Talk:IBM POWER

Rename proposed to "IBM POWER architecture"

The following discussion is an archived debate of the proposal. Please do not modify it. Subsequent comments should be made in a new section on the talk page. No further edits should be made to this section.

The result of the debate was no consensus. -- tariqabjotu (joturner) 02:13, 6 August 2006 (UTC)

Requested move

• IBM POWER → IBM POWER Architecture … Rationale: "IBM POWER" makes no sense when the suject is "POWER Architecture" … Please share your opinion at Talk:IBM POWER. —Power2ThePeople 21:43, 31 July 2006 (UTC)

---

Survey

Add *Support or *Oppose followed by an optional one-sentence explanation, then sign your opinion with ~~~~

• Support maybe if the title were corrected the gross inaccuacies might disappear. having "POWER" in all caps is the way its been printed since 1990, yet because of typo's on the internet "Power" is easily typed so now there's ambiguity were there should be none. —The preceding unsigned comment was added by Power2ThePeople (talk • contribs) 14:43, 31 July 2006 (UTC).
• Oppose, given that Power.org explicitly tell you not to call it "POWER Architecture" (see my comments, and references, below). Rename to "Power Architecture" (not all-caps), maybe, although the questions then would be 1) how much of the stuff here would belong here and how much would belong on a page specifically talking about IBM's RIOS/RSC/POWERn processors and 2) what relationship should there be between this page and the PowerPC page? Guy Harris 22:41, 31 July 2006 (UTC)
• Oppose, sice I wanted to rename this page "POWER ISA" and add a new page called "Power Arcitechture" as I descibe below. -- Henriok 23:34, 31 July 2006 (UTC)
• Rename to something. IBM POWER is not a good name. A rename to IBM Power Architecture would be better while any necessary changes are made. Vegaswikian 02:12, 1 August 2006 (UTC)

The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made in a new section on this talk page. No further edits should be made to this section.

Discussion

• Oppose, I would like to recant my previous support for a name change to "IBM POWER Architecture"; Yet personally i do like "POWER Architecture". Thank you.Power2ThePeople

Add any additional comments

So what is the subject here? The IBM processors that have POWER (in all caps) in their names (as well as the original RS/6000 processor and the RSC single-chip version)? The instruction set architecture, with the all-caps acronym POWER, that the first two of them implemented (apparently all the later ones implemented the 64-bit PowerPC instruction set architecture)? The "Power Architecture", not all caps, that IBM and others are touting now, and which appears to be, at the instruction-set level, the successor to PowerPC 2.02? Or all of the above? Guy Harris 22:11, 31 July 2006 (UTC)

The subject is the evolution of the "POWER Architecture" from a historical point of view. Not as if this where a news paper and every day the article should change. POWER IS A SUPER SET OF (Power or PowerPC or CEll or Power 64 AS or PowerPC 64 AS} and has allways been such, if PowerPC adopts a 64-bit Flavor where do you think this design or "ISA" were tested hint{POWER3,RS64,etc}. POWER HAS ALWAYS BEEN A 64-ARCHITECTURE from its inception in The "AMERICA PROJECT". circa-1990 POWER{IS THE MISSION CRITICAL aspect of the ARCH} PowerPC{embedded market,Cheap PC Yet more reliable than x86}. Belatrix project was to Merge all three{OS/360,OS400,AIX} of IBM'S ARCH'S INTO ONE, by 1995 they couldnt do but they did not give up. This is POWER6. so the 64-bit architecture which has been refined in over 7-generation of cpu, is given to the average consumer as {Power or Power64 or POWER64 or PowerPC 64 or PowerPC AS or CELL or CELL2} there all a subset of "The POWER Architecture"{were POWER is the superSET}. The work done by IBM/Sony/Toshiba will find its way into POWER6(you think INTEL CoreCrap/Itanum is fast, well POWER6 will substantially raise the bar). Only IBM ADD'S TO POWERn, yet what appears to be a new 64-bit architecture "Power" will be modified by comity but never allowed to supersede POWERn. BECAUSE IT IS DEFINED AS SUCH.... -- —The preceding unsigned comment was added by 66.27.52.158 (talk • contribs) .

Aside from the inane babble (OS/360, OS/400, and AIX are not and were never architectures; they are OSes that ran on S/360; IMPI and later POWER; and S/370, S/390, and also later POWER) and blind praise (it's foolish to compare POWER5/6 to either a general purpose desktop microarchitecture or the failed IA-64 implementation), please work on your prose a little bit. I have a difficult time interpreting what you intend to convey. -- uberpenguin @ 2006-07-31 23:29Z

ouch! the guy who always starts a rebuttal by bashing one charactor, yet lacks commone sense.Power2ThePeople 17:59, 31 July 2006 (UTC)

That wasn't what I'd term a "rebuttal". More like a plea to produce comprehensible commentary. -- uberpenguin @ 2006-08-01 02:02Z

I'm not sure what "POWER HAS ALWAYS BEEN A 64-ARCHITECTURE from its inception" means; perhaps they'd designed the 64-bit part of the ISA then, but the first implementation in the first RS/6000 didn't support 64-bit registers or (non-segmented) pointers bigger than 32 bits. Guy Harris 23:41, 31 July 2006 (UTC)

The First implementation of the "POWER ARCH" consisted of MORE THAN 8 dedicated processors, and IBM's competitors laughed at POWER1/? complexity. Also this was during a time when 64-bit software was a rarity, so to reduce the allready complex POWER1 64-BIT SUPPORT WAS DELAYED. Power2ThePeople 17:59, 31 July 2006 (UTC)

the "POWER ISA" IS CLASSIFIED/CONFIDENTIAL OR AT LEAST MOST OF IT, so any effort to create an article "POWER ISA" WOULD BE futile because you would have to wait at least 10 years for pertinent information. its taken over 15 years for IBM TO release information to the general public about OpenPOWER(POWER4,POWER5,POWER3,POWER2) —The preceding unsigned comment was added by Power2ThePeople (talk • contribs) 17:02, 31 July 2006 (UTC).

"Most of it"? By that do you mean "the PowerAS stuff"? Or is there stuff other than the PowerAS stuff which isn't in any of the PowerPC specifications? Guy Harris 00:10, 1 August 2006 (UTC)

"POWERn evolves with the progression of technological advances. The FBI/NSA/CIA/IRS...ETC use zOS/system360 and the internals of these systems and information about them are trade secrets/highly classified Power2ThePeople

That's a non-answer. "POWERn evolves with the progression of technological advances." has nothing to say about what "most of it" refers to, "The FBI/NSA/CIA/IRS...ETC use zOS/system360" might be true (although you don't mean "system360", given that z/OS requires S/360's successor, the z/Architecture), but the mere fact that government agencies use a particular technology doesn't mean that the technology is necessarily a Deep Dark Secret, and "the internals of these systems and information about them are trade secrets/highly classified" is true only of some of the information - a lot of the instruction set is documented in your friend the Principles of Operation, just as a lot of the PowerPC instruction set is documented in the PowerPC Architecture Book, Version 2.02 (of which the Power ISA version 2.03 is presumably the successor). Guy Harris 00:41, 1 August 2006 (UTC)

Nice job! Guy Harris, for some of the information below. I agree that the name should be something likened to say "POWER Architecture" with the absence of "IBM" in the title, so as to encompass the entire spectrum of work applied to the "Architecture" in a top down approach. In a few months IBM's going to allow Freescale to have equal voting rights in the decision processes of future changes to the to the "Power Architecture", yet IBM will still maintain an independent development of {POWER6,POWER7}. So by having a title called "Power Architecture" and including any reference to POWER6/POWER7 the best you could do would be to imply theirs a connection between "Power" and "POWER" which would add more ambiguity, furthermore too exclude {POWER5/POWER4/POWER3/POWER2/POWER1} from an article title "Power Architecture" would surly be an act of heresy. So my reasoning for the all caps "POWER" is to maintain the the heritage, and to conjoin "Power" with "POWER". Which would allow and foster a logical interpretation of Past or Future events. i'll restate {POWER-is from 1990-POWER7),{Power-is from 2004-?). And to answer your question Guy, I Know it say not to use all caps, but when did WikiWriters start to care about Integrity, if at a later time IBM decides not to do a POWER8, then i would think and only after the article has matured past 50% halve truths that the ambiguity could be contained. If it it is not history for which we are trying to preserve, then we have learned nothing insofar as today.Power2ThePeople

You've alluded a few times to the wreched state of this article without citing any concrete examples. Why not try to fix some of the problems you perceive rather than complaining about them? -- uberpenguin @ 2006-08-01 22:36Z

It appears from this interview with an IBM marketer that the idea is that"

• the entire family of processors with instruction sets derived from the original RS/6000 processor's instruction set is the Power Architecture family (not all caps);
• "PowerPC" is going to be used as "a product-level brand", which sounds as if it means it's a brand name used for certain products, i.e. for certain processors that implement the Power instruction-set architecture, with the new Power instruction set architecture being a superset of the PowerPC instruction set architecture (or architectures) and superceding it.

The Power.org Brand System FAQ says, among other things, that:

• "The Power Architecture name should never appear in all upper-case lettering (e.g., POWER Architecture) or lower-case lettering (e.g. power architecture)."
• "The PowerPC wordmark, as referencing the PowerPC instruction set architecture (ISA) and products based on that ISA, will continue to exist. However, we encourage the ecosystem to start leveraging the new Power Architecture brand system, including the new identity, to help establish a more consistent voice in the marketplace for the Power Architecture platform."

and the page for The Power Architecture Brand Heritage(TM) says:

• "What began in 1990 as the Performance Optimization With Enhanced RISC (POWER(TM)) has evolved into the Power Architecture(TM) platform today, the foundation for some of the world's most pervasive applications in fields as varied as automotive control, wireless infrastructure, enterprise severs and home entertainment systems.", so the Power Architecture is a descendant of POWER;
• "In conjunction with the architectural transformation, a brand metamorphosis is occurring. The new Power Architecture brand system builds on the significant heritage of PowerPC and other Power brands before it.", so PowerPC also contributed to the Power Architecture (PowerPC being, of course, a descendant of POWER).

All this marketing noise appears to suggest that there should be a page called "Power Architecture" - not "POWER Architecture", as they quite explicitly say "The Power Architecture name should never appear in all upper-case lettering (e.g., POWER Architecture)" - which should, at minimum, describe the current state of the Power Architecture. I'm not sure what, if any, stuff should be moved from the PowerPC page to there, nor how much of the stuff on the current "IBM POWER" page would belong on the "Power Architecture" page, as opposed to, for example, an "IBM POWER" page talking about, for example, the history of RIOS/RSC/POWER2/POWER3/POWER4/POWER5/etc..

(Boy, is this all confusing. Perhaps it's not as bad as selling some chips as Intel Core and then announcing a new microarchitecture called Intel Core Microarchitecture which is not the microarchitecture the Intel Core chips use, though....) Guy Harris 22:39, 31 July 2006 (UTC)

---

Can we discuss this without resulting in SHOUTING? And can we all sign our entries, please? OK, back to the real question. Should this page be renamed "IBM POWER Architecture"? No, because..

1. IBM is not the source of the new Power Architecture, Power.org is. IBMs might be the largest contributor, but they have humbly seceded the control over to Power.org. I bet Freescale would be upset if we called it something with IBM.
2. The new name is "Power Architecture™". We can skip the ™ but Power.org really make a point in not using all caps since POWER != Power.

Let's just leave this page as it is.. if someone have issues with it, please exercise your power as a wikipedian and contribute or discuss it.. cordially! -- Henriok 22:15, 1 August 2006 (UTC)

Missing sources

"Floating point became a focus for the America Project, and IBM was able to use new algorithms developed in the early 1980s that could support 64-bit double-precision multiplies and divides in a single cycle."

This is questionable, because even the POWER5 does not have a fully pipelined divide. Can anyone comment on this?

....this commenter tried to express the inclusion FP-operations to the “801” design, during the america project{which would be false}, it reasonable to conceive that those issues would have been addressed during the “cheetah project”

the new algorithm this guy talks about, was not an addition to the ‘POWER Architecture”, but a new adaptation made to the “801 design” v2. in the late 70s.

the “64-bit double- precision multiplies” must be referring to IBM’s own 80-bit extended precision technique before the adoption/creation of the IEEE{1986-87?} double- precision standard.

the “multiplies and divides” he must be referring to the operations in POWER1/RS6000 which later evolved into what is used in {PowerPC604/POWER4}[assuming because of scaling issues]

Clean up and correct

This article includes quite a lot of errors, is incomplete and badly structured. I think it needs some heavy lifting to get it right. Things like:

• Belatrix was not the project name from which POWER, RSC, POWER2, P2SC and POWER3 sprung

>>>>POWER1 has been moved to {america project}, and RSC was not even listed yet

• PowerPC 620 and 970 was not a part of the Amazon project

>>>> PowerPC 620 {instruction set 64-bit PowerPC v0.x)

>>>>IBM's contract to APPLE/MOTOROLA was to supple three referance designs {601,603,604)

>>>>Amazon project was to design a 64-bit processor capable of being used in a multi-processor{64<amazon} workstation or server environment using (AS/400 and AIX) {instruction set 64-bit "PowerPC AS" v0.x}, PowerPC 620 was the working prototype for RS64

• • PowerPC 970 is a part of the GigaProcessor project (its codname being GP-UL)

>>>>PowerPC 970 is derived from work done in the "amazon" project RS64 [970 notequal POWER4]

>>>>POWER4 is derived from work done in "amazon" RS64 [but they are not equal]

• • PowerPC 620 is a part of the original PowerPC project. There's no reason for it to be mentioned in this article.

>>>i'll mention it again IBM was to supply APPLE/MOTOROLA with three designs (601,603,604) with those designs apple/motorola should have been capable of making future PowerPC compatible Processors.

• • PowerPC 630 was also a part of the original PowerPC project, but was later renamed POWER3, hence relevant

>>>>six time three 18 divided by 6{the day power3 release) hence the relevance POWER3

• PowerPC 970 is not a POWER-processor (but a part of GP project, hence relevant)

>>>>>yes 970 does not equal POWER4, but {POWER4,POWER5,POWER6,POWER7} are in th "GP-project",why speculate {and the relavance you talk about i have no clue}

>>>>PowerPC970 is part of the 5th generation of the {64bit powerpc instructions set) , hence the relevants to G5 (the project code name G5)

• • 970 is derived from POWER4, it is not a single core implementation of POWER4.

>>>>PowerPC 970 is derived from work done in the "amazon" project "RS64"

• Revolution is most likely not derived from Gekko in any significant respect.

>>>>well maybe the significance of it being based on PowerPC needs to be pointed out to you.

• The PPE in CELL is an original design, not similar to POWER3 in any relevant respect.

>>>>{your reading speculations from PS3 about the pre-release of the CELL prosessor)

>>>well i beg you pardon but the PPE{instruction set 64-bit PowerPC AS v2.x), POWER3(instruction set POWER-32bit v?.x,PowerPC64bit v1.x,support for OS/400)

• • Xbox360/Xenon consists of tripple PPE cores

>>>yea so whats you point "PPE" stands for PowerPC Processing Element

• • Revolution is most like to use PPE too.

>>>>onther statment lashing out at the obvious

• The IBM and Xilinx project is PowerPC based, not relevant to this article

>>>>The article is about IBM'S "POWER Architecture", which PowerPC-32bit is a subset of POWER, which later becomes a subset RS64

• CRS-1 is a PowerPC design, not relevant to this article (but the PowerPC article)

>>>>that (PowerPC article 98% wrong}

• POWER1 had a single chip implementation, the RSC.

>>>thank for pointing this out, but its called the RISC Single Chip, and this was the reason APPLE started talking to IBM about using RSC, because MOTO's new 88000 was not compatible with [68020,68030,68040,68060,and i think there was an 68080) so apple was going to be forced to rewrite there code one way or the other, and the performance of RSC were out-standing, {(PowerPC was a modification/based on RSC}

• • RAD6000 (manufactured by BAE Systems) is a radiation hardened variant of RSC which powered the computers onboard the Mars rovers (general interest)

>>>>>BEA is misspelled did you mean IBM, because IBM has been making extreme-condition microprocessors for military satalites for decades....

• PowerPC 630 came in a multi chip version 1995 and in a single chip package as POWER3 in 1997.

>>>>nice point but this is also the start of the transition to RS64, could be

developer-workstations

• • POWER3-II was a copper chip
• POWER4 is more similar to RS64 than POWER3.

>>>>yea

• • POWER4+ reached 1.9 GHz

>>>ok

• • There's no need to use subjective wording concerning the performance of POWER4

• POWER5 introduced eFUSEs.
  • POWER5+ reached 2.1 GHz
  • POWER5 is the reigning ruler of the TPC-C benchmark.
• POWER6 and beyond is most likely to be based on CELL

>>>here you seem to be confused CELL={PPE+SPE}

>>>POWER6 will most likly incorporate some {hopefully 12-per-core and 32-cores) SPE's

• • The design goals for POWER6 will be massive multi core, very high frequency.

>>>these were the design goals of POWER4 and they were met.

• • There's nothing other than very vague speculations of what POWER7 will be like. It's part of IBMs efforts to meet Darpa demands for a project due 2010.

>>>>that darpa project was started after POWER7

• Who licenses PowerPC is not relevant to this article

>>>then were is it relevent {pentium/opteron/mipps/huh?}

• The awards section needs to be completed and cleaned up

>>>>who are you? then complete it or wait for someone else too do it.

• There shoud be some explenation of the difference between POWER and Power (the latter is market speak for POWER, PowerPC and CELL).

>>>>because you make it so......

>>>>The "POWER Architecture"

>>consist of {POWER-32bit,PowerPC-32bit,PowerPC-64bit,PowerPC-64bit AS,IMAPI}and NOW plus (SPE)

>POWER1(POWER-32bit v1.x) AIX,BSD

>POWER2(POWER-32BIT v?.x,PowerPC-32bit v1.x) AIX,Linux

>POWER3(POWER-32bit v?.x,PowerPC-64bit v1.xx,SMP) AIX,Linux

>POWER4(POWER-32bit v?.x,PowerPC-32bit v1.xx,PowerPC-64bit AS v1.x,IMAPI v?.x,SMP,SMT) AIX,OS/400,Linux

>POWER5(POWER-32bit v?.x,PowerPC-64bit AS v2.x,IMAPI v?.x,SMP,SMT,field gate programable logic) AIX,Linux,OS/400

>CELL(PowerPC-64bit AS v2.0,SPE v1.x,SMP) Linux

>POWER6(POWER-32bit v?.x,PowerPC-64bit AS v?.x,IMAPI v?.x,SPE v?.x,SMP,SMT) AIX,OS/400,z/OS,Linux POWER6 meets the orginal goals of the Beltrix project....

>PowerPC601 (PowerPC-32bit v0.x)

>PowerPC603 (PowerPC-32bit v0.xx)

>PowerPC604 (PowerPC-32bit v1.x,SMP) AIX,WinNT4,Linux

>PowerPC620 (PowerPC-64bit v0.x,SMP) AIX,OS/400

>RSC OS/400

>RS64 (PowerPC-64bit AS v1.x,IMAPI v?.x,SMP) OS/400

>RS64II (PowerPC-64bit AS v?.x, IMAPI v?.x,SMP) OS/400

>RS64III (PowerPC-64bit AS v?.x,IMAPI v?.x,SMP,CGT) OS/400

>RS64IV (PowerPC-64bit AS v2.x,IMAPI v?.x,SMP,SMT) OS/400

>PowerPC970 (PowerPC-64bit AS v2.xx,SMP) Linux

>PowerPC970fx(PowerPC-64bit AS v2.01,SMP) Linux

Henriok 15:49, 21 December 2005 (UTC)

OK, the above is a bit hard to read, with the lack of attributions on individual items (and the lack of any attributions at all from Mr. 71.136.172.35) and the mix of bulleting, multiple bulleting, and ">" indentation.

It also seems to reflect some confusion on the part of Mr. 71.136.172.35 about instruction set architectures vs. implementations of those architectures.

It's also not helped by the somewhat mixed-up history of the various POWER-derived instruction set architectures IBM's produced - we have the original POWER ISA, the POWER2 ISA that added a few instructions for number-crunching, the PowerPC ISA that both added and subtracted instructions from POWER and added a 64-bit flavor, and the "Amazon" architecture which combined PowerPC with some or all of the POWER2 extensions and some additional proprietary extensions for the benefit of AS/400 (although what it did with the instructions deleted from POWER to make PowerPC, or the POWER2 instructions not adopted into later versions of PowerPC, I don't know).

So:

• According to the Frank Soltis interview linked to by the current version of this article, and his "Inside the AS/400" book, Amazon was an instruction set architecture; there was no "Amazon" processor. Amazon was to be shared by the RS/6000 and AS/400 lines. Belatrix was to be a processor that implemented the Amazon architecture, but that project was cancelled. It was intended to be used both for high-end commercial work in the AS/400 and number-crunching in the RS/6000, and was often called the "PowerPC 630" by the RS/6000 people. However, it was cancelled, and the RS/6000 people then "began working on a new version of the 630", which was "a 64-bit PowerPC design that would provide the NIC performance needed for future systems" ("NIC" being "Numerically Intensive Computing"). I suspect that 630 project turned into the POWER3. The AS/400 people began working on another high-end Belatrix substitute, "Northstar"; I'm not sure if that turned into POWER4, or if it appeared only in AS/400 and POWER4 was the successor to Northstar.

• Amazon was based on 64-bit PowerPC, but that, in and of itself, doesn't make the 64-bit PowerPC 620 part of the Amazon project; the 620, as far as I know, didn't implement any of the Amazon extensions to PowerPC.

• RS64 was an implementation of Amazon, used in more commercial-server oriented RS/6000's and in AS/400's; it was the first generation of "full" Amazon processors (Muskie and Cobra didn't implement the full PowerPC architecture, according to Soltis' book, and were used only in AS/400's).

• POWER4 was an implementation of Amazon, used in RS/6000's and AS/400's. The mere fact that it and RS64 both implement the Amazon ISA doesn't, in and of itself, make POWER4 a processor derived from RS64; it would be derived from RS64 only if the hardware design of POWER4 was based on the hardware design of RS64.

• An IBM item on the BladeCenter JS20 says that "The PowerPC 970 is derived from POWER4 technology". This appears to imply that the 970 is, in fact, derived from POWER4. Unless they removed the AS/400 extensions from it, that means it's based on the Amazon work, but doesn't mean it's derived from RS64 any more than POWER4 is derived from RS64.

• As noted above, RS64 is not an instruction set architecture, Amazon is. Saying that every chip that implements the Amazon architecture is derived from RS64 would be like saying that every chip that implements x86 is derived from the 80386 - there's probably very little in the hardware of a Pentium 4 that resembles the hardware of an 80386. That's the whole point of an instruction set architecture - you can change implementation techniques without causing existing programs for previous implementations not to run.

• "BAE" is not misspelled; they (BAE Systems) make the RAD6000, a radiation-hardened RS/6000 (RSC-based?), as well as the RAD750, a radiation-hardened PowerPC 750. As the page for "BEA Systems" notes, they shouldn't be confused with "BAE Systems"; BEA only makes software.

• There are separate "IBM POWER" and "PowerPC" pages. "PowerPC" could be viewed as relevant to the PowerPC instruction set architecture and to all processors that implement it, but "PowerPC" is often used only to refer to "merchant" semiconductors, sold on the open market, as opposed to chips IBM makes only for use in their own systems. Right now, the "IBM POWER" page nominally refers to the instruction set architecture, but also lists some of the IBM-proprietary implementations of POWER, namely the processors with named with "POWER" in them, and single-chip derivatives of them (RSC, P2SC). It also mentions PowerPC if, for no other reason, that POWER3 and all later POWERn processors implement the full PowerPC architecture, and are thus PowerPC processors as well as POWER processors. There will probably always be information that could credibly be argued belongs either in the "IBM POWER" page or the "PowerPC" page.

Hopefully this clarifies some of the issues. I'd suggest that followup comments

1. NOT be inserted after my items in that list, so that we don't get another intermixed mess - instead, put your followups after my entire list, in separate items, so that each of those individual items can have a separate discussion;
2. be signed with four tildes, so that it's clearer who said what;

to make the discussion easier to read. Guy Harris 00:40, 25 December 2005 (UTC)

As for "POWER" vs. "Power", IBM says "Power Architecture™ technology is an instruction-set architecture that spans applications from consumer electronics to supercomputers. Power Architecture encompasses PowerPC®, POWER4™ and POWER5™ processors." Guy Harris 03:57, 25 December 2005 (UTC)

Is "POWER" really just PowerPC in all modern processors?

Does IBM (or any other Power Architecture vendor) have any processors that still implement the POWER instructions not adopted into PowerPC, or still implement the quad-word load-and-store instructions that were the POWER2 instructions not later adopted into PowerPC, or otherwise maintain compatibility with POWER or POWER2, rather than PowerPC? Or, to put it another way, is anybody implementing the original POWER instruction set, or the POWER2 instruction set, any more, or has it been completely replaced by PowerPC in modern processors? (Does Amazon include any of the instructions from POWER or POWER2 not adopted into PowerPC? Did any of the Amazon processors implement them? Did POWER3, POWER4, or POWER5 implement them?) Is the "Power Architecture" now just another name for the PowerPC architecture? Guy Harris 11:19, 25 December 2005 (UTC)

The way I've commonly seen the term "Power Architecture" used is to refer to various CPUs and cores derived from IBM's late PowerPC/POWER lineage. Here, the term "architecture" is used for a family of similar IBM microprocessor designs, not necessarily as an indication of the ISA (though the ISAs are all similar as well). In other words, IBM is using the phrase to refer to their implementations, not the PPC/POWER ISA. Good job so far with the cleanups to this article, by the way. -- uberpenguin 23:08, 25 December 2005 (UTC)

Unified Power spec

Power.org recently revealed a unified Power spec, the Power ISA 2.03, replacing all older specifications like Book E. Power 2.03 includes support for virtualization and incorporates AltiVec.

Press release

I think someone more qualified than me should write something abour this. -- Henriok 12:10, 31 July 2006 (UTC)

So is this a successor to various 2.x versions of the PowerPC instruction set architecture? And, if so, does that mean, in effect, that what used to be called the PowerPC instruction set architecture is now called the Power instruction set architecture? And, if so, what should be done to the IBM POWER and PowerPC pages about this (e.g., what stuff about the various flavors of instruction set in the POWER/PowerPC/Power family belongs in which page)? Guy Harris 16:22, 31 July 2006 (UTC)

I think we should make a "Power Architecture" article, with the new logo, history (a lot from this page) and small subsections about the three main parts; POWER, PowerPC and Cell. Then we'll rename this page "POWER ISA", and make small linking adjustments to the relevant subpages like PowerPC, Xbox 360, the different implementations and so forth.

What is now the PowerPC ISA will still be called that, but PowerPC as a trademark will probably go away.. "Power everywhere", as IBM will have it. After a while the PowerPC and POWER ISA articles will freeze as there will be no new PPC products, only Power Arcitechture. There's a lot of relevant reading over at [power.org], which illuminates their take on this. -- Henriok 18:21, 31 July 2006 (UTC)

So which page - "Power Architecture", or "POWER ISA" - will be the page for the Power 2.03 (not all-caps, apparently) instruction set architecture? Guy Harris 23:38, 31 July 2006 (UTC)

Backronym?

Why is POWER claimed to be a backronym? When the RS/6000s were announced, IBM was calling POWER by the Performance Optimized name. I just came across the announcement ad in the May 1990 issue of Scientific American.Wdonzelli 16:38, 26 October 2007 (UTC)

Agreed. The IBM RISC System/6000 processor: Hardware overview, IBM Journal of Research and Development, Volume 34, Number 1, Page 12 (1990) This ref clearly defines: IBM POWER (Performance Optimization With Enhanced RISC) It was published in January 1990. --mikeu ^talk 13:48, 20 September 2009 (UTC)

POWER1

POWER1 redirects here. Why? Every other POWER microprocessor has its own article. The current amount of information for the POWER1 in this article is not enough. I am willing to write an much more detailed article if I have support for turning POWER1 into an article. What does everyone think? Rilak (talk) 14:31, 4 August 2008 (UTC)

Go for it! I'll support you all the way. -- Henriok (talk) 20:52, 4 August 2008 (UTC)

Excellent! I'll probably start editing in a week or so, depending on my other commitments. Rilak (talk) 05:53, 5 August 2008 (UTC)

Tip: Do not edit this article, start buildning in you own user space: User:Rilak/POWER1 and when it's done.. -- Henriok (talk) 12:20, 5 August 2008 (UTC)

Why? Is everyone doubtful of my writing skills? :)

Nah, its fine, if you want to review the article or collaborate on it before it gets released, I already have a page User:Rilak/02 that can be used. Rilak (talk) 12:46, 5 August 2008 (UTC)

Striking out the link to my subpage as the article has been released. Rilak (talk) 11:15, 27 August 2008 (UTC)

The first bit is up for review here. It isn't anywhere near finished, the microarchitecture section still has to be written. Rilak (talk) 07:36, 7 August 2008 (UTC)

wackypedia

WACKYPEDIA'S growth is sloping off and it appears it is due to the bizarre antics of some of its editors.

For years, the INQ has been saying that within Wackypedia is a closed circle of editors with blinkered worldviews who are killing off real information. —Preceding unsigned comment added by 205.127.250.153 (talk) 18:36, 6 August 2009 (UTC)

I think Anonymous coward above is referring to us "closed circle editors" engaging an edit war to rid the article of unsourced, originally researched and off topic edits made by an anonymous and inexperienced contributor that's proven time and again unwilling and impossible to communicate with. Personally, I think that Above Complainer is the same person making these unlawful edits. -- Henriok (talk) 22:59, 6 August 2009 (UTC)

Endianess

Nowhere in the article does it state the endianess of the POWER architecture. Is it simply always big-endian? Are some versions bi-endian, although certain operating systems, e.g. AIX, always run them in a certain way, etc. Even if the answer is simple, I'd expect it to be stated. -- Ralph Corderoy (talk) 09:55, 2 September 2010 (UTC)

It is big. I can't find any source to back it up though, but it i recall correctly bi-endianess was an extetension to PowerPC and is not in any architecture. -- Henriok (talk) 15:49, 2 September 2010 (UTC)

Bellatrix project

In the subsection "The Bellatrix project" the page says:

Sometime in the years of 1986-89, the Bellatrix project was started, with the premise of using the America architecture as the base for a common architecture that could host OS/390 for mainframe applications, OS/400 for multi-processor server transactional processing, and AIX for scientific applications.

Sometime between the years of 1990-95, the project was considered overly ambitious and was canceled.

and in the subsection "The Amazon project" it says:

In 1990 the Amazon project was started to create a common architecture that would host both AIX and OS/400. ... IBM management wanted them to use PowerPC, but they resisted, arguing that the existing 32/64-bit PowerPC instruction set would not enable a viable transition for OS/400 software and that the existing instruction set required extensions for the commercial applications on the AS/400. Eventually, an extension to the PowerPC instruction set, called "Amazon", was developed.

At the same time, the RS/6000 developers were broadly expanding their product line to include systems which spanned from low-end workstations, to mainframe competitor-large enterprise SMP systems, to clustered RS/6000-SP2 supercomputing systems. ... But mainframe and large clustered supercomputing systems required more performance and reliability, availability and serviceability features than processors designed for Apple Power Macs. ...

The project to develop the first such processor was "Bellatrix" (the name of a star in the Orion constellation, also called the "Amazon Star"). The Bellatrix project was extremely ambitious in its pervasive use of self-timed & pulse based circuits and the EDA tools required to support this design strategy, and was eventually terminated.

The latter speaks of Bellatrix as the first processor planned for the Amazon project, speaks of the Amazon project as developing a common architecture for OS/400 and AIX but not for OS/390, and speaks of the entire Amazon project being started in 1990. Either there were two Bellatrix projects or one or both of those statements is wrong. I think Bellatrix was mentioned in Frank Soltis' "Inside the AS/400"; I'll dig my copy out of storage at some point (or, if the relevant pages are available on Google Books, check there) at some point and check. Guy Harris (talk) 07:33, 8 January 2012 (UTC)

Fork to POWER Instruction Set Architecture and POWER microprocessors

I think this article needs to be forked into two articles. One describing the old, deprecated ISA that's IBM POWER Instruction set, and another that describes the microprocessors that share the name (but hasn't been using the above mentioned ISA since the late 90s). I propose these two articles:

• IBM POWER Instruction Set Architecture
• IBM POWER microprocessors

There's been a lot of arguments (here on Wikipedia and in the world at large) where the origin lies in the confusion in nomenclature that IBM have pushed upon us. Let's make it right. They are related, but they are not the same think. -- Henriok (talk) 12:19, 20 March 2013 (UTC)

Where would the "History" section of this article go? It goes all the way from the 801 to the POWER8, including PowerPC, so it's not just a history of the original POWER ISA, which would presumably stop at the POWER2 version. Should the history be put into both the "IBM POWER microprocessors" page and a new "History of Power Architecture" page, with the stuff not relevant to the hardware history (e.g., the section on PowerPC) removed from the copy in "IBM POWER microprocessors" and the stuff not relevant to the ISA history (e.g., the clock rates and feature sizes of POWERn processors) removed from the copy in "History of Power Architecture"? Guy Harris (talk) 18:09, 20 March 2013 (UTC)

Yes, sounds good. 801→POWER3 in POWER ISA article, and a condensed version →POWER8 (including PPC, AIM, AltiVec, RS64, Power.org, etc) in the POWER processors article I think. I think we can leave expanded history sections of those topics in their respective articles. -- Henriok (talk) 18:43, 21 March 2013 (UTC)

So I did it. I was bold. I forked it… Look at the links above. -- Henriok (talk) 11:53, 24 March 2013 (UTC)

Is it OK now to go through all the inbound links and redirect them to IBM POWER Instruction Set Architecture and IBM POWER microprocessors or other relevant articles (such as Power Architecture)? What to do with the comments? -- Henriok (talk) 13:53, 8 April 2013 (UTC)

I've already done some of them. Might as well do all of them.

If by "the comments" you mean the comments in Talk:IBM POWER, well, some of them apply to the ISA, some apply to the processors, and some apply to both, so that's a bit messy.

According to Wikipedia:Copying within Wikipedia:

Wikipedia's licensing requires that attribution be given to all users involved in creating and altering the content of a page. Wikipedia's page history functionality lists all edits made and its users. It can not, however, in itself determine where text originally came from. Because of this, copying content from another page within Wikipedia requires supplementary attribution to indicate it. At minimum, this means a link to the source page in an edit summary at the destination page—that is, the page into which the material is copied. It is good practice, especially if copying is extensive, to make a note in an edit summary at the source page as well. Content reusers should also consider leaving notes at the talk pages of both source and destination.

That page's "merging and splitting" section points to Wikipedia:Splitting, which discusses how splits should be done. This is arguably a split, although it's not a case of just "taking bits from one article and putting it into a new article", as both articles are being renamed. The section on "how to properly split an article" has a whole bunch of steps which might be the right way to handle both the edit history and the talk page. At this point, the right way to handle this might be to put Template:Histmerge onto both IBM POWER Instruction Set Architecture and IBM POWER microprocessors, asking to have the history moved from IBM POWER to both pages, and also, as this is a "more complex case", do as the documentation page for Template:Histmerge says and

...leave a description of the problem under the "List of cut-and-page moves to be redone properly" section at the Wikipedia:Cut and paste move repair holding pen.

and perhaps also ask what to do about the talk pages (some of the stuff in Wikipedia:Splitting#How to properly split an article might be the right answer).

(And, when the smoke clears, maybe IBM POWER should become a disambiguation page.) Guy Harris (talk) 01:05, 12 April 2013 (UTC)

I relinked all links from articles that points here. It was a mix of stuff that now goes to Power ISA, POWER processors, PowerPC and Power Arch articles. It was very satisfying. All that's left are links from Talk-pages which I'm reluctant to do anything about. -- Henriok (talk) 14:31, 12 April 2013 (UTC)