Talk:PowerPC

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Computing (Rated Start-class)
WikiProject icon This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
Start-Class article Start  This article has been rated as Start-Class on the project's quality scale.
 ???  This article has not yet received a rating on the project's importance scale.
 


Instruction set vs. Microarchitecture[edit]

The current version Special/Permalink:567967225 of the article does not distinguish clearly between Instruction set and Microarchitecture. I know that e.g. x86-64 is an instruction set and Intel's Nehalem (microarchitecture) is a microarchitecture implementing it. I am not sure about the PowerISA. It seems to me, like there is the either one instruction set called Power Architecture (with different versions) or even an instruction set family. Then, so it seems, there are three main microarchitecture families implementing it: PowerPC for PCs, PowerQUICC for embedded systems and POWERx for servers.

1. Is this correct? If it was, all the articles should IMO begin with PowerPC/POWERx/PowerQUICC is a family of microarchitectures implementing the Power instruction set.

No, this is not correct. There are a lot of micro architectures implementing the Power ISA. Pretty all POWERn processor (POWER7 and POWER7+ use the same, POWER8 does not) use different, all generations of PowerQUICC use different (the e200, e300, e500, e500mc, e5500, e6500 cores all use different microarchs). PowerPC G3 and G4 use different (three (iirc) different microarchs for the G4s). Most new iteration use a new microarch. There are some designs that doesn't obviously use the same, such as he G5 and POWER4, the PowerEN and Blue Gene/Q, and PowerPC 440 and Blue Gene/L. The word doesn't bear the same importance for PPC or Power Arch, more what cores goes into where. -- Henriok (talk) 13:38, 4 September 2013 (UTC)
In the current Power ISA, a microarch doesn't have to implement the complete set, there are "Books" and they range from common to more specialized features. Book I and II are required, but Book III is split into several topics where you can't implement them all. There are subjects dealing with server oriented issues, virtualization, auxiliary accelerators, SIMD-units, embedded applications, memory management, floating point variations, 32 vs 64-bit, variable length instructions, big/small/bi endianess, and so on.
A low end processor implementing the Power ISA is a single 32-bit single issue, in-order core with variable length encoding, 4 stage pipeline with 3 functional units, no cache nor MMU and no FPU, running at <50 MHz (ie the Freescale e200z0).
A high end processor is 64-bit 12x multi-core, 8 way hyperthreaded, multi issue, out-of-order core with hardware hypervisor, ~15-25 stages long pipeline, microops-cracker, 16 functional units, large, deep and segmented cache structure (4 levels, >100 MB large), integrated SIMD-accelerators and decimal floating point, running at 4 GHz (ie the POWER8).
The e200z0 is implementing Power ISA v.2.03 Book III-VLE, and the POWER8 implements Power ISA v.2.07 Book III-S. Both are Power ISA compliant but quite different. -- Henriok (talk) 15:55, 4 September 2013 (UTC)

2. The x86 Instruction set is __owned__ by Intel, ARM instruction set is owned by ARM Holdings, MIPS instruction set by Imagination Technologies, Ubicom by Qualcomm, etc. Who is/are the owners of the Power Instruction set and who are the licensees? ScotXW (talk) 08:35, 4 September 2013 (UTC)

The ISA is owned by Power.org, while Freescale and IBM hold permanent licenses. Power.org, IBM and Freescale can issue licenses to new parties, and there are specialized brokers of licenses too. New aspects are incorporated into the ISA on a regular basis (the latest is PowerISA v.2.07 with support for POWER8 and Freescale's e6500 core), and such improvements come mostly from Freescale and IBM but can come from other members of Power.org. There are -- Henriok (talk) 13:38, 4 September 2013 (UTC)
The Power ISA is owned by Power.org. The former PowerPC ISA was jointly developed and co-owned by Motorola/Freescale, IBM and Apple (IBM owned the trademark). Apple dropped out, Motorola's and IBM's implementations diverged, but came back together in 2006, joined their separate PowerPC ISAs into one and called it Power ISA. -- Henriok (talk) 15:55, 4 September 2013 (UTC)

Timelines[edit]

'In 1991, the PowerPC was just one facet of a larger alliance between these three companies. On the other side was the growing dominance of Microsoft and Windows in personal computing, and of Intel processors. At the time, most of the personal computer industry was shipping systems based on the Intel 80386 and 80486 chips...'

I think you'll find that in 1991 - before the release of Windows 3.1 even - the 386 was just about coming around and the 486 was still a bit of a way off. —The preceding unsigned comment was added by 62.1.133.99 (talkcontribs) 17:57, 19 February 2007 (UTC).

Seems reasonable, but it might be a discussion for the separate article: AIM alliance. -- Henriok 08:12, 20 February 2007 (UTC)
I think you'll find that in 1985 the 80386 was just coming around (go to the Intel Museum timeline and select 1985; it's Flash crap, so I can't just paste a direct URL, sigh), and that the 80486 was 4 years off (released in 1989[1]). Guy Harris 03:47, 21 February 2007 (UTC)

A big problem[edit]

This page seems to be messed up in some way. there is a curseword displayed prominently at the top of the page, and several sections appear to be missing. also, when i go in to edit the page to fix this, the text in the edit screen does not match the text displayed in the artical. dare i use the dreaded "hack" word? --Whiteknight 04:51, 15 Jun 2005 (UTC)

Well, yahoos come along and vandalize ages all the time; apparently, they get off on it. WikiJanitors spend a lot of time mopping up these messes (by "reverting" pages back to their more-pristine state). See the Wikipedia:FAQ for more details on the reversion process (I think).
With regard to the specific problem you had doing a reversion, it may have been caused by:
  • Someone already having done the reversion in the previous moment; I did a reversion about 12 hours ago on this very page (scrub, scrub, mop, mop...).
  • Some Wiki database problem or another; there are a few that cause peculiar results from time to time.
  • "Finger trouble" on your part.
In any case, it was almost certainly nothing much to worry about.
Atlant 11:34, 15 Jun 2005 (UTC)

AMCC/IBM deal[edit]

Also, what is the deal with AMCC? Did IBM Semi get spun off? Last I checked, it still exists... —The preceding unsigned comment was added by 70.16.29.77 (talkcontribs) 16:04, 30 October 2005 (UTC).

AMCC "[acquired] intellectual property and a portfolio of assets associated with IBM’s 400 series of embedded PowerPC standard products, in addition to a Power Architecture license", according to this AMCC press release. It appears, though, that IBM Semiconductor still sells PowerPC 440's as embedded cores, as well as selling 740's and 750's and 970's. Guy Harris 10:35, 31 October 2005 (UTC)
The deal is pretty simple. AMCC bought technology, (IP, developers and specific products) from IBM to resell and further develop processors based on the 400 cores (403, 405 and 440). There's been no official statement from either part that IBM would stop making, selling or developing processors based on the 400 cores, rather the contrary. Since 440 is the core of the processors in BlueGene/L, the notion that IBM would just dump their technology on AMCC is pretty silly and just speculation or misunderstanding from journalists and rumor mongers. -- Henriok 20:34, 21 June 2006 (UTC)
AMCC dropped at least 1 project from IBM: Artic PDA based on 405LP.

I took a stab at the Apple announcement[edit]

The AIM alliance ended yesterday, and I did not see an update. Therefore I made a new paragraph under "history" stating the new history. I also changed the paragraph under "design wins" to say PowerPC "has beed used" on these various platforms, not "is used" in present tense as previously written so that the design win for PowerPC with Apple can be recorded but shown to be not in use anymore.

---

History section needs some work[edit]

I was recently reading this article by Frank Soltis talking about the history of PowerPC and the ambiguity of the name. I think the history section of this article should be changed to reflect the fact that POWER was named such retroactively, to try and differentiate it from what became known as PowerPC. If nobody else minds, I'll try to work up something. -- uberpenguin 20:35, 2005 May 24 (UTC)

POWER was named retroactively? The Soltis article says "During the 1980s, IBM researchers created an enhancement to RISC processor designs that allowed the processor hardware to start the execution of multiple instructions in a single cycle. They called this enhancement superscalar. The first superscalar RISC processor appeared in the RS/6000 in 1990. To identify this superscalar enhancement to a RISC processor, IBM named the architecture POWER for Performance Optimization With Enhanced RISC. The POWER architecture was the starting point for the joint effort between Apple, IBM, and Motorola in 1991 to develop a new RISC processor architecture." That suggests that it was called POWER from the beginning (I remember the term POWER being used early in the life cycle of RS/6000).
At this point, I think all of the IBM POWER processors support all instructions in the PowerPC instruction set, as well as perhaps some or all of the instructions that were removed from POWER in PowerPC; what modifications were made in the PPC970 to make it "backwards-compatible with 32-bit PowerPC processors"? As far as I know, POWER4 has all of the PowerPC instructions, both 32-bit and 64-bit, although it didn't have Altivec^H^H^H^H^H^H^HVMX. Guy Harris 08:57, 28 October 2005 (UTC)
Uberpenguin is wrong and Guy is correct. Mirror Vax 15:38, 2 December 2005 (UTC)
The 970's dcbz (data cache block zero) instruction was changed and a dcbzl instruction was added for compatibility with the G4, etc. dcbz zeros a cache line, which is 32 bytes on earlier 32-bit CPUs and 128 bytes on the 970. Many mac programs used this as an optimization, and it would zero four times as much memory as expected on the G5, so IBM added a mode to the 970 where dcbz zeros only a quarter of a full cacheline, and dcbzl the entire thing. 31 July 2009 —Preceding unsigned comment added by 76.210.76.212 (talk) 03:41, 1 August 2009 (UTC)

I removed the sentence "One other major problem was that the 88000 (and all earlier Motorola designs) were big-endian, whereas the POWER was little-endian." because I believe it is incorrect. Evidence:

p. 234 of The PowerPC Architecture: A Specification for A New Family of RISC Processors (Second Edition, Morgan Kaufman Publishers, 1994.) indicates that the RS/6000 is a big endian architecture.

The archives of comp.arch contain many references to AIX and RS/6000 being big-endian.

I also re-wrote part of the history section, following the chronology presented in the Forward of The PowerPC Architecture (ibid.), written by Phil Hester of IBM. This appears to be a more reliable source than that of the "so the story goes" version which I replaced.

I removed claims that Apple is/was Motorola's "largest CPU customer". That is very unlikely to be true in terms of volume shipped (consider the embedded market).

Also corrected comment w.r.t. endian switch requiring a reboot. It doesn't. Added expository text on PPC endian modes in linked article. Removed reference to microcode, as it is either ambiguous or incorrect. --Kday

You're correct. As far as I know, the only general-purpose OS that used little-endian mode on PowerPC processors was Windows NT; the older POWER, RSC, and POWER2 chips were big-endian only. Guy Harris 08:57, 28 October 2005 (UTC)

Some suggested changes to the PowerPC article[edit]

>>>>The 970 is a 64-bit processor derived from the POWER4 server processor. To create it, the POWER4 core was modified to be backwards-compatible with 32-bit PowerPC processors, and a vector unit (similar to the AltiVec extensions in Motorola's 74xx series) was added.

It probably should also mention the new PowerPC 970FX, which debuted in 2004. The major changes were to switch from a 130nm process to a 90nm process (which allowed the clock rate to be boosted from 1.6GHz to 2.2GHz), and to reduce power consumption from 2.5V to only 1.8V. Oh yes, and the FSB was increased as well (half the clock rate). It's also worth noting that IBM married the PPC 970/970FX chips with AMD's HyperTransport Tunnel in the BladeCenter JS20 blade servers. Interesting combination.

Apple uses HT in their G5 offerings. The system controller is named U2 and U2H, which appears to be identical to IBM's CPC925 that JS20 uses. -- Henriok 10:24, 22 June 2006 (UTC)

>>>> * 970 (PowerPC G5) (2003) 64-bit implementation derived from the IBM POWER4 enhanced with VMX (AltiVec compatible SIMD extensions), at speeds 1.4 GHz, 1.6 GHz, 1.8 GHz, 2.0 GHz and 2.5 GHz

I don't believe 1.8/2.0/2.5GHz versions of the 970 were ever shipped (although I could be wrong). I think it stops at 1.6GHz (maybe 1.8). Also, my understanding is that VMX isn't merely "AltiVec-compatible". VMX, AltiVec and Velocity Engine are the same thing, just marketed under different names by the three companies.

Apple shipped 970 in speeds up to 2 GHz. The original PowerMac G5 (Im typing this on one) came with 1.8 and 2 GHz 970 processors, not 970FX. -- Henriok 10:24, 22 June 2006 (UTC)

You might add another line for the 970FX, such as:

>>>> * 970FX (PowerPC G5) (2004) 90nm version of the 970, at speeds 1.8 GHz, 2.0 GHz, 2.2 GHz*, 2.5 GHz

Note the addition of a 2.2 GHz speed. This is the chip used in the current IBM JS20 blade.

..and 2.7 GHz. Used in the last Apple PowerMac G5 and in IBM JS21. -- Henriok 10:24, 22 June 2006 (UTC)

Take care.

Mark (mchapman@us.ibm.com)

About Xilinx Virtex[edit]

Can somebody include some info about the PowerPC embedded cores present in Xilinx VirtexII-Pro and Virtex 4 LX FPGAs?

PPC405 in Virtex-2 and Virtex-4 families, PPC440 in Virtex-5 — Preceding unsigned comment added by 213.79.90.226 (talk) 14:45, 23 July 2012 (UTC) Jimgeorge

About BlueGene[edit]

Doesn't current fastest computer in the world, Blue Gene, use PowerPC400 32-bit cpus? It says in this article that PPC440 is used for embedded systems and not mainframe high performance supercomputers ?? Is this a shift for PPC440 from embeded systems to supercomputers? Why use such "weak" processor in world's fastest computer and what's so special about it? Other processors would be a much better choice (faster clock, 64-bit, etc.)..

Yes, Blue Gene/L uses IBM PowerPC 440 chips. The reason IBM did not opt for something as beefy as POWER5 or perhaps PowerPC 970 is because using these chips wouldn't buy very much in the way of cluster computing performance. CPUs designed for massive I/O performance with a hefty load of cache aren't necessarily desirable for a cluster computer. It makes a lot more sense to use cheaper, abundant, easy to manufacture, and slightly more "stripped-down" processors in compute nodes due to the nature of the kind of tasks generally assigned to a distributed computer. Faster clocks cause more power to be dissipated, 64-bit datapaths increase transistor count within the same (or similar) design, but it buys nothing in the way of performance (unless you are dealing with data sets of VERY large integers or HUGE data chunks, both being somewhat rare since a lot of supercomputing-worthy tasks are floating-point intensive and easily broken into many smallish pieces). What is special about the PowerPC 440 is that it is a fairly small and simple PowerPC implementation that can be run at a fairly high clock speed but doesn't dissipate the massive amounts of power that larger, more complex chips do. In other words, it's a good choice simply BECAUSE it isn't, in itself, all that spectacular. -- uberpenguin 01:01, 2005 May 12 (UTC)
One of the key points of BG/L is that it's extremely dense. They can pack 1024 processors into one standard rack. The reason they can do this is that the processors draws very little power, and requires very little cooling. The 440 core is excellent in this respect. The core is very versatile and IBM has a lot of tools and add-ons so they can customize it to suit specific needs, extreme throughput and double point flotas in thie case. It is just perfect for what IBM had in mind for BlueGene/L, ie. cool, customizeable, powerful.
The result is 1024 440 processors in a rack. That should be compared to a rack full of Apple Xserves with just 84 processors, or a rack full of IBM JS21 blades which totals 168 CPUs. Even if a 970FX@2.5 GHz is much more powerful than a 440@700 MHz, one can't argue with the fact that BG/L packs almost 10 times as many processors in one rack, and consumes much less power. The BG/L rack has a theoretical peak floating point perormance which is about 50% higher than the rack packed with JS21 blade modules, and that is taking the very powerful AltiVec-units in 970FX into account. -- Henriok 14:20, 22 June 2006 (UTC)

PowerPC 615[edit]

Back in 1995/1996 there was a rumored PowerPC 612/615 with builtin x86 intruction support (see [2] for example). Anyone have more info on this? It might make a reasonable addition to this article, since there was a lot of buzz about it in the early PowerPC days. -- Kaszeta 14:24, 6 Jun 2005 (UTC)

While I don't have the info off the top of my head, I belive I have read a book from my college library stating of a PowerPC 612. I'll go ahead and re check the book, and fully answer the question once i'm in the library [wireless nodes are great, arn't they?] CoolFox 14:39, Jun 6, 2005 (UTC)
OK! I'm in the campus library, and I found the book. Its called The PowerPC Revolution, page 175. [ISBN: 1-883577-04-7]. It states that in Autumn of 1993, IBM CPU engineers were working on a PPC chip that provides x86 instruction-set support as a subsystem directly on the CPU. The chip is being developed as a solely IBM product, outside of the PowerPC alliance [AIM]. In short, the 615 is a chip that runs native PPC code, and native x86 code, being able to run Windows and DOS without any emmulation. I hope this is all helpful. --CoolFox 17:36, Jun 6, 2005 (UTC)
Thanks, I'll add mention of this to the page. -- Kaszeta 18:22, 6 Jun 2005 (UTC)
Hey! you forgot to add the book I sourced too you to the references page! Tell you what, I'll post it. --CoolFox 18:36, Jun 6, 2005 (UTC)
Ok, I need to look a little closer at the text before saying somthing... haha... thanks Kaszeta for the collaberation! --CoolFox 18:44, Jun 6, 2005 (UTC)

another picture?[edit]

I just uploaded this picture that I think might be a good addition to this article (I didn't crop it so the metadata could be preserved):

http://en.wikipedia.org/wiki/Image:Powerpc_lucite.JPG TerraFrost 10:06, 9 January 2006 (UTC)

Clean up tag[edit]

copied from User talk:Motor

I've just done some cleanup work on this article, which you tagged. However work still must be done. We should convert the lists into prose, that would help a lot. What do you think? Thanks. — Wackymacs 13:14, 20 May 2006 (UTC)

You've covered the big reason I added the tag. IMO it could also use the following:
  • Copy-editing to clean up some clunky wording (although your edits have improved this... others parts of the article need work)
  • The Endian-modes sub-section is quite confusing and hard to follow
  • What's a "Design Win Summary"? It's a horrible section title.
  • In-line citations for the facts in the article (WP:Footnotes).
- Motor (talk) 15:12, 20 May 2006 (UTC)

Playstation 2 with PowerPC processor?[edit]

This is really strange.. I know nothing about PS2 but I know something about PowerPC-processors. This article startled me. In the last paragraph, the IBM prep claims that Sony are replacing MIPS with PowerPC-parts.. PowerPC 440 in the case of Playstation 2. Can someone make sense of this? -- Henriok 11:20, 11 July 2006 (UTC)

PowerQUICC[edit]

I think someone should make a separate article about PowerQUICC. It's an important part of PowerPC and Power Architecture, but it's a part that hasn't gotten a lot of main stream coverage since it's an embedded technology that pretty much just for industry insiders. The recent bloating of the "Motorola (now Freescale)" on this page is not appropriate, considering all the other similar entries on the page. A separate PowerQUICC-page, that'll both outline and bring some depth to what Motorola and Freescale have beed doing the pastyears in repsct to PowerPC development, would be very nice. There's certainly a lot to be said in such an article.

PowerQUICC ScotXW (talk) 08:20, 4 September 2013 (UTC)
Thanks! I posted that request in 2006, and just initiated the article by myself shortly thereafter :) -- Henriok (talk) 11:49, 4 September 2013 (UTC)

Northbridge Chipset[edit]

Is it possible to put a list of Northbridge chipsets compatible with PowerPC CPUs on this website? (Tundra, Marvel, Articia, IBM).

That'd be great.. but it would probably be better if you made a completely new article -- Henriok 07:46, 27 September 2006 (UTC)
Because the northbridge are made only for PowerPC, IMO it is better to put it on PowerPC page. I ommited MAI Logic Articia Sa/P products, because the product never reached the market. There is a free POP reference motherboard design from IBM. What is the northbridge for this motherboard?

Is it possible to create another section about PowerPC specific support IC, such as SRAM/cache, voltage controller, clock generator and internal buffer?

Digchip, splitting page[edit]

You can search PowerPC related electronic components using digchip.com and google. Search keyword: site:digsite.com <component number>. Is it possible to split this page into multiple pages? PowerPC_Implementation, PowerPC_CPU, PowerPC_Embedded, PowerPC_Chipset.

Splitting the page into 3: main, implementations (applications/products related to PowerPC architecture), intergrated circuits (desktop CPU, embedded CPU, and nortbridge related to PowerPC architecture). —The preceding unsigned comment was added by 210.210.145.10 (talkcontribs) 01:45, 16 October 2006.
I'm not entirely comfortable with this split, but since I agree that the former combined page was pretty bad I'm willing to give it a shot. But.. The way it stands now, I really can't see the improvement. It went from a pretty messy page to three pages where two might suffer by being obscure and boring. They need to be able to stand on their own.
And _please_ sign your entries here! -- Henriok 08:36, 16 October 2006 (UTC)

Define Acronyms[edit]

I think that acronyms should be defined here. What does MSR mean, for example? --69.159.228.177 14:56, 18 May 2007 (UTC)

I didn't know what MSR meant either, but hey.. Wikipedia is a encyclopedia after all so I looked it up and wikilinked it. I don't think that one should define acronyms on pages like this. They probably deserve a page of their own. -- Henriok 16:23, 18 May 2007 (UTC)
While we are at it, it's a wiki - so here is the wiki-link to MSR (Model-specific_register) for the curious. --Unixguy 18:40, 21 May 2007 (UTC)

PowerPC and MC68000 compability?[edit]

How compatible are PowerPC cpus and MC68000 cpu in terms of running software. Ie can one run any MC68000 code on PowerPC ..? Electron9 20:01, 2 September 2007 (UTC)

One can only run 68k code on a PPC processor if either 1) software running on the PPC processor reads the 68k code and interprets it as if it were a "virtual" 68k or 2) software running on the PPC processor reads the 68k code and translates it to native PPC code or both. See, for example, the Mac 68K emulator, to which the article refers.
PPC processors are incapable of directly running 68k code; the instruction sets are completely different. Guy Harris 21:33, 2 September 2007 (UTC)
Thanks for the answer. It has implications for wheather a PPC can replace a m68k while still maintaining backwards compability. (Which it can't..) Electron9 23:59, 2 September 2007 (UTC)

as well[edit]

I have watched the delightful edit skirmish over the words "as well" in the second sentence. Perhaps a rewrite of that sentence is in order. From my point of view, the PowerPC architecture is still used in some personal computers (including possibly the CherryPC and certainly many still-functioning Apple Macintosh machines) but does not enjoy anywhere near the success in that market envisioned at its creation. In addition, it is quite successful in the area of embedded, commercial, video-game and high-performance machines. Is there a wording that would satisfy both of the sparring parties?

Maybe: While seeing only limited use in personal computers following Apple's 2005 switch to Intel processors, PowerPC CPUs have become popular video game, embedded and high-performance processors. Philhower (talk) 20:26, 4 September 2008 (UTC)

I second that -- Henriok (talk) 21:35, 4 September 2008 (UTC)
Seems fine to me. Rilak (talk) 08:42, 5 September 2008 (UTC)

Motorola 88000[edit]

I think there need to be more detailed info about Motorola 88000's contribute to the development of PowerPC.

implementations vs. users[edit]

The implementations section seems to go a bit awry, going from actual PPC variants (i.e., IBMs 601 vs. 603) into listing things that are merely users of the PPC (Honda's ASMIO robot and parts of the F-35 fighter jet). I deleted the F-35 image as not really relevant or illustrating the CPU. Others may wish to edit further. Johnh (talk) 22:50, 6 September 2011 (UTC)

PowerPC's origins predate AIM alliance[edit]

is a RISC architecture created by the 1991 Apple–IBM–Motorola alliance, known as AIM.

The way I remember it the powerpc was already in development at IBM before apple picked it and started the AIM thingy. I would have edited this out except I'm too lazy to google up the evidence. — Preceding unsigned comment added by 75.85.54.137 (talk) 11:40, 14 April 2012 (UTC)

Yes that's absolutely correct and that's why the entite third paragraph in the History section describes it. -- Henriok (talk) 08:46, 15 April 2012 (UTC)