Understanding 64-bit PowerPC architecture

An anonymous reader writes "Each of the leading microprocessor manufacturers has announced the availability of one or more 64-bit desktop processors, but differences exist in architectural design, fabrication, support, and intended use of each processor. This article looks at the critical issues in a few of IBM's 64-bit POWER designs, covering 32-bit compatibility, power management, processor bus design, and the manufacturing process."

DUPE!!!

[Anonymous Coward]

It's a dupe.

Re:DUPE!!!

[LurkerXXX]

Slashdot, where you get all the editorial prowess you pay for.

Unless your a sucker who subscribes. Then your getting rooked!

Re:DUPE!!!

[SilentChris]

Not only is it a dupe, and it's at the same time on the front page, but it's within half a dozen articles on my screen. Is that a record? This must've been resubmitted as a joke to test the editors.

Power != PowerPC

[Computerguy5]

Power != PowerPC That is all.

Re:Power != PowerPC

[Arker]

Power != PowerPC That is all.

Hmm not exactly. Power > PowerPC. PPC is a subset of Power, a point which TFA does mention, and explain a bit.

The PowerPC architecture that was born of this partnership is -- and always was -- a 64-bit architecture derived from the IBM POWER architecture.[...]
Note that the performance of the PowerPC 970 family actually exceeds that of its award-winning parent, the high-end IBM POWER4 processor, in many areas. This is due to the fact that the circuit and process technology used for the POWER4 processor was designed to achieve levels of reliability necessary for the continuous availability server market -- levels that can be relaxed for the desktop and small-scale server market -- at the expense of transistor switching speed. Thus, the fabrication technology used for the PowerPC 970 was designed to eke out higher performance by trading away reliability; for these markets, the trade-off between reliability and performance is different.

And yes, folks, it is a dupe. [slashdot.org] And a very recent one too. At least this time they got it in two different sections, first Apple, then Hardware. I'd have to say that Hardware is a better place for it, it's definately NOT just Apple that uses these chips.

Re:Power != PowerPC

[Billly Gates]

Not to mention AIX runs unmoddified on IBM's blade powerpc970 servers.

If power were totally different, AIX would not run at all.

Re:Power != PowerPC

[kompiluj]

You are not 100% accurate. PowerPC is a subset of POWER + Multivec Engine (Single Instruction Multiple Data) by Motorola. This makes PowerPC better for non-server use.

Re:Power != PowerPC

[m0rph3us0]

The Power processors implement the PowerPC instruction set.

What you are saying is like saying Pentium 4 != x86

You can run OS X on POWER processors through Mac on Linux. And the PPC 970FX is a derivative of the POWER4 processor.

Basically, the PPC 970 is a POWER 4 with a better SIMD unit.

Re:Power != PowerPC

[dgatwood]

At a user-mode instruction level, the POWER3 and POWER4 CPUs implement the PowerPC ISA. At the supervisor level, this isn't true, and for the POWER and POWER2, it wasn't true at the user level, either. (POWER and POWER2 had additional registers, for one thing.)

Technically, the POWER series implement an ISA that is user-space-instruction-compatible with PowerPC. Subtle difference, I know, but....

Re:Power != PowerPC

[Darksiderstlouisafte]

No a PPC 970/fx is a POWER4 with one CPU core removed, SIMD unit add (at apples request), meny POWER instruction removed and fitted with PowerPC ISA that is all there is to it Darksider

What's going on?

[Moby Cock]

Isn't this the same story as five stories ago? Am I missing something?

Re:What's going on?

[vivek7006]

No you are not missing anything. It is fucking CmdrTaco who is missing half his brain.

Seriously almost every dupe story is invariably posted by CmdrTaco

Re:What's going on?

[Wapiti-eater]

Bad Dupes

And in light of what gets rejected, makes one wonder what this place is really all about anymore.

Re:What's going on?

[The Bender]

No, no, no. You fool, do you not pay attention? The other story was in the "Apple" section, and this one's in "Developers".

Completely different.

Re:What's going on?

[Minwee]

Remember, Commander Taco has decreased the dupe ration from from 20 grams per day to 30 grams.

Re:What's going on?

[acaeti]

Somebody must be having a case of the Mondays...

Third Apple story today

[Jucius Maximus]

I, for one, welcome our new 64-bit overlords!

First Dupe!

[ggvaidya]

Right here [slashdot.org]. And why yes, the dupe is still on the front page, why do you ask?

Re:First Dupe!

[ggvaidya]

Interesting look at Slashdot categorization. Apart from "Hardware", the two stories don't share a single category! This one is marked Programming (?!) and Hardware, while the other got Hardware, Technology, Technology (Apple) and IT.

Ehh?

[RAMMS+EIN]

``64 bit PowerPC architecture...desktop...POWER''

but wait...I thought PowerPC and POWER are similar, but not identical, and that PowerPC was aimed at the desktop, whereas POWER is more for servers. Do I have it wrong?

Re:Ehh?

[temojen]

Same architecture, different implementations.

Re:Ehh?

[Ohreally_factor]

and both stories are linked to the same article.

Re:Ehh?

[mrdisco99]

Yes.

However, with this iteration, IBM took one of the cores from the dual-core POWER4 chip, repackaged it as the PPC970, and sold it to Apple as the G5. So PowerPC and POWER have re-merged... sort of. Freescale is still developing their own PowerPC chips which do not fall under the POWER umbrella.

Re:Ehh?

[Ohreally_factor]

However, with this iteration, IBM took one of the cores from the dual-core POWER4 chip, repackaged it as the PPC970, and sold it to Apple as the G5.

Don't forget they also incorporated a simd (alti-vec).

Freescale is still developing their own PowerPC chips which do not fall under the POWER umbrella.

One of which is a dual core [theregister.co.uk] G4. Not sure if this qualifies as an irony. Maybe a bronzey.

Wow

[Algorithm wrangler]

That is the first time I've seen a dupe at the same time as the original [slashdot.org] is on the front page. Wow again.

Re:Wow

[pclminion]

Really? I've seen that many times.

Once, I even saw a TRUPE on the front page -- three fucking stories, all the same. Does anyone remember what that story was?

Re:Wow

[elzbal]

I think they did this on April Fools day one year - one particular story went up 5 or 6 times...

Trupe?

[douglips]

The correct word is TRIPE.

Re:Trupe?

[russint]

As in threesome?

Re:Trupe?

[drinkypoo]

Yak! Dog food!

Of course, tripe is cow stomach, right...?

Re:Trupe? How about Tripe?

[pclminion]

Except that tripe is already a word with an entirely different meaning.

Plus, "tripe" doesn't even have the right sound either. The appropriate word would be "trip," which is also already taken. And we can't use Tripp, due to the reference to the Clinton fiasco.

Thus, "trupe" is the most appropriate word in this case, by analogy with "dupe." Yes, I actually pondered this before posting.

These are the important questions of our lives.

Re:Trupe? How about Tripe?

[AndroSyn]

I think the entirely different meaning was intentional.

Re:Trupe? How about Tripe?

[BoomerSooner]

Tripe is delicious. Fried with lots of A-1 Steak Sauce.

(Kidding, I cannot believe I've even tried the stuff!)

That's "trip"

[Kourino]

That would be a trip. Remember, dupe is short for duplicate. The correct abbreviation for three stories would be "trip", which is short for "triplicate" (which is a real word, yes).

The real question is why I'm debating the correct abbreviations for Slashdot editorial errors on Slashdot ...

Re:That's "trip"

[tenton]

which is short for "triplicate" (which is a real word, yes).

People who don't know triplicate isn't a real word hasn't filled out enough forms. :D

Let's have some consistancy! :)

[F'Nok]

That would be a trip. Remember, dupe is short for duplicate. The correct abbreviation for three stories would be "trip", which is short for "triplicate"

Yes, but if you use the same transformation...

duplicate -> dupe
triplicate -> tripe

I'd argue that tripe is actually well fitting in that pure sense. That fact that it has pun value is an added bonus! :)

Re:Let's have some consistancy! :)

[pclminion]

Right, but in the case of "tripe," adding the final 'e' actually changes the pronunciation. I actually gave thought to this before deciding to go with "trupe," which at least has the benefit of rhyming with "dupe."

I'm very, very sad. I know. As are we all, in this thread :-)

Re:Let's have some consistancy! :)

[F'Nok]

Ahhh, if only we had a phonetic language...

Then this decision would be so much simpler ;)

This has to be..

[creep]

The fastest dupe I've ever seen.

Re:This has to be..

[_PimpDaddy7_]

That's the speed of 64-bit architecture at work, baby...

Re:This has to be..

[Drakonian]

No, there were back to back ones once. Not just during April 1st either. It was amazing. Truly a day that will live in infamy.

wow!

[eshefer]

a dupe posted by CmdrTaco to a story posted by Hemos.. It feels like 1998 again :-)

Re:wow!

[Lumpy]

It feels like 1998 again :-)

shut up!
You might make Jon Katz come back.

Re:wow!

[Anonymous Coward]

He went off to write about dogs. No, seriously.

http://www.workingdogweb.com/Katz.htm

Re:wow!

[Lars T.]

Katz on dogs?

Grammar Nazi!

[Anonymous Coward]

their means belonging to them.
they're means "they are"
there means not here.

Super duper

[Sabu mark]

Let's see if they go for the record and post this story a third time before the first one goes off the front page.

Re:Super duper

[jon787]

Yeah but that was on april fools' day and doesn't count. They were doing that on purpose.

link still loadable

[tomcio]

i think they did that because the other link is still loadable
unlike the watercooling article

OK!

[Bombcar]

Admit it! Some of you send in stories that are on the front page just to see if they get duped, right?

I'm on to you!

Re:OK!

[Anonymous Coward]

Ummm, yeah.. ..... It's called Q&A. Perhaps more of that should go on to keep the site in check?

Re:OK!

[daeley]

Try submitting an "Ask Slashdot" with that question and see what happens.

If it gets posted twice...well, let's just say this thread will be the ultimate power in the universe. ;)

1) . . .

[Ohreally_factor]

If it gets posted twice...well, let's just say this thread will be the ultimate power in the universe.

Threads are mammals?

Here are some other interesting stories

[Anonymous Coward]

Here are some other interesting stories

http://www-106.ibm.com/developerworks/library/pa-m icrodesign/?ca=dgr-lnxw01fkjfdsa [ibm.com]
http://www-106.ibm.com/developerworks/library/pa-m icrodesign/?ca=dgr-lnxw01dhifads [ibm.com]
http://www-106.ibm.com/developerworks/library/pa-m icrodesign/?ca=dgr-lnxw01jj5324543 [ibm.com]
http://www-106.ibm.com/developerworks/library/pa-m icrodesign/?ca=dgr-lnxw01hdshdgh [ibm.com]
http://www-106.ibm.com/developerworks/library/pa-m icrodesign/?ca=dgr-lnxw01jesusheadfuck [ibm.com]

Nondiscrimination policy

[JLavezzo]

Slashdot's nondiscrimination policy prevents them from excluding dupes [slashdot.org] on the basis of topic of origin.

The explination of the difference

[TyrranzzX]

First, I'm talking a risc/cisc architecture like the x86.

When you're talking about 64 addressing lines, your talking about addressing a fscksum of memory and devices. But, in addition, those lines allow for other possibilities: for example, sending 2 or 3 write commands with attached data and 2 32 bit addresses on the 64 bit bus simultaniously with an extra address decoder either on the chip or on the memory controller, or to some other device. Although, I don't know weither or not they've thrown that in as of yet. 64 bit numbers don't occur that often, afaik, but I'm not a coder so :P.

The data bus advantage, however, is bigger. The x86 architecture has a command decoder, whereas you can send several commands in a single clock. With 32 more bytes, you get twice as many commands in a clock. Additionally, you can address more commands (but seriously, the first x86 had 38 commands, and that has increased by 10x in the past few years).

Aside from that, you're throwing on more features into the processor. But, that's been here in the past 20 years of processor developement anyway. The article tends to be unclear on this. You're essentially expanding the bus to feed more buffers/pipelines.

Re:The explination of the difference

[TyrranzzX]

The input is CISC, the execution is RISC. 4 or 5 commands may come in at any given point, the execution unit does only 1 command at a point, per pipeline, anyway. I forget which command, but one command for the 8086 required 83 clock cycles to complete.

Although, I will admit, the newer processors are getting more and more complex, and the command sets are more cisc-ish in nature. I dunno, I guess it depends on the technology; if a logical 8 bit AND is your idea of a risc function, then by all means, t

Re:The explination of the difference

[mikefe]

"I forget which command, but one command for the 8086 required 83 clock cycles to complete."

So what's changed? You can get that with a Pentium 4...

Daja vu? Or is it just me?

[cyberwinds]

I saw the same black cat walking by a moment ago.

Re:Daja vu? Or is it just me?

[Stevyn]

and I just saw someone misspell a subject line a few minutes ago, must be deja vu all over again.

Is this page cached.

[acomj]

I did my afternoon slashdot break. I'm hitting "reloading" thinking, my company must be caching the pages now or something... No new stories. Whats going on.

and on and on and on....

Re:Gets more Apple fanboy visitors

[Warlock7]

Which is why you chose to use the AC name, so they don't mod your ignorant ass down...

Re:At least you admit it.

[Warlock7]

Just like the Linux fanboys that have modding capability.

Mod the stories, not the comments

[Wapiti-eater]

What we realy need is a method to moderate the stories themselves - not just the commnets. Can ya see it now? "-5 Dupe"

Some of the stories that do get accepted, it's stunning what gets front page exposure. While other stories that get rejected - you know, real news that nerds could use...

It's time to extend the voice of the community

double exposure

[Blitzenn]

no. It's not a dupe, it's the 64 bit bus sending the same 32 bit information, doubled up along the bus. It's just looks like it's posted twice. It's much more efficient this way. Trust us!

no. It's not a dupe, it's the 64 bit bus sending the same 32 bit information, doubled up along the bus. It's just looks like it's posted twice. It's much more efficient this way. Trust us!

It's NOT a dupe !

[raulfragoso]

You guys sux on mathematics: if it was a dupe, it would be an article talking about 128-bit stuff, you fools !

dupe detector?

[zygote]

In the abscence of institutional memory or /. editors sitting in the same room :-) , could Slashcode be tweaked to scan for URLs that are identical in stories and flag them somehow?

Some Perl script or such that looks at the post about to hit the front page, then looks at a say the last three or four posts that were on the front page. If it finds an identical or closely identical link, it then sends an email of the body of the two posts to the editor.

I don't think it would get the server's load up too much.

(much sarcasm, minus 1, bad dog.)

Re:dupe detector?

[BJH]

Let me put it this way:

The idea of a dupe detector is proposed every time Taco reposts a story, but it hasn't been implemented yet. If it hasn't happened in 6+ years, it's never going to happen.

Jesus hopping cabbage.

[b1scuit]

Jesus Hopping Cabbage. Does anyone else see the irony concerning twenty nearly identical posts saying "this is a dupe! What has slashdot become!"?

Anyone?

Re:Jesus hopping cabbage.

[awehttam]

Who cares!

Timothy, breaking records

[theolein]

Timothy, you're really the only editor on Slashdot who manages such an amazing amount of dupes. And you're also the only one that manges to post dupes of articles that are still on the front page.

I heard the Guiness Book of Records called. :D

Re:Timothy, breaking records

[Ohreally_factor]

It wasn't Timmy [yimg.com] this time. It was Taco [usatoday.com].