
The obvious application for these newly announced chips will be low-end servers and workstations, particularly high density servers for the low power chips. I say low-end not because the chips are slow, but because IBM's OpenPOWER machines are pretty reasonably priced already. In fact, there's enough overlap that I'm not really sure how appealing the 970MP will be to anyone. Apple would have been an obvious customer, but that is pretty temporary, now.
Whoever corrects a mocker invites insult;
whoever rebukes a wicked man incurs abuse.
--Proverbs 9:7
From the notes:
The dual 64-bit core PowerPC970MP(TM) (970MP) is the next evolutionary step in the PowerPC 970 family of microprocessors. The higher frequency grade versions of the 970MP consume higher amounts of power than earlier IBM microprocessors do, and that can cause temperature issues. Each 970MP processor core contains a thermal diode used to monitor its operating temperature. The thermal diode must be monitored to ensure that the maximum operating temperature of the 970MP is not exceeded.
Terrasoft http://www.terrasoftsolutions.com/ is a company that sells IBM powerd desktops that run Linux. I wouldn't doubt they start selling this chips now that they have started selling yellowdog pre-instaled.
Yes, most individual programs don't fully take advantage of multiple processors. However, if you have multiple applications open at a given time and each is actively doing something, that's when you're glad you have processor #2 available.
Foreground app has the first processor, some busy app in the background (file copy, MP3 encoding, DeCSS, photoshop filter, etc) gets the second. You're much happier because your system isn't taking a few seconds to respond to each mouse click.
"Speaking of Roadmaps... Why not would Apple not switch to AMD? AMD's chips run with less power consumption and way less number of transistors. When comparing the Dual core chips from AMD and Intel, AMD wins on power consumption. But I thought Jobs said Intel had the best Performance per watt? ADA4800DAA6CD (AMD Dual core 64-bit): 110W Intel® Pentium® Processor Extreme Edition: 130W These are the latest and greatest from Intel and AMD right?"
This has been discussed about a million times on any site that posted any news about the switch.
You're wrong in two ways. First, you don't understand what Apple cares about. Second, you don't understand the situation in the area that Apple cares about.
Apple cares more about laptops. Intel wins easily in this area. They beat every current or planned PowerPC laptop chip, and they beat every AMD laptop chip. There's basically no serious competition at this point (AMD is trying but they're not yet serious competition).
AMD wins on power consumption on the desktop right now, but Apple cares more about laptops and also Intel is going to be moving their laptop chips into the desktop because the P4s have dead-ended. In the 2006-2007 timeframe, Intel is going to have very powerful multi-core low power chips on laptops and desktops.
Intel supplies chipsets as well, and their chipsets are pretty nice. They're not always the best, but they're usually close and they're almost always better in laptops. Having the chipset provided by Intel cuts down on engineering costs as well, which is important for Apple. Their volumes are small by most OEM and motherboard maker standards.
I rarely criticize things I don't care about.
IBM Journal of R&D has a special on Blue gene. From the article which has details about the processing node in Blue Gene.
The BLC ASIC that forms the heart of a BG/L node is a SoC built with the IBM Cu-11 (130-nm CMOS) process. Integrating all of the functions of a computer into a single ASIC results in dramatic size and power reductions for the node. In a supercomputer, this can be further leveraged to increase node density, thereby improving the overall cost/performance for the machine. The BG/L node incorporates many functions into the BLC ASIC. These include two IBM PowerPC 440 (PPC440) embedded processing cores, a floating-point core for each processor, embedded DRAM, an integrated external DDR memory controller, a Gigabit Ethernet adapter, and all of the collective and torus network cut-through buffers and control. The same BLC ASIC is used for both compute nodes and I/O nodes, but only I/O nodes utilize the Gigabit Ethernet for host and file system connectivity. The two PPC440s are fully symmetric in terms of their design, performance, and access to all chip resources. There are no hardware impediments to fully utilizing both processors for applications that have simple message- passing requirements, such as those with a large compute- to-I/O ratio or those with predominantly nearest- neighbor communication.
Pentium Ms perform a bit better than G5 chips at similar clock speeds, and Pentium Ms scale to much higher clock speeds than these lower power FX chips.
The new FX chips would probably be a welcome replacement for G4s if only to replace the archaic bus (though I doubt Apple will bother), but they're not good enough to replace Intel's current laptop chips, much less the future chips.
Intel is releasing Yonah-core Pentium Ms early next year. They're going to address the Pentium M's floating point weaknesses, increase clock speed, lower power usage, and there will be dual-core versions within the same power budget as current chips. I don't think a single G5 at 1.6 ghz would do too well against a dual-core 2 ghz Pentium M.
Basically, IBM is releasing chips that fit into the lower end of current laptop chips a few months before Intel releases the next generation.
I rarely criticize things I don't care about.
I think you meant to say the Pentium D + LaGrande (DRM in silicon), not Pentium M. The Pentium D (with not-yet-released updates and fixes), does exactly what Apple is after - controlled access to media with an architecture that provides lower-power (iPod-like devices and battery-powered Powerbooks).
The 3GHz promise is completely irrelevant to the Intel switch. In the last two years, IBM has gone from 2.0 to 2.7GHz, which a proportionally larger increase than Intel going from 3.0 to 3.8. Everybody ran into the same problems at 90nm; it's not a case of IBM dropping the ball. The real motivation is laptop chips, where the Pentium-M trounces the G4 today, and Yonah will easily beat a 970FX at 1.6GHz.
How to solve most of our problems: 1.Lots of nuclear plants. 2.Cure aging.
Pentium D is still a ~100W chip. Look at Yonah instead.
Lower speed Pentium Ms also consume less power. I believe 2.13 ghz is currently the fastest Pentium M.
Besides, Apple isn't going to use these things. They'd have to redesign PowerBook chipsets and motherboards for a computer that, at best, they'd be selling for less than 2 years. It's much more likely that they'll transition all current G4 computers to Pentium Ms first.
These just-announced FX chips compete with the lower end of Intel laptop chips, while the Yonah-core chips Intel is releasing early next year will improve in every way. Higher clock speeds, lower power consumption, and dual-core versions that fit within current power envolopes.
You might prefer a 1.6 ghz G5 to a 2 ghz Pentium M, but would you prefer it to a dual-core 2 ghz Pentium M that takes the same amount of power?
I rarely criticize things I don't care about.
I think you meant to say the Pentium D + LaGrande (DRM in silicon), not Pentium M. The Pentium D (with not-yet-released updates and fixes), does exactly what Apple is after - controlled access to media with an architecture that provides lower-power (iPod-like devices and battery-powered Powerbooks).
Not sure. I'm basing that off of this article from The Register. I don't know if they plan on keeping hte Pentium D in the lineup (or even moving to the Pentium E if it ever comes about), but it still (as another poster has mentioned) consumes too much power. Apple wants lower power processors, probably under 50W per core (just a guess).
Here is a link on the Pentium M roadmap.
As listed below (and speculated for the Macs):
4Q 06-1Q 07:
Merom: A dual-core Pentium M (Banias) successor
Conroe: A 64-bit desktop version of Merom (see comments above about Conroe).
A 64-Bit dual core Merom is just what apple needs to be the successor to the 64-bit dual core G5s. And, surprise suprise, it is due out just when Jobs said the transition would occur/finish. It is also more than likely going to be fairly low power as it is in the Pentium M lineup. We won't know till it comes out if it is as low power as these G5s, but it should be lower power than the current high end P4s. The guy also speculates on why Intel over AMD on the next page of the article.
Fly me to the moon Let me sing among those stars Let me see what spring is like On jupiter and mars
Basically, the DRM is not intended to protect user-created content, it's intended to prevent that content from ever being distributed in the first place.
The conglomerates make a healthy profit by being the gatekeepers. For decades now, they've done that by owning the means of production and distribution, but that monopoly is disappearing.
Now they want to hang on to that monopoly for a while longer, and they're trying to do that by mandating DRM. Publishers will need to license the private encryption keys needed to publish content that will play back on consumer devices, and hardware makers will need to license the means of playback as well.
The piracy issue is just a smokescreen. It's a shame the crowd here is too busy defending their "right" to "share" the latest Metallica CD to even notice that they're playing right into the *AA's greedy little hands. (They're also too busy to notice that all of the songwriting talent in Metallica died in a bus crash a long time ago... but that's another story.)
I've said it before, and I'll keep saying it: Geeks should *support* DRM, embrace and extend it to ensure that it protects everyone's rights, not just the *AA's. DRM that was about mechanism, not policy, would not necessarily be a bad thing.
Lost: Sig, white with black letters. No collar. Reward if found!
Newsflash: AMD's fab is... Wait for it... IBM!
Been there, evaluated that.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
IBM has since released an English press release, available here.
This should be significantly more informative than the earlier available Japanese documents.
It's the older Banias core that draws 27W. The newer 765, based on the Dothan core, draws 21W at 2.1GHz - a 33% increase in clock speed at 24% more power when compared to the 1.6GHz 970FX. Or if you're most interesting in power, how about the 758? That draws 10W at 1.5GHz - 37% power savings over the 1.6GHz 970FX with only a 6% drop in clock speed.
On top of that, the Pentium M outperforms the 970FX core clock for clock by most metrics.
- Apple's sales, along with the rest of the industry, are growing most rapidly in notebooks
- Neither IBM or Freescale has a dedicated notebook division. Intel does.
- Intel is best able to meet Apple's supply demands.
Make sense?Remember that Apple stated the transition would take until 2007. That gives IBM and Freescale plenty of time to come out with a PowerPC that can justify Apple keeping a PowerPC offering in the lineup. The Intel move was mainly about laptop performance. The dual G5 2.7GHz with a 1.35GHz FSB is still capable of outperforming Intel CPUs on vector work (digital imaging, video, audio...Apple's core markets). And now that IBM has dual core CPUs, it looks like the G5 still has lots to offer. I imagine a 1.35GHz FSB would make a huge performance difference on a dual core G5 when compared to Intel's current dual core offering. Anyway, Apple kept an up-to-date version of OS X on x86 for 5 years. Why do you think they plan on completely abandoning PowerPC? That just gives them less opportunities in the future.
You're 100% right.
:)
One of the things we do at the company I work for is tell people the G4 is better
than the G5. The G4 is wonderfully more generic in performance - random memory
access is a good one to benchmark. The G5 is very good at streaming huge
contiguous blocks, but the high RAM access latencies and cache latency/line width
problems kill random access or impact code such as array lookups (best Vector
Permute trick on the planet, also hampered by a weak Permute unit).
But that's not to say the G5 doesn't have merits; it just has some VERY specific
applications that it's very good at. Perhaps too specific for Apple. Companies
like Mercury (www.mc.com) would probably have gone for the G5 if they hadn't
found an even more specific processor for their needs (Cell, in this case).
With lower power chips the G5 could actually start to replace the G4 in places
where performance in high memory and streaming data are paramount.
For laptops, desktops, and places where we don't need 16GB of memory, the G4 is
going to rock for years to come though. I actually wonder why there couldn't be
a special "pseudo-64bit" version of Linux for the G4, which used the 36-bit
addressing modes to implement high memory support. Maybe it's because IBM practically own ppc64 Linux and don't want to overshadow their own chips?
-- Neko
The 970FX already matches (at 1.6GHz) the power envelope of the 1.6GHz G4 that
:)
Apple are using in their PowerBook.
Apple's "we can't because it's too hot" bleating is yet another example of Jobs'
bullshit.
IBM's PowerTune beats the pants off of the DFS functionality in the 7447A/7447B
Apple are using. They could drop power usage to tiny levels and keep battery
life at their usual high values.
The real reason might be simpler and slightly more technical; clock for clock the
G4 would outperform their G5 version in everything except a memory bandwidth shoot-
out. They had a benchmark right there on the show floor at Motorola SNDF (Dallas),
in April 2004. You would expect it from the figures but to see it in real life..
it's quite enlightening. You could bet that Apple could gloss over this though,
like they gloss over everything else important (like the useful Mac Mini with
wireless and bluetooth being $700, Mac Mini's performance sucking cock wrt hard
disk and peripherals, the hard disk cooling problems in the iMac, iBook logic
board blowouts..
-- Neko
I'm not an expert, but I don't think that Apple will give up on the PowerPC processors for a long time. It seems that Apple is pushing the move towards Universal Binarys, not Intel. While IBM Powers are electricity sucking monsters, they are still tops when it comes to performance in servers and applications that require lots of vector processing (the multi-media creation tools that has kept Apple alive in the content industry). Universal Binaries gives Apple the choice of using a processor that meets the machines requirements, for laptops they could choose a Pentium M, for the iMac they could use a Pentium D (does dual cores mean SMP, or does it use an Intel proprietary technology like a sort of HyperThreading to schedule tasks?), for xServes and PowerMacs they could stick to the PowerPC.
Either way, Apple would be a HUGE account for AMD, and would require a substantial portion of AMD's manufacturing resources.
How would 15% of thier total production correlate to a substantial portion of their manufacturing resources? What the hell is the other 85%? A super duper substantial amount? I'd assume 15% of the production would require 15% of the manufacturing resources. Wouldn't you?
On a side note....
I think everyone underestimates the amount of growth AMD has had in the last year. Apple would appaer to be a drop in the bucket compared to the others that have started using AMD chips.
Here are some quotes from various sources:
Partners sure seem to be betting on AMD's chips. IBM, HP and Fujitsu-Siemens began selling AMD64 processor-based systems in the fourth quarter and Sun Microsystems will deliver AMD Opteron processor-based enterprise servers in the first half of this year.
Customers are also asking for more AMD Athlon 64 and AMD Opteron processors, the company said. AMD said in the last three months it gained new global customers such as Daimler Chrysler, QUALCOMM, Pirelli and Bristol-Myers Squibb.
Another one from the AMD Q1 2005 Earnings Conference Call Notes:
Processor business not just growing, it is accelerating.
CPU revenue up 31% year on year, AMD64 CPU sales doubled from a year ago
growth rate higher than overall CPU rate in 2004
believe we have grown faster than then total market in Q1
Sun launched 2nd gen v20z and v40z servers featuring 252 and 852.
HP added new proliant models including servers, blades, and workstations
Adding new enterprise customers including Met life of Mexico and Lucasfilm
50% fortune 500 our customers
63 of CPUs are AMD64, by end of 05 100%
90nm is ahead of schedule with better than expected yields which results in more capacity for demand
production planned for Fab 36 first half of 2006
dual core systems on existing Sun cray and HP workstations and servers
we are ahead of schedule on dual core, have been shipping to partners since january
Athlon 64 dual core for desktops and notebooks
we are expanding enterprise footprint with turion 64 for thin and light notebooks
xp pro 64 bit announcement has thrilled us. believe it to be outstanding and robust. like to thank microsoft.
AMD operates their own CPU fab in Dresden, Germany. AFAIK IBM has no direct role in the fabrication of K8-based processors.
AMD and IBM do work together on developing fabrication technology. But AMD is not fabless nor totally dependent on IBM for manufacturing.
g5 is to g4 as p4 was to p3 better overall IPC, less picky about memory latency, less power, basically a great thing to quad core if you're looking for perf/watt
I refuse to believe that the 28- and 31-stage Pentium 4 pipelines are a better thing than the 10-stage pipeline in the Pentium III, particularly when we're talking about IPC. Do you remember the fuss made about P4 being slower at the same clock speed than the PIII? That's proof it has worse IPC rate.
Neither the P4 or the G5 are lower-power than their predecessors and they fail to provide better performance/watt, in any configuration. This is why the P3 architecture has been adopted into the Pentium M line for low power use and the G4 processors remain the chips used at the core of Apple's iBooks and PowerBooks.
The great thing to do with the Pentium 4 architecture would be to put in on good Strained Silicon and SoI processes to push it above the 4.0GHz clockrate at which it is believed to be a very strong chip.
The differences between the G4 and G5 chips are what happens when you move from a desktop computer chip to a cut-down Big Iron chip (IBM's POWER4, IIRC). The G5's are inherently 64-bit capable in a way that the first three generations (Willamette, Northwood and Prescott) of the Pentium 4 are not, although there exist Prescott-based Pentium 4 processors with Intel's EM64T implementation.
BTW: http://arstechnica.com/ is your friend. Hannibal has done a good job of talking through the history of the Pentium chip family (1 & 2) and the PowerPC family (1 & 2, part 3 hasn't yet arrived) up to the G4's. There's discussion of the IBM POWER5 architecture too, and some commentary on pipelines in processor design (1 & 2). I learned a lot from these, and value their information. But I'm going to stop telling Granny to suck eggs now.
Geez,
Who cares if they make dual core turtles?
If they could just peak it up to 3.0 GHz, a Dual-Core powered,
Dual CPU Apple G5 Power Mac would be Sweet!
here already. I subscribe to InfoWorld, and this article discusses available systems from IBM using the dual-core Power5.
antipaucity