Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Google IBM Supercomputing Hardware

IBM Opens Up POWER Architecture For Licensing 131

Posted by Unknown Lamer from the finally-gonna-get-that-sweet-workstation dept.
New submitter HAL11000 was the first of many to write with news that IBM and others have formed a new consortium to license the POWER architecture to third parties "IBM puts up POWER architecture for licensing and announces the OpenPower Consortium with Google, Nvidia, Mellanox, and Tyan." Quoting El Reg: "The plan, according to McCredie, is to open up the intellectual property for the Power architecture and to allow customizations by licensees, just like ARM Holdings has done brilliantly with its ARM processors ... Nvidia is very excited about the prospects of marrying Power processors and Nvidia GPUs for both HPC and general purpose systems. ... Tyan will presumably be working on alternative motherboards to the ones that IBM has manufactured for its own use." There are mentions of the POWER firmware being "open sourced," but it is unclear if that actually means Open Source or something more like the Open Group's definition of open (vendors only).
This discussion has been archived. No new comments can be posted.

IBM Opens Up POWER Architecture For Licensing More

IBM Opens Up POWER Architecture For Licensing

Comments Filter:

  • A Little Late? (Score:5, Insightful)

    by Anonymous Coward on Wednesday August 07, 2013 @07:55AM (#44496143)

    Shouldn't they have done this while Apple was still using the PPC? At number of developers and developer tools available for PPC back then has to be orders of magnitude higher than it is today. Better late than never?

    • Comment removed based on user account deletion

      • Re:A Little Late? (Score:5, Informative)

        by drinkypoo ( 153816 ) <drink@hyperlogos.org> on Wednesday August 07, 2013 @08:20AM (#44496339) Homepage Journal

        PowerPC cores are incredibly more ubiquitous than you probably believe. They show up all over the place. Hell, Motorola Cellphones had old tired PPC cores in them for some time, since as a contributor Moto had a license to make embedded PPC chips. And of course, it's well-known that there's a tri-core PPC in the Xbox 360. There's also a castrated little PPC core in the front of the PS3's processor, where there was a MIPS core in the PS2's. And there's a ton of little MIPS-based portable computers out there, but in recent times their sales have been cannibalized by ARM. There's no reason to believe that there couldn't be a ton of little PPC-based portables out there, if PPC were licensed like ARM. Now, allegedly, it will be. Probably too little too late, though.

        • I believe many automotive onboard computers are PPC based also.

          • And spacecraft because there are rad-hard parts available.

            The thing I wonder about is all the years of ARM driving for low power (Intel has done this too) while IBM Power (uppercase) focused on being really fast and powerful for server work. Either they expect to buy a company that has the expertise to reduce Power's power consumption or they expect one of these companies to license the design and do it themselves, though I'm not sure why Power's architecture is better enough than ARM's for such a company

            • Re:A Little Late? (Score:5, Informative)

              by Carewolf ( 581105 ) on Wednesday August 07, 2013 @09:28AM (#44497143) Homepage

              POWER has been been 64bit and massively out-of-order superscalar for years where ARM is only just beginning to enter the market. Simply put POWER is not in the same market as ARM, but in the same market as x86. Which means it is Intel (and AMD) who is killing them.

              • Ah, good point. Intel is way far ahead in low-power ILP/OOE, but they don't license their architecture, so yeah, there could be a market.

              • Simply put POWER is not in the same market as ARM, but in the same market as x86.

                Well, POWER covers a lot of ground. PowerPC is derived from POWER, and there have long been embedded PowerPC cores. The problem is that they worked on an old-world licensing model where each new product required a new license, and that license pretty much had to be negotiated with Motorola because that's who was doing low-power PowerPC.

          • Back when Apple first came out with PowerPC (the PowerPC 601 based 6100, 7100, 8100) the Ford EEC controllers were PowerPC 40x cores.

            We had "PowerPC Inside!" marketing stickers (I worked for the campus Apple Reseller) that i wanted to stick on random Ford cars for the hell of it. I never did, didn't want to screw up their paint jobs.

            Im sure there still are PowerPC applications in cars. And think of the 68000/DragonBall. A family that came out in 1979 but was sold until the 2000's as an embedded controller.

        • Re: (Score:1)

          by Anonymous Coward

          And of course, it's well-known that there's a tri-core PPC in the Xbox 360. There's also a castrated little PPC core in the front of the PS3's processor

          Whoa there. If you're going to defend PowerPC, you should acknowledge that the PPC core in the Xbox360 is the same core as in the PS3. The PS3's main work is done by the 7 vector processors (alternately called "APU"s, "SPUs", "SPC"s, depending on who's talking), but the CPU core is the same repipelined Power4 with a VMX unit on it. My memory fails me, but

          • (alternately called "APU"s, "SPUs", "SPC"s, depending on who's talking)

            SPE = Synergistic Processing Elements.

          • Whoa there. If you're going to defend PowerPC, you should acknowledge that the PPC core in the Xbox360 is the same core as in the PS3.

            Not only am I not defending PowerPC or POWER (I give a shit if it lives or dies, which hopefully will happen on its own merits, ha ha) but the PPC core in the 360 is not precisely the same as the one in the PS3, though they're based on the same design. The one in the PS3 is stripped down further and then glued to the vector units. The core in the 360 is faster, there's three of them, they are symmetric and they have slightly more features.

            • As far as I know, the only difference between the PPE and a Xenon core is that the later has a modified VMX (AltiVec) unit. They upgraded the vector register count to 128 per thread compared to 32 per thread on the PPE and they replaced a few vector instructions with others that are more useful in gaming.

              • As far as I know, the only difference between the PPE and a Xenon core is that the later has a modified VMX (AltiVec) unit.

                Well, in the Cell the SPEs are supposed to do the vector processing, so that makes sense. But as I've written here before, it's an extremely puzzling decision on Sony's part. Did they just believe some total bullshit from IBM about how great Cell would be? That would be fairly ironic given the impact of Sony's bullshit about how great PS2 would be on the Dreamcast, to add to the irony of following up a console for which developers complained about difficulty of development due to a wacky architecture (the P

        • Re: (Score:1)

          by Anonymous Coward

          And what has any of that to do with the POWER architecture?
          PowerPC != POWER!

          I thought this was a geek site. Then again, that was a looong time ago.

        • Spot on. The PPC core is well received in the embedded world. The ARM folk have powered up and moved into space once owned by PPC. With newer process the PPC could move down and put a lot of pressure on ARM.

          The 64bit ARM is a reach.

          Now if someone could build a five watt PPC Raspberry-Pi/BBB with 2x more RAM and GigE for five bucks less the landscape would shift in no time.

          • I do like POWER and PowerPC more than ARM. But ARM get the leg up by licensing to third parties. Still PowerPC still has the edge in higher performance (and power sucking) applications.

          • Now if someone could build a five watt PPC Raspberry-Pi/BBB with 2x more RAM and GigE for five bucks less the landscape would shift in no time.

            Realtek makes RTD1186 750mhz mips + PowerVR SGX 531 + 1Gbit + USB 3.0 + hdmi + sata + pcie, and it costs $3
            That doesnt mean shit because Realtek is one of those companies (like Broadcom) that never ever give out documentation.

            • Realtek makes RTD1186 750mhz mips + PowerVR SGX 531 + 1Gbit + USB 3.0 + hdmi + sata + pcie, and it costs $3

              Cheapest finished product on ebay is eighty bucks...

              • Chip is $3.
                Chinese are able to churn out $40 android tv sticks using $10 Rockchip chips, but are unable to make anything close to that price range using 3x cheaper Realtek - because Realtek hates open documentation.

                This is my point - availability is not enough, products wont just magically happen just because someone is selling cheap chips. Rasppi happened ONLY because people inside Broadcom invested their own time to be a buffer between bunch of closed source dicks and community.

        • But they've been losing a whole market segment that they had initially gained - at the expense of MIPS and Pentiums. Last time, they were there in Xbox 360, PS3 and Wii. Now, Xbox & PS3 are both going AMD, and I dunno about Nintendo. That's not good news for a platform that started out well w/ RS/6000 workstations, PowerMacs and PREP systems, then lost them, then won some game systems, then lost that, and is now there only in some high end IBM servers. Power.org does list some vendors who do make PP

      • Re:A Little Late? (Score:5, Informative)

        by bill_mcgonigle ( 4333 ) * on Wednesday August 07, 2013 @08:57AM (#44496763) Homepage Journal

        Are IBM hoping that people migrate to AIX or something? (good luck with that!)

        A few days ago on the Fedora homepage was announcement of the full release of Fedora 19 for IBM Power, presumably with Linux 3.10. You can get RHEL 6 if you want support and certainly there are debian and netbsd ports in various states. If there's a market for the hardware, the software is ready.

        • A few words about Linux technologies that originated from solid positions within the IBM camp...

          YUM (Yellowdog Updater, Modified)
          Yellowdog is a well-known RPM-based Linux distribution for the POWER architecture.
          JFS
          The OS/2 native filesystem was incorporated into Linux and released to production in June, 2001.
          NUMA
          IBM's acquisition of Sequent eventually led to NUMA code releases for the kernel which have been particularly appropriate for Hypertransport and QPI - high-performance Linux ows much to IBM.
          D
        • IBM actually promotes PowerLinux for server, datacenter & cloud solutions. IIRC, AIX is now just there for legacy sites that started off w/ it when it was the only option for the RS6k

      • Better late than never..... but better never late.

        Better late than pregnant

    • Wasn't POWER already an open architecture [power.org]?

  • Compiler support good for general PPC? (Score:1)

    by Anonymous Coward

    I recently got computing time on a BlueGene/Q (PowerPC A2) and I needed to run my C++ program on it. The compiler support was atrocious. It uses OpenMP for parallelization. GCC is damn slow on it and LLVM does not support parallelization using OpenMP. The IBM in-house compilers are crappy too for anything besides Fortran or baseline C.

    My question is: Is it like that also for the more "general purpose" PowerPCs? If yes, I really hope nobody licenses it. IBM supercomputers really do not deserve the TOP XXX ti

    • From the perspective of most supercomputer users, the problem here is that you want to use something which is not Fortran or C. Very little hardcore numerical stuff is done outside of those two languages. I'm not saying that's a good thing, but it is the reason why there's so little incentive to have good compiler support.

    • The compiler support was atrocious.

      Visual C++ is the best compiler I've used for PPC. (It's a shame that's not available outside of the 360 devkit).

    • You can get the 'advance [ibm.com]' toolchain. It's basically an upstream gnu toolchain, with a lot more optimizations and support for ppc chips.

      Looks like LLVM is getting improved [phoronix.com] as well.

    • The standard compiler for the PowerPC Architecture is xLC which is made by IBM.

      xLC is supported on AIX and Linux and it has all the optimization features needed for BlueGene/Q and other big iron (inter-procedural analysis, auto-vector optimizations etc...)

  • Nvidia... (Score:5, Funny)

    by MachineShedFred ( 621896 ) on Wednesday August 07, 2013 @08:22AM (#44496363) Journal

    Nvidia is very excited about the prospects of marrying Power processors and Nvidia GPUs for both HPC and general purpose systems.

    Nvidia hasn't quite figured out how to get their thermal energy per square centimeter to the level of a nuclear reactor, so I'm sure opening up the POWER series of chips has them quite excited on that front.

    • Kirk: Do you think it will work?
      Spock: It will depend on what Mr. Scott can coax out of the systems.
      Kirk: Scotty, Spock thinks that if we can boost the precision of the sensors and overlay the data on the navigation computer we may be able to navigate through the interphase rift to escape the Tholian web. Can you do it?
      Scotty: Aye Captain. With that last maintenance overhaul at Star Base 11 our computers were updated with the new Multitronic GPU processors. For once I have the power.

      Power, the next front

  • ...that spy^H anonymously profile your behaviour at the microcode level? I'll pass, x86 SMM is already evil enough for me.
    • If you wanted to subvert a processor in that way, it would be easiest to add a secret knock that would allow the attacker to run any code at all on it. And any CPU you buy can have such a secret knock built into it. I don't see any particular reason to trust one CPU manufacturer more than another. If you have some evidence that one is untrustworthy, please share it.
      • My comment was tongue-in-cheek. All current mainstream processors already have operating modes that allow the firmware to run code without the OS knowing (or having a way to know). But then no CPU vendor is, yet, in the market of collecting and analysing personal information.

  • I use a lot of IBM software and hardware on a daily basis. I /really/ feel like this is more of a 'corporate alliance' than an 'opening up' of their 'intellectual property."

    I guess I just dislike the fact it's called the "OpenPower Consortium". Somehow I feel it dilutes the word "open", which has a lot to free/libre.

    KPH

    • Somehow I feel it dilutes the word "open"

      You're thinking about it the wrong way. Consider instead:

      IBM are open to the idea of taking your money.

      That is the fundemental idea of openness behind it.

    • no, it's more like the open API of Unix, open specs of Sparc. IBM finally doing something Sun had success with decades ago but not useful marketing ploy now. Too little way too late.

  • I can't wait for the return of Motorola Starmax, Umax Supermac and Power Computing's Power Tower Pro. I remember my Power Tower Pro was upgradeable to 1 GB of ram in 1997! Shut up and take my money!

    • Great hardware until Apple killed them.

    • Actually, when Apple killed that business, they should have ported BeOS to those boxes, and sold them. At that time, they still had a chance, given that Apple was still running System 7 based OSs on the PowerMacs. Instead, they simply folded completely. Another disappointment was Be discontinuing the BeBoxen

  • Power Licensed (Score:5, Informative)

    by LoRdTAW ( 99712 ) on Wednesday August 07, 2013 @08:42AM (#44496583)

    The headline and summary are confusing, Power is licensed and Power based chips are produced by third parties. Applied Micro (AMCC) along with Freescale make power core based CPU's/SoC's for embedded use and Xilinx has power cores in their high end Virtex 5 FPGA's. A-EON uses the AMCC Power CPU on mATX motherboards for modern Amiga systems. What they mean is that IBM is making it easier for others to license and adopt Power for their needs. Though the Gamecube, Wii, Wii-U, Xbox 360 and PS3 use power processors, they are all made by IBM like the Apple Power CPU's.

    Its good to see more RISC architectures that have been around for a while becoming more popular. The mobile market pretty much bought RISC back into the spotlight and is giving x86 a run for its money. And more interesting are the partners and the task Power is looking to solve: the cloud (I feel dirty using that phrase). Intel better watch out, with everyone pushing software as a service and mainfr^H^H^H cloud computing, companies are looking to create hardware targeted towards those tasks while also reducing power.

    • Re: (Score:2)

      by afidel ( 530433 )

      Is this about POWER or PowerPC? There's a significant difference and I seriously doubt IBM is going to risk that sweet, sweet large system revenue by allowing others to produce POWER based CPU's.

      • Re: (Score:3)

        by LoRdTAW ( 99712 )

        Its confusing because different Power architecture versions may support both PowerPC and Power ISA's or PowerPc or Power only. Power 1/2 evolved into PowerPC which was renamed to Power ISA. Both Freescale and AMCC call their processors Power processors and support the Power ISA v.2.03 spec which also supports PowerPC. Newer Power ISA versions are called both PowerPC and Power, e.g. CPU's which comply with Power ISA v.2.05 are called POWER6 and the PowerPC 476. The latest power spec, Power ISA v.2.07, does n

    • I believe the intention is to make the high-end POWER chips more ubiquitous in the server room - heavy duty RISC/Unix(Linux) server platform. The intention is to squeeze x86 out of the datacenter with AMD systems at the low end, and POWER-based gear for the serious number crunchers.

      IBM hopes that by bringing competitors into their platform, they can use economies of scale to make their systems more cost-competitive, and name recognition to separate themselves from the other POWER platform providers. Reduce

  • POWER support is dead on all enterprise Linux distributions, Red Hat dropped support with EL5. Furthermore OpenPower boxes are contractually prohibited from running AIX.

    You've got a box of hardware with nothing to run on it and it can only deliver half the performance of comparatively priced Intel equipment. If you outsource support to IBM, their support specialists in the delivery centers will accidentally nuke your whole frame during routine maintenance, and you could be down for days. If you can manage

    • WRONG! (Score:5, Informative)

      by Funk_dat69 ( 215898 ) on Wednesday August 07, 2013 @09:46AM (#44497359)

      Impressive. You are wrong on just about *everything* you wrote:

      >>POWER support is dead on all enterprise Linux distributions, Red Hat dropped support with EL5.
      Nope [redhat.com] and nope [fedoraproject.org] and nope [novell.com]

      >>Furthermore OpenPower boxes are contractually prohibited from running AIX.
      You are confusing this announcement with a previous attempt at the Linux market that was also called OpenPower. Those systems only ran Linux and could not run AIX. This announcement is about opening up the entire platform and licencing out parts or whole cores of the actual high end chips to companies like Google, who recognize that the single most expensive component in servers is the CPU - and they want choice and customization.

      >>You've got a box of hardware with nothing to run on it and it can only deliver half the performance of comparatively priced Intel equipment.
      The recently released Power7+ chip running Linux is the fastest [sap.com] thing [spec.org] on [spec.org] the market right now.

      >> If you outsource support to IBM, their support specialists in the delivery centers will accidentally nuke your whole frame during routine maintenance, and you could be down for days
      Umm..ok I'm stopping now

    • Hey, I still have some Mac System 9 floppies lying around. This should run those, right?

  • You want to start a revolution, make the damn equipment simple and available.

    Build it and they will come.

  • It's loo late, dude.
    25 years too late.
  • IBM and company's use of variously-capitalized forms of "Power" can be a bit confusing. When the RS/6000s first came out, IBM described the instruction set architecture they implemented as POWER, for "Performance Optimized With Enhanced RISC"; see the "IBM POWER Instruction Set Architecture" Wikipedia article [wikipedia.org] and its references. Starting with the second-generation RS/6000 processor, they started naming the processors "POWERn" as well.

    PowerPC was an instruction set architecture based on the POWER ISA; a f

    • Power.org is responsible for the Power Architecture ISA. (ie architecture)
      OpenPOWER will be the consortium which will license out the actual POWER Cores and associated peripherals.(ie microarchitecture and implementation)
      • Why not have just one org - power.org, w/ its pithy URL - handle all of that? Have divisions within that org handle the stuff, if you need

  • Anyone want to surmise whether we'll get a desktop machine anytime soon?

    Quite fancy a 5Ghz desktop beast running Amiga OS 4.

    Just imagine - Full - motion - video. Less than 0 second shutdowns. Deluxe paint loading quicker than you can thumb a floppy in.

    Or you could run ubuntu and have the dash load up in the time-frame your short-term memory works in.

    D

    • My 1st thoughts were to the Amiga.

      . . and while this news makes me hopeful, as an Amigain, I have learned not to get disappointed :)

Slashdot Top Deals

A committee takes root and grows, it flowers, wilts and dies, scattering the seed from which other committees will bloom. -- Parkinson

Close