back to article IBM opens up Power chips, ARM-style, to take on Chipzilla

With its embedded Power chip business under assault from makers of ARM and x86 processors – and to a lesser extent MIPS chips – and having lost the game console business to AMD, IBM had to do something dramatic to expand the addressable market for its Power processors. And that something, which Big Blue has just rolled out, is …

COMMENTS

This topic is closed for new posts.
  1. John Savard

    Too Little, Too Late

    It's obvious why one would want an x86 chip: to run Windows. As for an ARM chip, one would use it to run Android. And there's lots of z/OS software that gives IBM its mainframe customers.

    But there's no particular reason to choose the PowerPC architecture. Some time back, IBM was giving away IP to people wishing to design PowerPC-compatible chips, and nothing much resulted.

    1. John Gamble

      Re: Too Little, Too Late

      Sure, and the reason one would want a 68000 chip is to run a MacOS. Oh, wait...

      The x86 chips are not going away yet, true, and they may hold the lead for a very long time, but it's not guaranteed that they'll still hold the majority. What's going to drive chip choice now is the software that runs on it, and there are plenty of platforms that would be equally happy running on a PowerPC as on an x86 chip.

      1. itzman

        Re: Too Little, Too Late

        well for some of us not wedded to Gates at all, there is little difference between ARM PowerPC or Intel, except on price and performance.

        Since linux ports exist, or could exist for any of them. And on big servers most of the code is compiled specifically for the platform, or is platform independent anyway using java, oracle or WHY..

        There is a world beyond the desktop PC. Although Redmond wont accept it.

        And even ON the desktop, the vast majority of people run nothing more than the office suite, a mail client and a browser. Or use webmail anyway.

        All of which are FUNCTIONALLY OS independent. (avoiding the need to get into msoffice versus open office/libre office wars)

        1. This post has been deleted by its author

    2. Voland's right hand Silver badge

      Re: Too Little, Too Late

      Errr... excuse me... It is me being too thick this evening. Last time I recall Android was java that is not java but is kind of java but not java at all running on top of Linux.

      That runs fine on x86, MIPS and shall run fine on Power too. The only advantage Arm ever had here was the ability to bake in various special hardware accels. MIPS moved to cancel this advantage a while back. Power did that now. I am going to take a very safe bet - x86 will follow.

      That is as far as Android. As far as other stuff, Power has a _MAJOR_ advantage over arm. It has mature, time tested 64 bit architecture. So while I agree that it is a bit late to the party I would not be so sure about too late.

      1. ThomH

        Re: Too Little, Too Late

        Some Android titles are built using the native development kit, which builds directly for the ARM and whatever else the developer chooses. If an app is compiled for native execution then usually it's a game — it's especially likely with ported titles.

        Besides customisability, amongst ARM's other strengths are code density (in thumb mode, anyway) and power efficiency.

      2. Anonymous Coward
        Anonymous Coward

        Re: Too Little, Too Late

        Power has no heritage as a battery powered mobile processor. I can't think of a single device that used it for mobile.

        1. Rampant Spaniel

          Re: Too Little, Too Late

          I believe a fruity computer manufacturer did have a fling with power chips, including in laptops. PASemi also developed a low power version (at least relatively low powered for the time). Over a decade ago ibm was mumbling about it's LP chips running pda's, embedded systems etc, no idea if they ever did though.

          Many people said x86 couldn't be a realistic low power contender, buggering with compiliers aside they at least are showing some promise in that.

          I can vaguely see a chance for this in phones \ tablets if they can compete on power per watt and only if it can be put in phones etc in a manner which means the user doesn't notice it isn't an arm chip. There could be a market in enterprise as well. It's not likely to worry intel in the desktop or laptop markets at least in the near future.

          It stands a chance but will require luck and investment. This is probably better than letting it die, not a sure fire win but a chance.

        2. seven of five

          Re: Too Little, Too Late

          Thinkpad 860 - 166MHz PPC 603e

          more of a white elephant, though...

        3. solid gold suleyman

          Re: Too Little, Too Late

          I will give you at least one. The DCS220 and DSC260 digital cameras used MPC823 processors from Motorola (now Freescale), which had a PowerPC core.

    3. Anonymous Coward
      Anonymous Coward

      Re: Too Little, Too Late

      As soon as Apple ditched the PPC it was the end for Power on the desktop.

      PowerPC was being directed at low cost embedded and high end servers, there wasn't a middle ground desktop product.

      Part of the reason why Apple supposedly ditched PPC was that there wasn't a G5 CPU that would have worked in a laptop. Given how many laptops were sold after that decision (as people ditched desktops), I'd say that was the right choice.

      One problem will be a lack of expertise in PPC Linux and lack of support.

  2. Paul Crawford Silver badge

    It will be interesting to see how this pans out. Sun was one of the first to open up a CPU for such things but it ultimately failed to make enough money to survive, and Oracle have, it seems, little real interest in this.

    Given the "limited success" of Itanium, it seems the only significant player left is IBM so maybe it can work this. But...I find it hard to see what most users will find that makes it sufficiently desirable compared to the current market leaders of x86 (lots of legacy software) or ARM (cheap license, good for systems with lots of cores).

  3. xperroni

    But why?

    I'm curious as to why would someone prefer to use a PowerPC chip on a device, as opposed to a x86 or ARM chip?

    ARM's generally sell themselves on being customizable, providing "good enough" performance on a restricted energy envelope, and remaining affordable at relatively small batch sizes.

    Intel seems to believe x86 is preferable for its larger user base and software library, that it offers the best performance while (for current models at least) still being energy-efficient enough, and tries to be price-competitive by producing large batches of identical units.

    Where in this picture does PowerPC (or MIPS, for that matter) fit? Does it offer its own advantages that make it appealing to customers? Or is this just a "me too" effort from IBM, akin to HP's open-sourcing of webOS?

    1. ThomH

      Re: But why?

      I'm curious about this too, but with the caveat that the PowerPC architecture obviously has some advantages for some demographics — the first generation XBox was x86 based but the 360 switched to PowerPC, even at the cost of Microsoft having to buy a company to supply emulation software for backwards compatibility. Similarly you'll find PowerPC-derived parts in the PS3 and every Nintendo since the Gamecube.

      Then again, both Sony and Microsoft have switched to essentially the same x86 architecture for the next generation so maybe that demonstrates nothing.

      1. Anonymous Coward
        Anonymous Coward

        Re: But why?

        I would suggest that given IBM's shedding of low margin business that they simply didn't want (or see enough value in) supplying chips for the consoles. Or that Microsoft and Sony squeezed the margins so tight that IBM couldn't be profitable (but where AMD could be...)

    2. Anonymous Coward
      Anonymous Coward

      Re: But why?

      PowerPC is a RISC chip, it was good but failed to keep pace with x86 in terms of power and cooling requirements. The Mac G5 was very power hungry, there was no G5 laptop.

      ARM and x86 are mature and power efficient. That's not to say that PPC isn't or can't be, but there isn't really much experience of it as a phone CPU.

    3. Anonymous Coward
      Anonymous Coward

      Re: But why?

      AS others have mentioned, this isn't about PowerPC.

      But as to why someone would want to use Power over x86 and ARM? I think that SPARC and POWER both have SMT4 (I think the newest SPARC has SMT8, if I am not mistaken). So for large applications where you have software running many large lightweight threads, this would seem like an advantage to me - the cost of a context switch is markedly less.

      As for choosing POWER over ARM? I think that they are in significantly different markets. You can have low power clusters of hundreds of ARM chips, but good luck delivering data to all of those cores so that they can do useful work; and, no, sitting around spinning waiting for data to arrive from the L2/L3/main memory is not useful work. Admittedly, I don't know how fast POWER's memory bus is, but I think it'd be at least as fast as the current generation Xeon's memory bandwidth of about 25GB/sec. Now, you just got to find a way to use all of that available memory bandwidth :)

      I don't see the point of handicapping POWER cores (and still maintain ISA compatibility) so that everything fits into a low power application like where ARM currently dominates - unless you want the full 64-bit capability and an application to scale from the very small to the very, very large.

      Additionally, POWER has been known to scale to large SMP systems - much larger than what x86 currently does (Sandy Bridge-EP had 2-sockets, Westmere EX had 4 sockets. I'm not up on what Ivy Bridge or Haswell offers). What applications would it be preferable to have large SMP over, say, a very large cluster? I guess as long as the NUMA memory effects aren't too bad and your system requires memory faster than what can be delivered over your cluster fabric. But these sort of scale-up scenarios are increasingly rare, I suspect....

    4. andro

      Re: But why?

      But why? Where I work I see large corporates buying IP and making custom chips for custom applications (think engine computers, motorola cpus, and other bits of hardware baked in to make a purpose build low cost application platform.).

      This is something that people will keep wanting to do, and with IBM having been a large long time supporter of linux they will not worry about not being able to get vendor support.

      In these applications it doesnt matter what chip it is. It'll come down to what options you can put in the silicon, and price point, and the existence of software tools. This will make ppc hardware or ppc + linux os a viable option in this field. Sure the chips were falling behind vs x86, but were not talking ppc from 3 years ago. Were talking the next gen, and if they have done the engineering right with regards to heat and power they may remain relevant in some markets.

      Just dont expect one on your desktop pc or your phone. If thats what your thinking, you have not seen the deeper picture.

    5. Ian Ringrose

      Re: But why?

      -> I'm curious as to why would someone prefer to use a PowerPC chip on a device, as opposed to a x86 or ARM chip?

      You cannot build an x86 into your own system chip design – combining CPUs with custom networks may be what Google wants. ARMs cpus are not very fast, they just use a lot less power than anyone else, ARM is also not great with large amount of RAM.

      (I don’t think they are aiming at smart phones….)

  4. Beachrider

    IBM hasn't done mobile, true-HPC and other variants for over a decade...

    IBM is close to bringing out Power 8, just like the article says. They probably feel that they have been leaving money on the table, because they hadn't been bringing out an array of processors.

    It COULD be that they don't believe in their internal capability to switch-on this business capability, since it has badly atrophied with a decade+ of disuse.

    In any event, this is about generating income ONLY. When IBM wants to control a market, they DON'T open their products, like this. It is a very interesting move by IBM.

  5. FutureShock999
    Thumb Up

    This is an HPC and server play...

    To all of the above that couldn't figure out what you use Power for - you use it when you have SERIOUS need for compute cycles with a relatively advanced instruction set. It's not about power efficiency a la ARM, it's not about Windows compatibility a la Intel...it's simply a very good architecture for high workload compute-intensive tasks. Yes, so is Intel, but the PowerPC arch and instruction set are in many ways cleaner, and always have been. This can be important if you are going to use them as building blocks, say by designing your own CPU/GPU hybrid on-die. This is the stuff you take as a core building block if you are going to design your own HPC modules - a solid, well-documented architecture that is pretty clean and scales well.

    I highly applaud IBM's thinking on this - MIPS was always underpowered (after the 90s at least), SPARC in many ways too...but PowerPC has always kind of deserved that name. It's exactly what a chip designer needs if they are going to build the hardware for the next Deep Thought. It's a niche market, but it's where the excitement is in terms of cool HPC applications, particularly if someone closely-couples it with GPU parallelism on-die. "Imagine a Beowulf cluster of..." indeed. The ONLY part of this that is surprising is that IBM is not opening up AIX to run on the resultant customer-derived chips...AIX could really be leveraged on some of the resultant architectures to provide a solid HPC platform. But overall, good show IBM!! Let's see what gets born from this...

    1. itzman

      Re: This is an HPC and server play...

      Yup.

      Or simply port linux with some custom extensions to loads spread amongst multi-threaded tasks.

      Power PC might well have a niche market in the big tin and massively CPU bound graphical processing workstation.

      AND with more and more servers going linux over hypervisors or whatever its called, there's a big market for enterprise class virtual server clusters that don't need to tun MSDOS-with-a-pretty-face any more.

      Desktops no, there X86 will probably stand the test, but big servers? they have a damned good chance, and on power workstations - you know the sort of stuff that used to run on Irix or SPARC machines. CFD and massive computation and evaluation of parallel mathematical stuff. And THAT is where the GPU integration comes into play.

    2. Jesper Frimann
      Linux

      Re: This is an HPC and server play...

      "The ONLY part of this that is surprising is that IBM is not opening up AIX to run on the resultant customer-derived chips...AIX could really be leveraged on "

      You forget the SCO lawsuit, which is still lurking in the shadows..

      // Jesper

  6. danR2

    Power still kicking?

    601? I thought that architecture was dying a slow death inside embedded systems, such as solar-powered climate-controlled coffins.

    1. boatsman
      Devil

      Re: Power still kicking? imagine the

      number of devices that western digital sold like live book. (44x power chips)

      or the powerchips in almost every motormanagement system of many car brands

      or sat receivers, wifi routets, and so on.

      there's millions of those chips.....

      1. Anonymous Coward
        Anonymous Coward

        Re: Power still kicking? imagine the

        But no phones or tablets? which is where the growth is.

        1. Bronek Kozicki
          Pint

          Re: Power still kicking? imagine the

          Actually, no. You can only sell this many "phones or tablets" to a single customer. On the other hand, embedded CPU, as the name suggest, can be put in many devices which your missus would never suspect to have a CPU inside. Meaning, you can have many more of those - think majority of kitchen appliances, media room devices, home and car controls etc. Apropos cars: do you know how many CPUs single modern car has? Last time I heard it was going well beyond dozen.

          Have a beer from your modern, CPU controlled fridge.

        2. itzman

          Re: Power still kicking? imagine the

          But no phones or tablets? which is where the growth is.

          Not necessarily. The age of consumerism is coming to an end. This household probably has about a dozen CPUS. two desktops, one server, one laptop. two simple mobile phones a router. two cameras, a car with at least 3 CPUs in it. a smart TV. Oh hang on, there two more set top boxes..the supermarket I shop at has handheld devices, tills and my bank has ATMS..god knows how many CPUs this post goes through as well..

          there's a bunch of dead boxes as well that have CPUS in em :-)

          VERY few of these are X86. three I think. some are arm, some are - well probably - custom embedded chips i've never heard of.

          Only two are 'mobile devices'

          Oh and there's an old PABX,. That's got a CPU, and some model radio stuff. THAT has CPUs as well..what today does NOT have a CPU of some sort? the radio thermostat, the central heating timers..??? What about the two DVD players? Bet there's a CPU in THOSE too.

          Smart appliances, smart meters, the control gear that looks after everything from power stations to petrol pumps. everywhere you look there is digital processing going on, more and more of it, and an internet that needs CPU grunt just to switch messages around a vast and growing network.

          mobile devises? Pah! a flash in the pan.

      2. danR2

        Re: Power still kicking? imagine the

        I appreciate all that, but I assumed Power was simply filling in a humble role where it could no longer have the market-forces to evolve into leading edge computing functions. Apple dropped the chip because they simply felt it had run out of gas, and that was years ago. I was disappointed of course.

        1. jmsavoy

          Re: Power still kicking? imagine the

          Actually @danR2 and others, Apple dropped Power because IBM had no interest in spending untold millions to develop a lower power/lower heat Power processor designed specifically for laptops - which is what Apple was requesting. IBM's justified business reasoning for this is that at the time Apple sales amounted to only about 2% of all IBM Power chip sales (hope that puts it in perspective for everyone here, as there were a LOT of PPC MAC's floating around then).

          Power continues to be very successful in many embedded markets that are as others have pointed out constantly growing and expanding pervasively into our lives, and Power servers with AIX hold not only more market share than all other Unix/RISC manufacturers combined, but also their market share is growing in that space (See the IDC report on this from May 2013).

          This is, as others have said, a play for HPC market share - has nothing to do with phones/tablets/mobile. I assume folks are inferring this because of the correlation to the model that ARM Holdings used - thats simply a business model, it does not dictate market.

          Many of you probably recall that IBM sold off the low end Power chips (400 series ... 440BX, etc) to Applied Micro back in 2004 and has continued to shed low margin business in every IBM business unit since then. Phones/tablets are extremely low margin and don't fit IBM's high value business model.

  7. amanfromMars 1 Silver badge

    Q: When is a Fab not a Fabulous Solution? A: When ITs Control Levers are a Fabless Business?

    In days gone by, people might have wanted to innovate at the instruction set level, but McCredie says this is less important today, and being able to add functions to a Power core (or a collection of them with cache memories and such) is what people want to be able to do.

    Innovation at the instruction set level is always the prime driver for every move forward and/or backwards into the future. It is only the incumbents who people the Power core who would tell you it is less important today when adding extra OS functionality from/for/with Other Sources which have ….. well, superior disruptive intellectual property portfolios is the key driver which energises all fundamental shifts in base methodologies, and in others responsible and accountable for providing and protecting global power bases and virtual control grids ……. Sublime InterNetworking Networks.

    It would appear that IBM have at least realised that they do not lead in such fields with/for Uncle Sam and seek to hitch a ride in the caboose on the intellectual property train of a new generation of prime primary movers and shakers.

    Bravo, IBM, that's a fine best move in the present current circumstances. One trusts in GOD you will enjoy the rides, although the markets may prefer to ensure, as they always have done and like to do so that they can prosper rather than crash, that monies are better invested in lead pushing and pulling engines.

  8. John Smith 19 Gold badge
    Thumb Up

    POWER did rule in the HPC field.

    People who do geophysics simulation and visualization (IE oil companies) bought big RS/6000 SP2 systems over the SPARC multiprocessors others were touting at the time.

    And they always wanted bigger

    Going IBM gives you a)One of the worlds biggest computer companies b)Fallback wafer fab facilities if needed (the in house fab used to be pretty good, but I don't know if that's still true). c)US based (which Americans seem to like).

    This consortium puts them in close touch with potentially very big users so they get feed back on what ISA people want in the core, which is always useful.

    Thumbs up to IBM for this move, wheather it delivers the benefits they hope for is another matter.

    1. swissrobin

      Re: POWER did rule in the HPC field.

      In the "proper" computer market I would guess that PowerPC is as good as anything else, but I am struggling to see why anybody would need to do anything dramatically different than what IBM is already doing - i.e. more clocks/more cores/more cache/keep up with memory and peripheral I/O bus developments. From a s/w perspective I would imagine that most things Linux will either "just work" or can be made to work with some effort.

      For those old enough to remember, PowerPC was always a shared architecture - like Open but with just two players - IBM and Motorola. Motorola became Freescale along the way but still makes and sells these chips in a wide variety of configurations - some SoC like, some host processor like. They appear not to be innovating there any more, though, with all their new chips being ARM (full range of Cortex stuff) so at least in their (vast!) markets it would appear ARM has won out.

      I cannot see the other big embedded SoC vendors bothering about this either (TI, Atmel, NXP, ST).

      So, if the plan is to embed PowerPC in new embedded SoC designs, it will most likely fail.

      If the plan is to have somebody else produce some sort of super-server-chip out of this, I guess it might happen - looking at the partners that seems more like where it is heading - but don't expect it to have any impact on your life unless you happen to be the poor sucker managing the data centre software :-)

      1. FutureShock999

        Re: POWER did rule in the HPC field.

        I expect it will have an impact on my life in the same way that the supercomputers that developed the new engines for the A380 did - a lot quieter experience at 38,000 ft due to the super efficient blade and compressor design, that makes my semi-frequent flights from London to Melbourne a lot more bearable. And a raft of other hidden, tech-driven design improvements that we take for granted, but required someone doing a whole lot of calculations that we never see or think about. Computational Flow Dynamics and other simulations (weather, climate, etc.) take massive cycles, and the more cycles you can throw at the models the quicker you can develop accurate models, and the finer grain simulation you can run.

        1. swissrobin

          Re: POWER did rule in the HPC field.

          Sure, but that's thanks to the algorithms and engineers that deployed and analysed them. The PowerPC is just another calculator - it's not like it's 10x or 100x faster than another calculator built in the same era. So whether Rolls Royce use a PowerPC or x86 "super" computer to do the number crunching isn't going to change the fact that you can enjoy a quiet flight.

          Unless of course you really are saying that the PowerPC is the only CPU that we can do this number crunching on in reasonable time .. . in which case tell me more.

          1. Bronek Kozicki
            Thumb Up

            Re: POWER did rule in the HPC field.

            Correct, the difference between Power or x86 won't be as big as to make massive difference to compute power in the short term. What this will do, though, is to make Power architecture more widely available, thus helping to sustain innovation in the CPU industry, as a whole. Meaning, in the long run there might be indeed 10x or 100x more powerful chips compared to the situation when x86 was allowed to take all the deals.

            Competition is a healthy thing and I'm very happy to see IBM up the stakes. There are few companies to take on Intel, but IBM is one, as far as HPC is concerned.

    2. FutureShock999

      Re: POWER did rule in the HPC field.

      Totally correct - I once (1997) installed and configured an entire football pitch of SP/2 hardware buried deep under the United Airlines World HQ parking lot for customer demand simulation...the models it ran took 18 hours to run in a single batch cycle, and had to be run every day...we installed several terabytes of RAID 1 disk with the drives being ultra-fast Fibre Channel 4Gig drives. People bought SP/2s to do SERIOUS work.

      Interesting fact - the very normal-looking employee parking lot of United Air WHQ has reinforced concrete/macadam that is about 6 feet thick, for bomb resistance to protect their underground data centre from truck bombs. Or crashing planes...

    3. Anonymous Coward
      Anonymous Coward

      Re: POWER did rule in the HPC field.

      Some may remember that in the nineties was created the PowerOpen Association. A complete fail! I doubt reversing the order of name constituents will give a different outcome...

      Anyway, as a player in the said PowerOpen venture, I personally can witness that associating IBM and openness is quite a bit of oxymoron

  9. MacroRodent

    > The third factor is that the IT industry wants choice – meaning something other than an x86 option.

    Sounds like the same fallacy Nokia fell for with WP. Actually there already is a major competitor to x86: The ARM. A third choice has no change without some extremely compelling advantages, and maybe not even then.

    Remember Yoda: "Always two there are, no more, no less. A master and an apprentice. ". Applies to technology competition as well as to Sith lords.

  10. Anonymous Coward
    Anonymous Coward

    POWER is not PowerPC!

    There is a difference between POWER and PowerPC. PowerPC are for smaller computers, such as Xbox360, Apple, etc. POWER are for IBM's large Unix servers.

    IBM is adressing POWER with this move. Since earlier, Motorola was involved in the PowerPC (remember Apple using the 601 PowerPC cpu?).

    Anyway, this act smells desperation from IBM. IBM has always relied and heavily worked for monopoly like markets (prime example are the Mainframes). It is the IBM market model. IBM letting a market go, is not done lightly, it is only done in desperation. Why is IBM so desperate? This is a sign of weakness. Just like that Sun did, at the end: opened up all their products - because Sun lacked business. This tells us that IBM POWER lacks business. It seems that POWER margins are very weak, otherwise IBM would not have been forced to do this desperate act. What will happen with POWER cpus? Rumours say POWER will be shut down because x86 is catching up, at a much lower price. Why buy a slow expensive POWER, when you can get a cheap and faster x86? IBM POWER is going down the drain. Sad to see.

    1. Bronek Kozicki
      WTF?

      Re: POWER is not PowerPC!

      I think your pessimism is unjustified. Licensing CPU architecture has done wonders for ARM, why would it fail to deliver at least small chunk of (different, admittedly) market for IBM?

  11. kventin

    CHRP, PAPR...

    remember those?

    btw. that mentioned "open sourced firmware" -- any connection to OpenFirmware?

    so: what went wrong with CHRP/PAPR/OF? why wouldn't the same thing happen now?

  12. Justin Stringfellow
    FAIL

    smells

    Smells of IBM desperation to me. They're not in the game of giving anything away for free. Apparently they think the only chance they have of a future for power is to attempt to get traction with the open source crowd.

This topic is closed for new posts.

Other stories you might like