Slashdot Mirror


48 Core Vega 2 in the Making

TobyKY76 writes to tell us The Inquirer is reporting that upstart Azul Systems is planning to integrate 48 cores on their next generation chip. From the article: "The first-generation Vega processor it designed has 24 cores but the firm expects to double that level of integration in systems generally available next year with the Vega 2, built on TSMC's 90nm process and squeezing in 812 million transistors. The progress means that Azul's Compute Appliances will offer up to 768-way symmetric multiprocessing."

206 comments

  1. Oh wow!! by liliafan · · Score: 1

    Oh my God, 768 way processor!!! I almost had a fit when I saw suns http://www.sun.com/smi/Press/sunflash/2006-03/sunf lash.20060321.3.xml'>t1 processor I need to sit and catch my breath, this is a multithreader's wet dream. Hmmm I wonder how oracle will price running on this processor?

    --
    GeekServ Unix Consulting Services (http://www.geekserv.com)
    1. Re:Oh wow!! by AuMatar · · Score: 2, Funny

      If you have to ask, you can't afford it.

      --
      I still have more fans than freaks. WTF is wrong with you people?
    2. Re:Oh wow!! by SeeMyNuts! · · Score: 1

      Huh? Only 768? Bah!

      I will be designing and selling an Earth Simulator on one chip. Only $50,000,000,000...paid in advance.

    3. Re:Oh wow!! by liliafan · · Score: 2, Interesting

      lol lol, hmmm $40,000 per processor / 2 for virtual processors $20,000 * 768 = $15,360,000 licence fee per chip, no wonder Larry isn't worried about the sales of oracle stagnating, as soon as people upgrade from their old duel core to these he beats billy boy in the billionaires list :op

      --
      GeekServ Unix Consulting Services (http://www.geekserv.com)
    4. Re:Oh wow!! by AKAImBatman · · Score: 1

      You're far too trusting. Remember, this is the same "online journal" that printed the Google Office Confirmed scoop right after a press conference that announced nothing of the sort.

      The Inquirer is taking Azul's word for it at the moment, which is probably why the article is so light on details. About a billion questions pop in my mind when I hear a story like this. The only answer I get is that, "Sun is banging on Azul's door for IP infringement."

      Sure.

      Does anyone have any real info on these guys? About all I can find is an unsourced Wikipedia article titled Network Attached Processing. It is also lacking in details. (A shill, perhaps?)

    5. Re:Oh wow!! by MindStalker · · Score: 1

      If you've ever read oracles policy on per processor license.
      http://www.oracle.com/corporate/press/2005_dec/mul ticoreupdate_dec2005.html

      You would see that it depends upon your architeture. For example when running on UltraSpac you pay for one processor for every 4 cores you have. AMD/Intel multicore you pay for every 2 cores you have. Either way they would have to sit down and devise a per processor license for you. Though you can always purchase per user instead of per processor. This would probably be the best route if your runnings hundreds or thousands of processors.

    6. Re:Oh wow!! by advocate_one · · Score: 1
      Oh my God, 768 way processor!!!

      woo hoo... at last... something that has processors left over after allocating one to every task running on this Kubuntu box... ;)

      --
      Donald 'Duck' Dunn: We had a band powerful enough to turn goat piss into gasoline.
    7. Re:Oh wow!! by br0k_sams0n · · Score: 1

      Azul sells specialized processors for high-end Java applications (last I looked). You can't just slap any binary you want on these things and run it. All apps run inside a proprietary OS inside their boxen, there's no installing what you want here, just uploading jar files.

    8. Re:Oh wow!! by networkBoy · · Score: 1

      Though you can always purchase per user instead of per processor. This would probably be the best route if your runnings hundreds or thousands of processors.

      That assumes you don't also have 80K+ users :-)
      -nB
      --
      whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
    9. Re:Oh wow!! by bfischer · · Score: 2, Funny

      "duel core"? Are they fighting or something? They will not share the work very well if they are fighting. Maybe you should give them a time-out or something.

    10. Re:Oh wow!! by CliffClick · · Score: 1

      I work for Azul - and yes the chip is real.

    11. Re:Oh wow!! by AKAImBatman · · Score: 1

      Cool. Now how about some details? MHz, MIPS, FLOPS, instruction set, JVM version support, OSes ported, what exactly "Network attached processing" services are, etc, would all be nice to know. Is this just a really fancy DRMAA machine, or what?

      Inquiring minds want to know! :-)

    12. Re:Oh wow!! by CliffClick · · Score: 1

      We've (Azul Systems) have long been aware that this is an issue and have reasonable liscencing worked out.

      (my first SlashDot posting & I'm learning about posting-throttling)

    13. Re:Oh wow!! by CliffClick · · Score: 1

      Most of this stuff is proprietary.
      Chip is our own design, our own ISA, etc.
      JVM: it's a port of Sun's HotSpot, v 1.4.2 & v 1.5.0
      OS: doesn't really matter - you can't see the OS on the box, and we work with most host OS's.

      - and what's a DRMAA machine?
      (caught in JFK waiting for a flight, so very soon no email...)

      Cliff

    14. Re:Oh wow!! by dotgain · · Score: 1
      I work for Azul - and yes the chip is real.

      Are you getting a kick out of some of these replies?

    15. Re:Oh wow!! by AKAImBatman · · Score: 1

      DRMAA is a standard for distributed computing. By your response, I take it that your hardware isn't intended for that function. :-)

      So what exactly is "network attached processing?"

      BTW, thanks for responding. Sorry to bug you during your relaxing downtime in JFK. ;-)

    16. Re:Oh wow!! by Anonymous Coward · · Score: 0

      Or you can always just download Postgres for free.

    17. Re:Oh wow!! by somersault · · Score: 1

      I'm sure the competition will result in better performance ;)

      --
      which is totally what she said
    18. Re:Oh wow!! by liliafan · · Score: 1

      eh dyslexia is a bitch

      --
      GeekServ Unix Consulting Services (http://www.geekserv.com)
  2. Take that pop ups! by fernandoh26 · · Score: 1

    Now I can finally deal with the 768 or so pop ups that come up whenever I attempt to use my computer!!

    --
    Chums up, let's do this!
    1. Re:Take that pop ups! by paintballer1087 · · Score: 1

      that's what adware protection's for... and using linux instead of windows will also help

    2. Re:Take that pop ups! by Anonymous Coward · · Score: 0

      Tomahawk Desktop is highly recommended.

  3. AutoCAD by turtleAJ · · Score: 5, Funny

    Behold the power of copy/paste!


    Yeah, yeah... my Karma is SUPER negative...

    1. Re:AutoCAD by pnatural · · Score: 1

      Um... the correct autocad command would be array.

      Sorry for being one of those lame-asses who corrects others jokes. It just rankled me a bit to think of a million copy and pastes --- array is a nice command if you're a drafter. the polar array option comes in handy more often than you might think.

  4. 48 cores is a nice start, but.... by Raul654 · · Score: 4, Informative

    I know of a certain project that's working to put over a million cores into a system (160 into a single chip), and it should be finished and available off-the-shelf within a year or so.

    --


    To make laws that man cannot, and will not obey, serves to bring all law into contempt.
    --E.C. Stanton
    1. Re:48 cores is a nice start, but.... by joe545 · · Score: 1

      The wikipedia link itself says 80 cores per chip not 160.

    2. Re:48 cores is a nice start, but.... by Raul654 · · Score: 1

      No, each thread unit (TU) is a core; there are two cores per processor. (and 80 processors per chip)

      --


      To make laws that man cannot, and will not obey, serves to bring all law into contempt.
      --E.C. Stanton
    3. Re:48 cores is a nice start, but.... by questionlp · · Score: 1

      Rather: 80 cores per processor package and 2 threads per core.

      That equates to: 160 threads per processor package.

      For instance: even though the Montecito Itanium is capable of four-threads per processor package, it is only a dual-core (each core is capable of two threads).

    4. Re:48 cores is a nice start, but.... by Raul654 · · Score: 1

      Don't call it "processor package" - it has 80 processors inside of a single integrated circuit, which is called a chip. When the chip is connected with external devices, it's called a board (physical view) or node (logical view).

      --


      To make laws that man cannot, and will not obey, serves to bring all law into contempt.
      --E.C. Stanton
  5. Finally! by lax-goalie · · Score: 4, Funny

    Enough CPU power that even Microsoft Office will run with a little pep!

  6. Are they x86 compatible? by Stoutlimb · · Score: 1

    I read the article, and I was a little confused on that point. I would think that they are definately not.

    1. Re:Are they x86 compatible? by LLuthor · · Score: 2, Informative

      They are not x86 compatible. They are RISC like chips with an instruction set optimized for for running VM based applications like Java and .NET.

      That said, its still very impressive to get that many cores working together, though not as impressive as x86.

      --
      LL
    2. Re:Are they x86 compatible? by Anonymous Coward · · Score: 0

      It will emulate x86!

    3. Re:Are they x86 compatible? by WilsonSD · · Score: 1
      They're not x86 compatible. They're special purpose chips for Network Attached Processing devices for driving up utilization on Java applications. For another approach to driving up utilization on server-side Java applications you should check out Cassatt.


      Cassatt was discussed on Slashdot a few weeks ago.


      -Steve

    4. Re:Are they x86 compatible? by Anonymous Coward · · Score: 0

      ...using all 48 cores

  7. Not for nothing... by Red+Flayer · · Score: 3, Insightful

    But I'd tend to take a website's articles with a grain of salt when the links at the bottom of the page are:

    "Home Discuss on our Forum Flame Author

    Recommend this article Print"


    Sounds to me like someone issued a press release and wants a share of the excess VC floating around... and the Inquirer took the bait. They did a good job of not loading the buzzwords, though -- they didn't say they would 'leverage their experience with graphics chip design' or anything like that.

    I'd expect this company to turn around and sell out to AMD or Intel at the earliest opportunity, if given the chance.

    --
    "Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
    1. Re:Not for nothing... by CTho9305 · · Score: 1

      The Inquirer is generally regarded as a somewhat-reliable rumor mill that publishes a lot of tech secrets from leaked documents, plus general industry news.

  8. The next generation PC? by Anonymous Coward · · Score: 0

    Will there finally be a PC which can actually meet Windows Vista system requirements? Sign me up!

    (posting as AC because n00bs with mod points lack any sense of humor)

  9. Finally! by robyannetta · · Score: 1

    768 cores. Finally, a box that will run Everquest II.

    --
    - Just my $0.02, take with a grain of salt, your mileage may vary.
  10. Upstart? by Anonymous Coward · · Score: 0

    "...upstart Azul Systems."

    The editors are so useless they can't even copy and paste... TFA calls them "startup Azul Systems"

    1. Re:Upstart? by Anonymous Coward · · Score: 0

      Both terms exist, and are correct in the context that they are being used in.

      http://dictionary.reference.com/search/?q=upstart

  11. No, not clusters, but .... by Anonymous Coward · · Score: 0

    ... video cards.

    Given the super pace that ATI and nVidia are going at, how long before they adopt this tech for their video cards. Heck, their top end cards have more ram, suck more juice, and cost more than most pc's anyway. I can just see some ATI guys drooling over the opportunity to deliver a 512 core (x2 with SLI) video card with 2GB RAM. Of course it'll need it's own separate power supply and cooling tower, but that's another issue and the game freaks will buy it anyway so they can run their favourite game at 1.5x frame rate AA.

  12. Re:imagine... by Anonymous Coward · · Score: 0

    A minute earlier - and you could've been first to make that tired joke - but YOU FAIL IT!

      by BigZaphod (12942) Wednesday March 29, @06:42AM (#15013114)
    Imagine a beowulf cluster of these!

      by pxuongl (758399) Wednesday March 29, @06:43AM (#15013117)
    imagine a beowulf cluster of these guys...

  13. 768 cores, why? by Coopjust · · Score: 2, Insightful

    Dual cores, quad cores, whatever, I can understand that for multitasking and programming. But 768 cores? What would possibly use that many cores? And for any single task, the thing would not be efficient. What exactly is the point of this? Bragging rights?

    1. Re:768 cores, why? by ltwally · · Score: 1
      "...But 768 cores? What would possibly use that many cores?..."
      If you'd RTFA, you'd note that they're building chips with 48 cores, not 768.

      To answer your question, anyways: More cores == better efficiency == less heat == lower electric bill. Desktop users may not need a 48-core chip (yet), but server farms love designs such as this.

      --



      /dev/random
    2. Re:768 cores, why? by fayd2003 · · Score: 1

      Think virtualization ....

    3. Re:768 cores, why? by Anonymous Coward · · Score: 0

      It'd be used for servers, mostly. Webbservers like to devote one thread per request -- this processor, it seems, could handle 768 simultaneous requests from users. Useful for when you get slashdotted.

    4. Re:768 cores, why? by thrillseeker · · Score: 3, Insightful
      What would possibly use that many cores?

      Any task that can be split into multiple processes. An example is an array of data, where a single algorithm is going to be applied to each element. An array of data can represent anything - an image, or stock prices, or DNA, etc.

      for any single task, the thing would not be efficient.

      It depends on what you really mean by a single task - a given process consists of multiple sequential tasks, where a task may be as fine-grained as a single CPU operation, or perhaps due to overhead of communication between tasks, a tuning effort can be made to say a "task" is some multiple of operations.

    5. Re:768 cores, why? by Carnildo · · Score: 1
      But 768 cores? What would possibly use that many cores?


      Think "real-time raytracing". Think computer games with real shadows from multiple lights, with curved mirrors, with glass that actually bends light.

      Think about a first-person shooter where you can notice someone sneaking up behind you when you spot his reflection off the hood of the car you're using for cover.
      --
      "They redundantly repeated themselves over and over again incessantly without end ad infinitum" -- ibid.
    6. Re:768 cores, why? by x2A · · Score: 1

      "What exactly is the point of this?"

      Simulations, of weather, bio/chemical/nuclear reactions, searching for extra terrestrial signals in the radio spectrum (okay, kidding about that one)... sciency stuff :-p

      --
      The revolution will not be televised... but it will have a page on Wikipedia
    7. Re:768 cores, why? by makapuf · · Score: 2, Funny

      yeah, frankly, 640 cores should be enough for everybody, no?

    8. Re:768 cores, why? by Alef · · Score: 1
      What would possibly use that many cores? And for any single task, the thing would not be efficient. What exactly is the point of this?

      Remember, the human brain has 100 billion cores, each in itself very inefficient, and yet it is pretty powerful.

      There are huge amounts of unexploited parallelism in the tasks our computers perform. The problem is mainly that most of the tools we use (programming languages etc.) are very serial in how they describe and handle problems and solutions. This is natural, of course, since it reflects how computers historically have been operating. But there are limits on how much computational performance one can focus into a single computational unit, and we are reaching them. So expect to see more development in this direction.

    9. Re:768 cores, why? by njvic · · Score: 1

      Beowulf.

    10. Re:768 cores, why? by Anonymous Coward · · Score: 0

      To reply to this, I will repeat something that was once said, in the not too distant past: "Who would ever need a 1 gig hard drive? Who could ever fill ALL THAT SPACE up? What is it, bragging rights?"

      Just to put that into even worse perspective, I'm carrying that much around my neck, and feeling like I should quadruple it to actually handle the things I need it for.

      Here is another quick quote: "Build it, and they will come!"

    11. Re:768 cores, why? by super_bob711 · · Score: 1

      ... heh hook a couple of these in a cluster :D might as well grab a couple petabytes of storage and become the internet..

    12. Re:768 cores, why? by Splab · · Score: 1

      High performance computing...

      Also try looking up c++csp, jcsp or occam (kroc), since we cant go any faster, we go parallel...

    13. Re:768 cores, why? by oddaddresstrap · · Score: 1

      Bill Gates: "768 cores is more than anyone will ever need".

    14. Re:768 cores, why? by Jozer99 · · Score: 1

      Ever used SETI or UD? Those are running on hundreds of thousands or even millions of processors. This is just faster.

    15. Re:768 cores, why? by timeOday · · Score: 1
      What would possibly use that many cores?
      The simple answer is, "anybody who currently has a cluster or server farm, plus those who would like to if only they were cheaper and more power efficient."
    16. Re:768 cores, why? by ChrisA90278 · · Score: 1
      Many servers (http and databses) use one process per client. It's not uncommon to see 20 copies of Apache running on a web server. I typically see a few copies of ostgreSQL running on my system.

      for desktops video rendering can uses as many CPUs as you can get. Many desktop apliations are now written to tae advantage of multiple CPUs. Check out Apple's iTunes. I've seen iTunes run a dozen threads.

      huge numbers of CPUs, as in "thousands" will likely be required for true AI.

    17. Re:768 cores, why? by SirKron · · Score: 1

      The chip is designed for VM type apps. I would love to see vmware ESX ported to it, it would certainly use it.

    18. Re:768 cores, why? by TummyX · · Score: 0, Redundant

      This co-processor is specifically designed to accelerate JVMs. Specifically, they've modified sun's hotspot java vm to use their co-processor. Think about massive Java appservers where you've got thousands of concurrent users with each user using one or more threads. The accelerater would allow the application and business logic from many users to be, physically, processed in parallel.

      The points about graphics and other algorithms made by others are valid but aren't *really* the target market for this chip when you look see that the company's partners are java app server vendors (BEA etc).

  14. Chevy Vega Returns! by Anonymous Coward · · Score: 0



    Way to go Vega! Watch those cylinder sleeves if you got 'em!

    1. Re:Chevy Vega Returns! by Bimo_Dude · · Score: 1
      48 cores and 2 doors

      What a beauty!

      --
      "Teleporting Rodents with D-Cell Battery Displacement" theory -- IgnoramusMaximus (692000)
    2. Re:Chevy Vega Returns! by Lawrence_Bird · · Score: 1

      vega always rocked pinto's world.. and vega cpu upgraders were the best

    3. Re:Chevy Vega Returns! by Duhavid · · Score: 1

      Imagine the core charge....

      --
      emt 377 emt 4
    4. Re:Chevy Vega Returns! by miller701 · · Score: 1

      Didn't the signals in the movie and book "contact" come from Vega. There's that one scene where the come back to the array and there's the Chevy Vega fan club.

    5. Re:Chevy Vega Returns! by Anonymous Coward · · Score: 0

      Alas, there were no cylinder sleeves in the aluminum block of the Vega. They relied, instead, on surface hardening of the aluminum cylinder walls. It didn't work very well.

  15. Would'nt wanna... by Revellion · · Score: 3, Funny

    Would'nt wanna see how htop looked on that possible SMP setup... I would have to page down atleast 80-100 pages of per-processor load meters before i got to the processes list :S

    --
    htop(top on stereoids): http://htop.sf.net
    1. Re:Would'nt wanna... by BlahBlech · · Score: 1

      well, it could be displayed as an N x M grid of colors where each cell is more green for small utilization to more red for large utilization. then you can assess the system as a whole.

    2. Re:Would'nt wanna... by psbrogna · · Score: 1

      I expect the process list will have to be kept on a high end external database server so that reasonable performance metrics can be collected and monitored.

  16. Blade servers by Seanasy · · Score: 4, Funny

    So, chip manufacturer's have adopted the Gillette approach to marketing chips. I guess it was inevitable after they went from one core to two. The only difference, I expect, is that they'll increase by powers of 2. Soon, we'll have a Intel Mach 512 Core Sensor Extreme or something :P

    1. Re:Blade servers by p!ssa · · Score: 0

      YES AGREED! And it will have the one single core on the back side that will actually be useful.

    2. Re:Blade servers by reverend_rodger · · Score: 0

      But will each new processor warrant its own brand new type of thermal paste, keeping in the same line of Gillette? "New Intel Mach 512 Core Sensor Extreme Gel! Reduces your processor's temp by a tiny fraction of a degree, if that, and smells great!"

    3. Re:Blade servers by Anonymous Coward · · Score: 0

      then we can have little cores that vibrate so you can code closer to your skin

    4. Re:Blade servers by Anonymous Coward · · Score: 0

      Intel Mach 512 Core Sensor Extreme, now available in manual or battery powered vibrating edition!

    5. Re:Blade servers by Criffer · · Score: 1

      Which only serves as further proof of the approaching singularity.

      Only nine years away. Can't wait till I get my hoverboard.

    6. Re:Blade servers by Anonymous Coward · · Score: 0

      Powers of two don't make as much sense as the sequence of squares: 1, 2, 4, 9, 16, 25, 36, 49, ...

  17. The wiki link says 80 not 160 - read by joe545 · · Score: 4, Informative

    "Each 64-bit Cyclops64 chip (processor) will run at 500 megahertz and contain 80 cores." While it may have two threads per core, that is not what you claimed. You stated "...that's working to put over a million cores into a system (160 into a single chip)". 160 threads per chip, yes, but not 160 cores.

    1. Re:The wiki link says 80 not 160 - read by Raul654 · · Score: 1

      Yes, I noticed that after I posted my reply. I've fixed the wikipedia article accordingly (I wrote most of it). When I was writing it originally, I got processors and cores confused

      --


      To make laws that man cannot, and will not obey, serves to bring all law into contempt.
      --E.C. Stanton
  18. I don't know much about CPU internals but by rolfwind · · Score: 1

    It would seem to me, that a CPU's workload is roughly limited by the number of transistors it has multiplied by it's MHz speed. No matter how many cores one has, the transistor count should remain roughly the same for a 1-core, 2-core, 8-core chip of the same nm process and is limited by that process (90 nm in this case).

    I would suppose (but am not sure) extra cores reduce the number of transistors being idle at any one moment. The downside would seem that each extra core reduces the capability to process highly sequential problems in favor of highly paralell problems.

    I know extra cores are nice to an extent, but isn't there a point where the paralellized gains aren't worth reduction in individual core capability? So these chips may be great for networks (routers, etcetera) but not so nice in desktops.

    Are have I missed something completely?

    1. Re:I don't know much about CPU internals but by Anonymous Coward · · Score: 0

      Yeah, you did miss something. You're crazy! absolutly nuts. WTF are you talking about??!?!?! Hmm?

      I'm high... way too high.

    2. Re:I don't know much about CPU internals but by AKAImBatman · · Score: 4, Informative
      It would seem to me, that a CPU's workload is roughly limited by the number of transistors it has multiplied by it's MHz speed.

      The number of transistors can go up for a variety of reasons. Chief among them is designs that utilize complex performance enhancements. To name a few:

      • Superscalar processing
      • Branch prediction
      • Hyperthreading
      • Out of order instructions
      • Pipelining


      The secondary source of transistor usage is coprocessors like Floating Point Units and SIMD Units.

      The latest craze in processor design is to simplify the microprocessor back down to the most basic level. From there, the processors are ramped up through shear numbers of parallel pipelines (i.e. threads) and cores as opposed to ramping up the individual CPU horsepower. These multi-core chips typically share coprocessors among a pipelines or cores, and may even have entire cores dedictated to specific tasks like SIMD. As a result, a properly designed program will be able to execute within a very short period of time, thanks to the parallel nature of the multi-core architecture.

      Now the only problem is in finding these "properly written programs".
    3. Re:I don't know much about CPU internals but by AndyG314 · · Score: 0

      It really depends on what a "dual core" chip really is, if you have 720 little processors each with everything they could need, then your right the total processing power is not improved that much, but if you start to share bits between cores (Tomasutra (sp?) could tell you more on this) then you can start to reap big savings. By shavring multipliers, dividors etc... you can reap big savings. Every single core computer has a dividor, but how often does it use it? By sharing bits between multiple cores you can reduce the number of currently unused transistors and increase your overall throughput of instructions.

      --
      If it's dead, you killed it.
    4. Re:I don't know much about CPU internals but by Sebastopol · · Score: 1

      properly written programs

      Almost.

      Sooner or later you have to branch, or speculate, or go to main memory (or *gasp* I/O). So exploiting parallelism in arbitrary code is quite a challenge. Having worked on optimizing the linpack benchmark for a small "supercomputer" (64 CPUs) during college, it is quite a bit of work to decompose a problem this way.

      However, the trend is not "well written" parallel apps, but rather parallel process scheduling: like antivirus + network stack + device drivers + MP3 player + excel + VNC + compiler... all getting their own processor core.

      That's the future, baby.

      --
      https://www.accountkiller.com/removal-requested
    5. Re:I don't know much about CPU internals but by A+Commentor · · Score: 1
      I don't know much about CPU internals but
      So if you don't know CPU internals, why make these statements:
      It would seem to me, that a CPU's workload is roughly limited by the number of transistors it has multiplied by it's MHz speed.
      - NO, number of transistors has nothing to do with it.
      No matter how many cores one has, the transistor count should remain roughly the same for a 1-core, 2-core, 8-core chip of the same nm process and is limited by that process (90 nm in this case).
      - NO, transistor count WILL increase with the increase of the number of cores.
      I would suppose (but am not sure) extra cores reduce the number of transistors being idle at any one moment.
      - NO, more transisters will be idle since each core will have it's own set of idle transistors to add to the total number of idle ones.
      The downside would seem that each extra core reduces the capability to process highly sequential problems in favor of highly paralell problems.
      - NO, not for the individual core, only if the cores are competing for shared resources such as the memory or some I/O bus.
      I know extra cores are nice to an extent, but isn't there a point where the paralellized gains aren't worth reduction in individual core capability?
      - Hardware-wise, NO, each core will have the same capabilities, although as stated above may have to compete with other cores for shared resources.
      So these chips may be great for networks (routers, etcetera) but not so nice in desktops.
      Are have I missed something completely?
      - YES, totally.
      --

      Looking for any old 8-bit Heathkit/Zenith software/hardware - http://heathkit.garlanger.com

    6. Re:I don't know much about CPU internals but by Anonymous Coward · · Score: 0

      Yup, with 3 caveats:
      a) if you can't fill all the individual execution units (adders, multipliers and so on) with are parallel, then it's more like the frequency times the square root of the transistor count (you're wasting "width"); hence parallel/out-of-order execution, hyperthreading, SIMD... But we kinda reached a wall here.
      b) you can't arbitrarily increse the "depth", with the same result (square root again); hence superscalar execution, pipelining and so on. These have the advantage of also allowing greater clockspeed (you so less in one step, so you need less time to go through it). But there's a problem: if there's a mispredicted branch, you just wasted loads of power, and there's limits on how good branch prediction can be. Wall there too.
      c) as you increase width and frequency (depth), you need to feed more data to the processor, so you need ever bigger caches (that now gobble up much more die space than the core itself). Problem: caches need to predict what to fetch and what to flush (altho you can help the proc), and you got the same problem as with branches.

      So the current solution to these problems is to add cores and get the programmer to do more parallelism himself.

      There's several schools:
      - AMD/Intel, Power, Sun's "Rock" use several standard cores: you get the same performance that you used to with one core core (except for the freq reduction to deal with heat), programmers don't have much to relearn, but you still have big caches, big misprediction penalties and all the plumbing for the internal parallelism, so you're wasting loads of potential power (like just how many apps use SIMD, really).
      - this here, like Sun's Niagara (or the XBox 360's proc, for what matters) have many more, but much simpler cores. You don't need big caches or heavy branch prediction, and you drop out-of-order execution (you hope that if one thread is blocked on memory, the other can run). Advantage: IF it's used properly, much more of your transistor count * frequency budget is used, peak and on average. Problem, you need those programs that use it properly.

      The jury's still out on what the best mix is; as of now, servers like multicores, but there's no desktop apps.

      I'll just mention that while it's popular to bash multicores on the desktop (altho less than when Intel dropped the P4), there's *loads* of parallelism in desktop tasks as well: spreadsheet recalc (obvious), document search & replace or reformatting (do it per page), and so on. It's just that at of this moments, that's not how they're written...

    7. Re:I don't know much about CPU internals but by Anonymous Coward · · Score: 0

      > Now the only problem is in finding these "properly written programs".

      Simple, make all the improperly written programs fight to the death using glow-in-the-dark frisbees....

    8. Re:I don't know much about CPU internals but by rolfwind · · Score: 1

              No matter how many cores one has, the transistor count should remain roughly the same for a 1-core, 2-core, 8-core chip of the same nm process and is limited by that process (90 nm in this case).


      - NO, transistor count WILL increase with the increase of the number of cores.


      But it would seem to me, that for the same sized chip, regardless of number of cores, the process (90 nm in this case) limits the number of transistors.
    9. Re:I don't know much about CPU internals but by Anonymous Coward · · Score: 0

      You don't have the faintest idea what you're talking about.

      The grandparents is completely right, the amount of work a proc can do is, in absolute, the number of transistor it has times the number of times per second you can get them to do something.

      The transistor count is only limited by process. If you go for simpler cores (like in the article, FOR INSTANCE), you'll get more cores with the same transistor budget.

      One of the idea of having simpler cores is that you waste less space on "non processing power" task, i.e. cache, branch prediction, out-of-order execution and so-on - even tho that's not the way AMD, Intel and others are doing it (largely because they have no choice, by the way, existing apps must run *fast*). So yes, you'll get less "idle" transistors.

      Of course, if you have simpler cores, then you usual sequential app will be slower. Not to mention that multicores tend to drop frequency to help with heat. Did you check the frequency of the X2s when they were released?

      Like I mentioned, using cores with similar capabilities to the one we had in unicores is not the only way. Do you think these guys have 24 cores (in Vega) with full branch prediction, superscalar and out-of-order execution? Hell even the 360's processor doesn't, and it has only 3!

    10. Re:I don't know much about CPU internals but by A+Commentor · · Score: 1
      But it would seem to me, that for the same sized chip, regardless of number of cores, the process (90 nm in this case) limits the number of transistors.
      Correct, that would set the upper limit if everything was constant. But the article does not state what process was used in the first-generation chip:

      The first-generation Vega processor it designed has 24 cores but the firm expects to double that level of integration in systems generally available next year with the Vega 2, built on TSMC's 90nm process and squeezing in 812 million transistors.

      The way I read this statement is that they are moving to the 90nm process and will be able to squeeze more transistors on the chip because of that. The word 'squeezing' implies to me that they are using more transistors than the first chip. Also, the article does not state that the size of the chip would not increase. The actual silicon of most chips are much smaller than the package size, thus even if the silicon increased, the packaging can remain the same.

      --

      Looking for any old 8-bit Heathkit/Zenith software/hardware - http://heathkit.garlanger.com

    11. Re:I don't know much about CPU internals but by A+Commentor · · Score: 1
      The transistor count is only limited by process. If you go for simpler cores (like in the article, FOR INSTANCE), you'll get more cores with the same transistor budget.
      Did you even read the article?
      While system performance in the Wintel world often lags behind theoretical microprocessor performance gains, Sellers believes that "it's a safe bet" that actual system performance for the next-generation Azul appliances will be more than twice current speeds thanks to the extra cores, faster clock speeds and new instructions. This, he says, gives Azul a good shot at being chosen over traditional non-uniform memory access (Numa) server architectures for consolidating workloads.
      If they are adding new instructions, the cores are not being simplified. Although, I didn't find where/if they stated the transistor count for the first chip with wording like: "squeezing in 812 million transistors", I'm sure they are adding to the transistor count.
      --

      Looking for any old 8-bit Heathkit/Zenith software/hardware - http://heathkit.garlanger.com

    12. Re:I don't know much about CPU internals but by timeOday · · Score: 1
      It would seem to me, that a CPU's workload is roughly limited by the number of transistors it has multiplied by it's MHz speed.
      - NO, number of transistors has nothing to do with it.
      Your statement is just as wrong as his. Yes, you can design a faster CPU given more transistors - up to a point. (If more transistors don't help, why do you think CPUs have had increasing transistor count for entire history of digital computing?) Only recently have designers failed to find new ways to use more transistors to speed up a core, hence multi core chips.
    13. Re:I don't know much about CPU internals but by Courageous · · Score: 1

      Only recently have designers failed to find new ways to use more transistors to speed up a core, hence multi core chips.

      Um.... hello? You think those extra cores didn't require transistors?

      C//

    14. Re:I don't know much about CPU internals but by TheRaven64 · · Score: 1
      There is no limit to the number of transistors you can fit on a chip. There is, however, a limit to the number of transistors you can economically fit on a chip. As you increase the transistor count, you dramatically decrease the yield, and drive up the cost. If you want your die size to be more than a 300mm radius circle then you are going to have to build a new fab (several hundred million dollars) and then expect yields of a few percent (and even then, you will have some defects on each chip).

      Beyond this limit there is a limit to what you can usefully do with these transistors. You can increase pipeline length (which also helps you drive up clock speed), but then branch miss-predictions cost a lot and you start to get into diminishing returns. You can add more pipelines (make the chip more superscalar), but then you run into the limits of instruction level parallelism, not to mention the fact that the chips spends more and more of its time (and power) running calculations that are going to be thrown away and you get into diminishing returns. You can add more cache, but once the working set is in cache you get into diminishing returns. You can add a vector unit, but that only benefits code specifically designed for it.

      Eventually, the only thing that it's really sensible to use those extra transistors for is adding more cores. The problem with that approach is that we are going to run into I/O problems in a few years. Memory speed is not advancing at anything like the rate at which we can add cores to a chip, so in the next ten years we will be able to build chips with so many cores that they are starved for data, and then we will need to re-think things a little. My money's on a transputer-like approach, programmed in a language like Erlang. IBM and friends seem to be taking the first step in this with the Cell. Sun are doing something similar with their 'vertical threading' support in Rock.

      --
      I am TheRaven on Soylent News
    15. Re:I don't know much about CPU internals but by Hotsphink · · Score: 1
      Now the only problem is in finding these "properly written programs".

      And finding properly parallelizable problems.

    16. Re:I don't know much about CPU internals but by incabulos · · Score: 1

      Idle CMOS transistor groupings dont contribute to heat though, as they dont consume power, at least in the case of the field effect transistors used in most ICs ( basically a digital on/off switch, rather than the amplifying bipolar junction type you might be more familiar with ). Each state change consumes power while the logic 1 is transitioning to a logic 0, and vice versa. The speed at which the state can change ( aka how high a frequency you can clock the thing before states are being incorrectly stored and/or detected ) is governed by the gate size ( and hence the process size of the chip itself ), which is also correlated with how much power the CMOS transistors will consume during the transition.

      So idle transistors arent quite the devil they are made out to be. They take up chip real-estate sure, but dont contribute to heat and power problems. SRAM-type caches on the other hand tend to really suck power and generate heat, due to lots of closely packed transistors all of which change state frequently as the cache is refreshed. Its why the xeons/ppro were crazily power hungry and hot compared to the rest of the cpus at the time. The core itself is a far more efficient piece of circuitry.

      That said, the trend with multicores on one chip seems to really be racing and changing at a pace thats given the big CPU makers another lease on life, there was really no way they could keep on increasing the clock speed of the core indefinately as consumers have been led to expect.

    17. Re:I don't know much about CPU internals but by Anonymous Coward · · Score: 0

      Uh, you forgot:

            o cache

      Probably one of the biggest contributors to transistor count.

  19. A new beginning by jellomizer · · Score: 2, Interesting

    Now except for the MHZ/GHZ wars the new standard will be how many cores your processor has. By 2026 My PC has 2k cores while your PC has only 1.5k cores, Thus my PC is superior, It will be just a pointless as comparing PCs using MHZ/GHZ

    --
    If something is so important that you feel the need to post it on the internet... It probably isn't that important.
    1. Re:A new beginning by Anonymous Coward · · Score: 0

      How is comparing PCs by Ghz pointless? Try underclocking your desktop CPU to 100 Mhz some time and see how much you like waiting for menus to appear.

  20. Cisco/IBM SPP by Anonymous Coward · · Score: 0

    A bazillion application specific cores isn't a new idea. Cisco's Silicon Packet Processor has 188 cores per chip to help the CRS-1 get to 92 Tbps.

    1. Re:Cisco/IBM SPP by Raul654 · · Score: 1

      You missed the point - they *are not* application specific. It's a general purpose architecture designed for running all sorts of scientific applications.

      --


      To make laws that man cannot, and will not obey, serves to bring all law into contempt.
      --E.C. Stanton
    2. Re:Cisco/IBM SPP by r_jensen11 · · Score: 1
      It's a general purpose architecture designed for running all sorts of scientific applications.

      So how is being able to run "all sorts of scientific applications" going to help me get more FPS in Duke Nukem Forever?

    3. Re:Cisco/IBM SPP by Raul654 · · Score: 1

      General purpose does not mean suitable to run every application known to man (by this defintion, there is no such thing as a general purpose architecture, because there is no architecture that can get very good performance on every conceivable application). General purpose means it's not limited to one or a small set of related problems. The signal processor attached to your car's tire (which triggers the anti-lock brakes) is an application-specific processor; a MIPs, Pentium, or Cray architecture is a general purpose architecture.

      --


      To make laws that man cannot, and will not obey, serves to bring all law into contempt.
      --E.C. Stanton
    4. Re:Cisco/IBM SPP by bufalo_1973 · · Score: 1

      Speeding up DNF development? ;-)

  21. I'm not afraid of Vega by Anonymous Coward · · Score: 0

    The chip to really watch out for will be the M. Bison

  22. Re:Ohhh boy by Anonymous Coward · · Score: 0

    Yeah, but does it run Linux?

  23. Cost of Hardware Failure by Dareth · · Score: 1

    How well do these multi-core chips fail? Do they fail silently? Do they come crashing down if even a single core on them fails?

    Are we putting too many eggs in one basket? I thought modular design was good.

    Oh well, back to setting up Linux on old dell boxes. Maybe I will get a real server one day. *grin*

    --

    I only look human.
    My mother is a halfling and my dad is an ogre, so that makes me an Ogreling
    1. Re:Cost of Hardware Failure by Anonymous Coward · · Score: 0
      Do they fail silently?


      No, they fail with a scream worthy of a little girl. You truly want to turn these things off at night because a CPU failure usually leads to you peeing the bed...
    2. Re:Cost of Hardware Failure by vadius · · Score: 1

      If this is properly designed, this shouldn't be too big of a deal. The only time I've ever had a CPU go bad is when it is improperly cooled.

      Besides, think of how many billions of transistors are being used in, say, your memory? Or any other integrated circuits we have? If a single chip on your memory dies, does the whole stick die? Of course!

      This is just taking this kind of concept to another level. Look at old memory architectures: 8088s used to have dozens of memory chips to support 256K+ of memory, and they worked fine. Today this is typically accomplished with 8 or 16 chips on a DIMM, but nothing is there to stop us from 32 (which I have a few of), 64, etc.

      CPUs are a similar game, it's just that synchronizing more processors is a bit more difficult than memory, since memory accesses are much easier to parallelize.

      Sure, it's a bit more difficult, and will decrease yield, but the idea is far from new or dangerous.

    3. Re:Cost of Hardware Failure by droopycom · · Score: 1

      Actually the more interresting question is how Intel can keep up the yield of their wafer when they have multi core chips.

      When the multi core chips get tested, if only one of the cores is bad the whole chip is thrown away.

      Now imagine that when they are doing single core cpus, they have 10% (random number) defects, thats 90% chance of having a good chip.
      Now assume (for simplicity) that there dual core waffer is basically the same except they bundle the core two by two. So they have the same number of core per waffer (but half the chips), and each individual core has 10% defect rate. Now assuming that the defects are evenly distributed on the waffer, the probably that both core of a single chip are good is only 81% instead of 90%.

      4 cores : 65%
      8 cores : 43%
      16 cores : 18%

      How are they going to keep this up ?

      Actually the problem is not new, its the same issue with just increasing the number of gates, but while adding gates add some value to the chips that might offset the reduction in yield, it doesnt seem very obvious to me the value added of a dual-core chip compared to two single core is going to offset the reduction in yield.

      Well, I guess the guys at Intel found out that it does offset it at least for dual cores...

    4. Re:Cost of Hardware Failure by Detritus · · Score: 1

      You can put N+1 cores in your N core chip, and disable one after testing.

      --
      Mea navis aericumbens anguillis abundat
  24. extreme by matt328 · · Score: 3, Funny

    If Intel's two cores are 'Extreme Edition' what should these be called? Ludicrous Edition? With a little sign by them saying 'Never Use'?

    Trying to be humorous, not seriously comparing the two chips.

    --
    Check out the cave on the east side of lake Hylia. Strange and wonderful things live in it.
    1. Re:extreme by Tmack · · Score: 1
      Plaid edition...

      tm

      --
      Support TBI Research: http://www.raisinhope.org
    2. Re:extreme by geekoid · · Score: 1

      What's wrong col. Sanders......Chicken?"

      --
      The Kruger Dunning explains most post on /. http://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect
    3. Re:extreme by reverend_rodger · · Score: 0

      Intel OMGWTFBBQ Edition

    4. Re:extreme by Daneurysm · · Score: 1

      Trying to be humorous, not seriously comparing the two chips.

      It sounds like you already knew the joke bombed.

      I bet this happens often....you must be a blast at office parties.

    5. Re:extreme by llamaxing · · Score: 1

      now THAT'S funny

    6. Re:extreme by otis+wildflower · · Score: 1

      Bug report 31337: Users report that the background on dynamic pages has 'gone to plaid'. Please advise.

    7. Re:extreme by matt328 · · Score: 1

      Some found it funny, some obviously did not. Thanks for actually spending the time to clarify that you're among those who did not. Bonus points for adding an insult.

      --
      Check out the cave on the east side of lake Hylia. Strange and wonderful things live in it.
  25. Vista? by Hoi+Polloi · · Score: 2, Funny

    Finally, a hardware platform that will support Windows Vista.

    --
    It is by the juice of the coffee bean that thoughts acquire speed, the teeth acquire stains. The stains become a warning
    1. Re:Vista? by Anonymous Coward · · Score: 0

      Vista? Are you saying a computer with 48 cores on chip can open a portal into the distant future?

    2. Re:Vista? by x2A · · Score: 1

      "a hardware platform that will support Windows Vista"

      Is more likely than Windows Vista supporting the hardware!

      --
      The revolution will not be televised... but it will have a page on Wikipedia
    3. Re:Vista? by Anonymous Coward · · Score: 0

      Not quite. I'm still waiting on terrabyte ram chips. Then again they'll probably be avaible before Vista comes out. The twist will be that it requires 1 terrabyte but only supports 512 gig of ram.

    4. Re:Vista? by DigiShaman · · Score: 1

      YES! A dedicated core in which it's resources is fully dedicated to the "windows update" thread. Finally, Vista owners everywhere can stay secure faster!

      --
      Life is not for the lazy.
  26. So what's the memory model? by Tired+and+Emotional · · Score: 4, Insightful

    So what does the memory interconnect look like on this thing? They say its not NUMA but I see no mention of what it is.

    There's no way you can feed that many processors over a single bus and if you've got symmetric access to a bunch of busses, that's one heck of a cross bar switch and I don't see that its any easier to program than NUMA. Instead of making sure data you need fast is local you have to make sure you load balance - that has to be harder much of the time.

    --
    Squirrel!
    1. Re:So what's the memory model? by Bios_Hakr · · Score: 1

      Not knowing much about processor design, I'm going to talk out my ass here for a second; bear with me.

      First off, not every core needs to be as powerful as a AMD or Intel core. There are some problems that are easier to solve using a lot of low-power cores vice one or two uber cores.

      You could also reduce the memory bandwidth to each core. You could keep a fairly powerful core, while only feeding each core a limited bandwidth.

      You could also completely change the idea behind how the proc works. Maybe they can interleave access to memory on a time-slot-based system. Or they could be building a lot of cache on die.

      --
      I'd rather you do it wrong, than for me to have to do it at all.
    2. Re:So what's the memory model? by Jeff+DeMaagd · · Score: 1

      I would think that a crossbar switch would be completely transparent to the OS or app programmer. The bus for the original Athlon (IIRC, Alpha's EV6 bus) was a crossbar switch system and I don't remember any changes to the OS, apps, drivers or anything difficult to support it.

      I've never heard of these companies or projects, so until they demonstrate something or show some credible people in the project, I'll file it in the same category as Infinium Labs and Duke Nukem. I know the current well-known chip makers are planning to scale beyond that, but they are better situated to do that than the typical dark horse company.

  27. stating the obvious by Anonymous Coward · · Score: 0

    Insert obligatory, 'One core should be enough for anybody,' comment by ignorant Slashdotter.

  28. The inquirer? by TriZz · · Score: 0

    ...next you'll be reporting that the processor will be produced by aliens who decended from clouds in the shape of Satan. The processor burned the world's fattest baby and melted his arm in half.

    --
    No matter how hot a girl is - some guy somewhere is sick of her shit.
    1. Re:The inquirer? by Anonymous Coward · · Score: 0

      Is that "Satan-shaped aliens" or "Satan-shaped clouds"?

  29. "However... by fak3r · · Score: 0, Troll

    reports show that Java still runs dog slow, even with the 768-way configuration."

    Ah, the more things change, the more they stay the same!

  30. Neat stuff. by SoupIsGood+Food · · Score: 3, Informative

    I'm not as impressed by the sillicon as I am by their product... it's a platform-agnostic application accellerator, designed to make Java apps (or any other VM app) optimized for multithreading go like stink. It does for processing power what a storage server does for disk space. Plug it into the network, and go... all it does is run a gajillion threads for the VM living on your general purpose servers. Each core probably isn't very powerful (altho they are 64bit RISC designs), but if you're in dire need of cramming as many lightweight transactions through as possible, lots and lots of little optimized processors are going to be more help than one or two big, fat general-purpose Opterons.

    It's a very neat concept, and the careful wording ("virtual machine accellerator") indicates that they aren't tied to just Java... Azul's Compute Pool could be something future Parrot-lovers can use to sneak LAMP into places where Java rules all.

    They're using some serious sillicon know-how to fuel an innovative and original product... gives me hope we aren't doomed to a wintel-only world, after all.

    1. Re:Neat stuff. by dagnabit · · Score: 1

      Their CEO is Steven Dewitt, former CEO of Cobalt Networks, so he's got the "why buy an appliance?" sales pitch down pat.

    2. Re:Neat stuff. by mtaht · · Score: 1

      Yea, well, there's a slight flaw in the ointment. Every java program I ever saw that was worth a damn linked to tons of external C libraries that did the real work.

    3. Re:Neat stuff. by afabbro · · Score: 1

      Azul is profiled in the issue of Forbes that's on newsstands now. Unfortunately, I found it rather shallow technically (which isn't surprising for Forbes).

      --
      Advice: on VPS providers
    4. Re:Neat stuff. by Ginger+Unicorn · · Score: 1

      if this C linking is due to performance issues, then it looks like this appliance eliminates the need to link to C libs to get stuff to run fast enough.

      --
      (1.21 gigawatts) / (88 miles per hour) = 30 757 874 newtons
  31. arm melting processors !!!! by pdxguy · · Score: 1

    get 'em while they are hot!!!!

  32. Memory interface by karvind · · Score: 1

    Does anyone know how you keep all the cores busy ? What kind of ulta-high-bandwidth memory architecture do they use (not clear from architecture) ? What kind of cores are these ? In-order, simple 6-10 state pipleline or more complex ones ? And more importantly, what is the power consumption ?

    1. Re:Memory interface by CliffClick · · Score: 5, Informative

      The box is a flat SMP - if a core misses in L2 it's the same cost to any piece of memory (or remote L2).

      The cores are our own design, not MIPs, not ARM, etc. Simple, short in-order pipeline, decent caches (not huge) caches.

      Power consumption is very low compared to the equivalent stack of P4 blades or other main-frame solution.

      The first-gen box (368 cores) is about 2700 watts in an 11U rack mount.
      Next-gen box isn't much bigger, nor draws very much more power (a little more of both I belive).

  33. So Which is Better? by Nom+du+Keyboard · · Score: 1

    So which is better? Vega, or IBM/Toshiba Cell Processor?

    --
    "It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
    1. Re:So Which is Better? by Zantetsuken · · Score: 1

      and people were bitching about how cell is supposed to be a pain in the ass to develope for... now they'll be thinking they've got it made in the shade!

  34. To clear up a few things by Sgt_Astro · · Score: 0

    I've noticed a few people getting the numbers wrong already... 1. It is NOT 768 cores, it is 48 cores with 768 paths through them. 2. They never mention the die size so this processor could be significantly larger in terms of silicon than others. 3. Its score in 3DMark is INSANE, not Ludicrous.

  35. When will people stop thinking like this? by Anonymous Coward · · Score: 0

    But 768 cores? What would possibly use that many cores?

    "No one would ever need more than 640K"

    "We can just use two digits and assume it's always 1990 something"

    sigh...

  36. Unhappy Math by Nom+du+Keyboard · · Score: 1

    768 Threads = 768 Memory Accesses.

    --
    "It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
  37. Obligatory by phekno · · Score: 1

    Yeah, but will it run Linux?

    1. Re:Obligatory by hentaidan · · Score: 1

      Beat me by 11 minutes.

      May you forever be condemned to run Windows Millenium Edition.

  38. OK, but... by behindthecamera · · Score: 1

    ...what's the framerate for Quake when you run this baby?

    1. Re:OK, but... by DarthStrydre · · Score: 2, Informative

      You meant that in jest... but this is the kind of parallelization required to run such things as Ray-traced applications or games in realtime (or at least jerky realtime - which is much better than several seconds watching each frame draw).

      Refer to:
      http://graphics.cs.uni-sb.de/~sidapohl/egoshooter/

      and for a screenshot with multiple reflection:
      http://graphics.cs.uni-sb.de/~sidapohl/egoshooter/ screenshots/mutlipleReflectiveSpheres.JPG

    2. Re:OK, but... by lxs · · Score: 1

      I love the term "ego shooter" instead of "first person shooter". It perfectly captures the attitude of the average FPS enthusiast.

  39. So True! by Dareth · · Score: 1

    because a CPU failure usually leads to you peeing the bed...

    Especially if you are the sys/admin who did not order a spare X core chip or worse budget for the inevitable cost of ever replacing one or more of these behemoths!

    --

    I only look human.
    My mother is a halfling and my dad is an ogre, so that makes me an Ogreling
  40. No it isn't by Cybert14 · · Score: 1

    There are fundamental limits to speed, but not to multiple cores (well, quarks in the universe maybe). The speed race was due to laziness in writing multithreaded applications.

    1. Re:No it isn't by networkBoy · · Score: 1

      screw multithreaded apps, the simple truth is that on an average personal computer there are an impressive number of processes running that have nothing to do with each other and as such could benifit from multi-cores, even if they were all single threaded.
      -nB

      --
      whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
  41. Vega by nschubach · · Score: 1

    Why must I think of Tom's Diner every time I see Vega?? Damn you Suzanne !!

    --
    Every time I start to have faith in humanity, I ruin it by driving to work between 7 and 8 am.
    1. Re:Vega by LWATCDR · · Score: 2, Funny

      Funny I think leaking head gaskets and rust when ever I see Vega. I must be getting old.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    2. Re:Vega by BigCheese · · Score: 1

      I thought the same thing. We're both getting old.

      For those of you who don't know (or had the good fortune to forget) here it is:
      http://www.stationwagon.com/gallery/1974_Chevy_Veg a.html

      --
      The obscure we see eventually. The completely obvious, it seems, takes longer. - Edward R. Murrow
    3. Re:Vega by LWATCDR · · Score: 1

      I had a friend in highschool that put a small block chevy into a Vega. I think it lasted about 3 weeks before the windshield broke from the twisting.
      It was a scary car.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
  42. Wow!... by Illbay · · Score: 1, Funny
    ...imagine a Beowulf cluster of...

    Ah, never mind.

    --
    Any technology distinguishable from magic is insufficiently advanced.
    1. Re:Wow!... by Anonymous Coward · · Score: 0

      I can't believe it took this many comments for that joke to pop up.

      But seriously, imagine a Beowulf cluster of those. *ponders*

      Ack no don't! I just hurt my head!!

  43. Points-of-Failure by bradgoodman · · Score: 1

    I hate this idea. Clustering - okay. But SMP? No way. The more CPUs - the more points-of-failure - and for SMP - that bring everything down.

    Can you imagine the MTBF on a puppy like this?!

    Now if were talking about two-by-n redundancy - (or something of the equiv.) - that's a whole other story - but now were not just talking CPUs - caches too.

    But since that technology itself is more complicated than just copy-and-pasting a bunch of cores - I don't think that's the kind of think that can be "slapped-in" at the last minute - nevermind the software implications...

    1. Re:Points-of-Failure by shmlco · · Score: 1

      An Intel Core Duo has 151 million transistors. This has a little over 4.75x that. So the MTBF wouldn't be too bad, I would think.

      --
      Any sect, cult, or religion will legislate its creed into law if it acquires the political power to do so.
  44. Doom 3 by Mikya · · Score: 1

    The articles continues:
    "with the added processing power of the Vega, Doom 3 is expected to run in excess if 20 FPS."

  45. I'm still figuring this out... by tlynch001 · · Score: 0

    If I say "Finally, Microsoft will run quickly" I score a 5, but if I say "Finally, Java will run quickly" I am a troll? I need to know because since I seem to be stuck in some sort of bad Karma spiral and need to get out.

  46. so... this is a video card processor for... by Anonymous Coward · · Score: 0

    Duke Nukem Forever?

    (ducking... but c'mon... it had to be said)

  47. Re:imagine... by pxuongl · · Score: 1

    actually only a millisecond difference would've been enough to give the 6:42, 6:43 difference. bastard beat me to the punch

  48. Graphics work by phorm · · Score: 1

    My first thoughts were graphics applications, particularly 3d and rendering apps. This could be especially useful when it comes to things like raytracing for an animation, where you could dedicate parts of the work to different cores. Kinda like having a server farm all in one machine...

  49. Ghostbusters by DragonHawk · · Score: 3, Funny

    Right now, it seems like they're a company, but they don't have a product yet. I guess that means...

    There is no Vega only Azul.

    (Thank you, thank you, I'll be here all week.)

    --

    dragonhawk@iname.microsoft.com
    I do not like Microsoft. Remove them from my email address.
    1. Re:Ghostbusters by Meph_the_Balrog · · Score: 1
      Right now, it seems like they're a company, but they don't have a product yet. I guess that means...


      I don't think they got that :P (I'd mod you funny, but no mod points for me..)
    2. Re:Ghostbusters by kernelfoobar · · Score: 1

      Thank god! I thought I was the only one that thought of Ghostbusters when I saw the name!

      Dr. Peter Venkman: [discussing the creature Dana saw in her fridge] Zuul was the minion of Gozer.
      Dana Barrett: What's Gozer?
      Dr. Peter Venkman: Gozer was very big in Sumeria
      Dana Barrett: Well, what's he doing in my ice box?
      Dr. Peter Venkman: I'm working on that.

      --
      Here we go again!
  50. Weather Simulation by nurb432 · · Score: 1

    DNA work, or other easily parallized processes.. For high end work.. Its not so you can run word or something.

    --
    ---- Booth was a patriot ----
  51. Heat, and power by phorm · · Score: 1

    What I would wonder about is how this would work for heat. From a software perspective dual-core chips appear as multiple processors, as it seems to essentially be having multiple processing units in a single die (correct me if I'm a bit off here, searching for the simplest answer).

    The issue I see with this is:

    Multiple processors generate more heat, and consume more power. Would it not be the same for multiple cores, thus making such a machine a power-chugging space-heater? Are special cooling devices required when you start hitting so many cores?

    1. Re:Heat, and power by eh2o · · Score: 1

      The practical limit for dissipation by air-cooling is around 100-150 Watts. Heat is a function of the number of transistors which are powered on *and* the clock rate. If you want to make a single core chip run faster there really only two options -- 1) make more fancy circuitry (branch prediction, new instructions, etc) on the chip but only switch it on when necessary and when heat constraints allow you to and/or 2) ramp up the clock rate. Intel put their money on #2 and they hit the heat wall, hard. But going with #1 gives you diminishing returns after some point because the fancy logic becomes too specialized to have broad applicability.

      So the solution is to hit the ball into the other court -- that is, make a chip with multiple cores but a modest overall clock rate so it still fits the heat profile, and tell the OS and application designers to rewrite their code for the parallel architecture.

  52. My users.... by Rhys · · Score: 1

    Keep 1280 cores at a 90-100% load, 24 hours a day, 365 days a year. They show no sign of stopping.

    It is called supercomputing.

    --
    Slashdot Patriotism: We Support our Dupes!
  53. Slashdot gettin' slow? by otis+wildflower · · Score: 0, Flamebait

    Is it just me or has /. really been behind the pace of news lately? I'm seeing links come up like hours or even days after I see 'em on theinquirer, arstechnica, macnn, etc...

  54. The answer is 42! by gwayne · · Score: 1

    I wonder how oracle will price running on this processor?

    Obviously, the answer is 42!

  55. It Runs Java not X86 Code!!! by TheNarrator · · Score: 3, Informative

    Have any onf you people visted Azul's website? This is not an Intel compatible machine.
    It is going to only run a Java Virtual Machine so anything written in Java will run on it.
    Windows will not run on it. I took some operating system courses in college and the intel
    architecture is a huge mess of hideousness of backwards compatibility that luckily only operating system implementers have to deal with. By only running Java these guys get to sidestep the whole mess and focus
    on massively optimizing the hardware architecture for running java code.

    http://www.azulsystems.com/products/nap.html

  56. traditional programming languages obsolete? by Khashishi · · Score: 3, Insightful
    Languages like c are inherently serial. All statements must be placed in sequential order even when there are no serial dependencies between the two. E.g.
    x = zeta(y)
    w = gamma(z)
    print(x+w)
    The code explicitly states that x should be calculated before w although they could certainly be calculated concurrently. Of course a smart compiler could figure out the dependencies, but the programming language shouldn't force the programmer to specify an order when none exist.

    I predict that non-procedural languages will dominate the future of programming. Some currently used languages seem already well-suited for taking advantage of multiple cores, like HDL languages, functional languages, Labview-style languages.

    1. Re:traditional programming languages obsolete? by Anonymous Coward · · Score: 0

      Just like a lazy-evaluation language defers computation as long as it can, a compiler doesn't need to spit code immediately for every line of code (or statements/expressions) it reads. It can generate a dependency graph for a whole block of code and then rearange instructions so you can keep your CPU's pipelines as busy as possible. The trouble is: some algorithms are very serial by nature, so even with a fancy compiler and a superscalar CPU, or a VLIW/EPIC one, can't do a thing about it. This is quite low-level, though.

      With a bit more work, and probably a more higher-level language, the compiler could take a good look on the computation done inside loops, check if there's any serial dependencies between steps, and if not, use whatever SIMD mechanisms your CPU provides, or if it's a really lengthy piece of code, it could try to run them in different threads and throw each of them at different cores.

      Trouble is, it's not that simple, and for very low level languages it can be pretty hard to be sure no weird serial dependencies exist. Such optimizations also may take quite some time to compute, and may make the compiler really resource hungry. It would also require a very smart, thus very complex compiler. Some of this stuff is already being done on today's compilers, but more aggressive optimizations can be pretty hard to implement properly.

    2. Re:traditional programming languages obsolete? by null-sRc · · Score: 1

      // non blocking thread init (all non blocking to avoid ordering issues
      x.zetathread(y).start();
      w.gamethread(z).start(); // blocking waits for each to avoid ordering issues
      x.waitforthread();
      w.watiforthread(); // do whatever u want to do when they're all done
      print(x + w);

      It's up to the programmer to optimize the code... the compiler isn't psychic! if it were... we wouldn't need programmers anymore.

      you can use threads and signals to do things in parallel and start them in any order you wish if you wanted to.

      --
      -judging another only defines yourself
    3. Re:traditional programming languages obsolete? by Anonymous Coward · · Score: 0

      I predict that non-procedural languages will dominate the future of programming.

      Gee, when did i read that before? Somewhat around 1985 maybe? BWAHAHAHAHA!

  57. Still not quite right by grahamsz · · Score: 1

    The grandparents is completely right, the amount of work a proc can do is, in absolute, the number of transistor it has times the number of times per second you can get them to do something.

    In a way that's true, but most cpu's are nowhere near the bound of what you can get from the transistor count. The general purpose nature of them makes that true.

    If you've ever used a hardware definition language like verilog, it becomes apparent that you can design silicon that can do FAR more per clock cycle than most CPUs. As a novice I was able to get a 100Mhz FPGA doing computations faster than I would have been able to get from an 800Mhz P3.

    Now the challenge of course is getting a general purpose CPU to have that sort of efficiency which is not trivial. Azul seem to have the right general idea, they'll make a chip optimized to particular tasks and try to get a little closer to the ideal.

  58. Much Ado About Not a Damn Thing! by Anonymous Coward · · Score: 1
    Rewind backwards about 6 years and listen to the mindless drone of support for Java processors. They did not last long because Java compiled into x86 machine language actually ran faster than Java executed directly by a Java processor.

    Now, we have this multi-core, multi-threaded Java processor. It might be faster than a single/dual-core AMD processor.

    However, I would bet good money (on AMD stock) that Azul processor is, in fact, slower than an AMD processor with the same number of cores and threads. The whole purpose of Azul is to sucker some venture capitalists into paying fat salaries to the managers and slaves working at Azul.

    Also, the lawsuit by Azul against Sun is a mere publicity stunt. The supposed patents for the supposed technology in the Java processors of yesteryear are not worth the paper on which the patents are written. Don't believe me. Fire up Live Search (by America's modern-day Bell Labs) and try to find any of yesteryear's Java processor still being sold.

    1. Re:Much Ado About Not a Damn Thing! by Anonymous Coward · · Score: 1, Informative

      > They did not last long because Java compiled into x86 machine language actually ran faster than Java executed directly by a Java processor.

      this is NOT (repeat repeat repeat) a *JAVA* chip.

      its a general purpose RISC chip that has hardware support for many common java functions, such as GC and locking.

  59. Size by necro81 · · Score: 2, Interesting

    For comparison purposes, the Cell processor in the PS3 has something like 100 million transistors, comes from a 90 nm process, and has a die size of about 1 cm square. The Cell has a modestly-sized cache, which means that its transistors are mostly given over to functional blocks. This is in contrast to something like a P4 Extreme edition, which has a higher transistor density because more than half its die is cache memory.

    TFA does not mention anything about this new processor's die size. But, if we scale up the Cell processor's transistor density, the Vega processor, with 812 million transistors, would result in a die size of about 800 mm^2, which is more than one square inch. In the processor industry, that kind of die size is just plain ridiculous. I wonder what the yields are?

    1. Re:Size by Anonymous Coward · · Score: 0

      where i work, our next-gen asic is going to tapeout ~500mm^2 in 80nm. the predicted yeilds on early wafers are not particularly high... of course, i find the projected power numbers to be more interesting: ~250W (@1.2V) that's just our asic. nice and toasty...

    2. Re:Size by JollyFinn · · Score: 1
      TFA does not mention anything about this new processor's die size. But, if we scale up the Cell processor's transistor density, the Vega processor, with 812 million transistors, would result in a die size of about 800 mm^2, which is more than one square inch. In the processor industry, that kind of die size is just plain ridiculous. I wonder what the yields are?

      If those transistors are most of cache then the yields are pretty good. If they are logic then there needs to be more clarification. Firstly they probably have some sort of redundancy there. Probably having something like few spares in case there is problem in one of the cores. As for yields of cache divide area dedicated to cache by 4. What they are doing is quite different from cell so they might have larger portion of area in cache, which bring transistor densities up, and your area estimate way off. They might even have less logic transistors than in SINGLE athlon core on *ALL* their cores combined, and rest as cache.

      There is reason why they would put lots of cache instead of logic transistors. A simple Arm is 30k transistors,ARM9 is 110k transistors & jazelle unit is 12k transistors. Thats amount of transistors take to execute most of javabytecode, without floating point support directly on hardware designed to run risc ISA. It could be even simpler than that, but adding fpu support probably costs some of the simplicity but I doubt seriosly if each core would have large execution resources trying to find ILP. Now what about scaling 48 cores or so. At that point most of need is memory bandwith and cache bandwith on shared cache levels. It should be obvious that increasing L1 cache hit rate is important to avoid stalling on shared memory bandwith. And Increasing L2 cache hit rate would be important to avoid wasting memory bandwith. The cell vs azul cores is partially dependant on inside core parallerism and very high clockspeed target of cell to add more logic transistors on cell design. And of course high disparity on number of cores. If they chooce not do bytecode directly but have jit then still vliw or simple risc would still be a good choice.

      Montecito has 1.65Billion transistors in 0.9u process and has die size of 600mm. Thats dual core itanium. With its transistor density the Vega2 would be 300mm with most of die area spend on cache, and yield wouldn't be bad at all.

      --
      Emacs is good operating system, but it has one flaw: Its text editor could be better.
  60. yeah but by hatori · · Score: 1

    developers are already having a hard time coping with Cell's 7 or 8 cores for the PS3. IBM had to release a proprietary compiler (octopiler) to calm down the coding cries. Wouldn't 48 cores be a developer's nightmare?

    1. Re:yeah but by Anonymous Coward · · Score: 0

      Yes, 23485723 cores would be horrible for a video game console.

      However there are other uses for microchips. Some of which are very parallelizable.

      Imagine distributing a frame of Quake to a hundred PCs on the internet and waiting for each to spend 1ms on calculations and to return the result for their portion of the workload. It would be a horrible frame rate compared to single fast PC. Now imagine you've got some protien that needs folding.

      Choose your best weapon.

  61. These are Java Appliances by Anonymous Coward · · Score: 0

    These systems are Java appliances designed to run JVMs. It basically, attaches to your regular intel/amd/whatever box and looks like just another mount point. It has software that proxies any java server calls to the azul appliance and runs your jvm. Its basically throwing hardware at the 4 way 4GB problem that JVMs typically have.

  62. Chevy Vega? by Anonymous Coward · · Score: 0

    What kind of company is stupid enough to name its product after the worst car ever produced by Chevy? Man will it be funny if they walk into Baby Boom Corporate America trying to pitch hardware with the "48 core high-tech Vega engine inside", complete with a Hindi accent and a 1970s wardrobe.

  63. Duke nukem forever by Anonymous Coward · · Score: 0

    Your just jeolous that Duke nukem forever, that I got on a 300GB 3D optical disk, runs at 300fps on this machine.

  64. To paraphrase... by Biomechanical · · Score: 1

    "The desktop is the computer."

    --
    His name is Robert Paulsen...
  65. 3D rendering and multi-cores by Anonymous Coward · · Score: 0

    I'm currently evaluating the feasibility of using Blender included in a Linux distribution named Tomahawk Desktop.

    Using Blender has a significant cost advantage. Our bottleneck is the rendering time. We expect these chips may be available soon at a price SOHO can afford.

  66. chicken and the egg problem by GunFodder · · Score: 1

    Many others have posted their proposed uses for massively parallel processing units. Personally I see this as a great server platform for running software that is already optimized for multiprocessor architectures. But you also bring up a good point in that most software is not optimized for multithread CPUs, especially on the desktop.

    A few years ago only expensive servers had multiple processors. Now every major CPU maker produces a multicore chip as their flagship product. Many of these chips have hit consumer level prices. I think it is only a matter of time before we start seeing a proliferation of multithreaded software.

  67. Re:Azul longevity by HaydnH · · Score: 2, Informative

    How can pointing out there's a current Sun Lawsuit with these guys be a troll? TFA even mentions the Sun lawyers:

    "The scalability is showing is attracting big-name early adopters, including Credit Suisse - and even enough to have Sun Microsystems lawyers hammering at the door, alleging intellectual property infringements."

    Basically Sun are saying that Azul are infringing on Sun's patents and have illegaly obtained Suns trade secrets. Sun have tried to take part ownership of the Azul and charge ongoing license fee's. Azul have given Sun a chance to look at their documents etc to prove that they haven't infringed on the patents, but Sun haven't taken them up on the offer - I believe Azul are trying to sue Sun also as they believe they're just trying to distract their companies resources.

    Personally I wouldn't like to pin my hopes on a chip that has so much politics going on behind the scenes - I'd rather wait until all of this is sorted.

    --
    Time is an illusion. Lunchtime doubly so. - Douglas Adams
  68. 640 processors by cool-code-guy · · Score: 1

    Who needs 768 processors? 640 ought to be enough for anybody!

  69. Re:Azul longevity by Bitmanhome · · Score: 1
    Personally I wouldn't like to pin my hopes on a chip that has so much politics going on behind the scenes - I'd rather wait until all of this is sorted.
    Yes, that's exactly why Sun is suing. By the time it's "sorted", Azul will be gone and you'll buy Sun's chip.
    --
    Not that this wasn't entirely predictable.
  70. how to use those cores by pensivemusic · · Score: 1

    If the O/S allocates a core to an icon, that would work.