Future of SGI? - SGI

This is a discussion on Future of SGI? - SGI ; Le Sun, 29 Jan 2006 17:52:08 +0100, kolaric a écrit*: > > Yes. > It would be interesting to hear from SGI management on what was the > rationale behind these decisions. Stop developing OS entirely in-house. Stop developing complete ...

+ Reply to Thread
Page 2 of 3 FirstFirst 1 2 3 LastLast
Results 21 to 40 of 46

Thread: Future of SGI?

  1. Re: Future of SGI?

    Le Sun, 29 Jan 2006 17:52:08 +0100, kolaric a écrit*:

    >
    > Yes.
    > It would be interesting to hear from SGI management on what was the
    > rationale behind these decisions.


    Stop developing OS entirely in-house. Stop developing complete CPU
    architecture in-house. They succeeded in both. However they are now
    dangerously dependent upon Intel.

    --
    Quis, quid, ubi, quibus auxiliis, cur, quomodo, quando


  2. Re: Future of SGI?

    In article <4447vgF9dc7U1@individual.net>, "S.C.Sprong"
    wrote:

    : A company that would concentrate itself on high reliable _systems_ with much
    : of the hardware and software integrated and hidden could make it, when|if
    : reliable systems start to become popular.

    The problem is that high-end reliable is a hard sell these days. Most people
    just don't expect anything they buy to break, so see no value in buying
    something that's designed to break less.

    There was recently an article on Slashdot asking about implementing a
    departmental file server. Hundreds of people chimed in about cobbling together
    bottom-line PC with a few consumer-grade IDE-Raid cards and using that. I'm sure
    if you mentioned "fiberchannel" they'd think it's a laxative powder that you mix
    with orange juice. You can't sell quality to these people.

    This race to the bottom has driven down the price of low-end systems to the
    point where in a lot of cases it's more cost effective to buy multiple low-end
    systems and cluster them in an n+1 or n+2 setup than it is to buy one large,
    ultra-reliable supersystem. The market for quality is shrinking every day.

    And don't forget that SGI already tried selling to the PC workstation market
    with a custom system, the 320 and 540, and failed. They then tried selling to
    the PC workstation market with non-custom systems like the 550 et al and again
    failed.


    Cheers - Tony 'Nicoya' Mantler

    --
    Tony 'Nicoya' Mantler -- Master of Code-fu -- nicoya@ubb.ca
    -- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

  3. Re: Future of SGI?

    In article <443dlrF5brqU1@individual.net>,
    Michael Laajanen wrote:

    : Now we have to hope for the Fujitsu guys to create a great Sparc so we
    : still have a few high end CPUs in this world.

    IBM's Power is still alive and well.


    Cheers - Tony 'Nicoya' Mantler

    --
    Tony 'Nicoya' Mantler -- Master of Code-fu -- nicoya@ubb.ca
    -- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

  4. Re: Future of SGI?


    >> You're talking about a full-blown workstation here.
    >> Such machines are usually sold to academia, government
    >> and industry.

    >
    > Well, what else is there? Ah, right, the "hobbyist"... So, are you saying
    > sgi to survive must become another Dell?!?!?


    Not the "hobbyist", but the "advanced home user".
    And no, SGI shouldn't become another Dell -
    they should limit themselves to CG systems
    (when talking about desktops/PCs).



    >> For SGI to get there (that is, be able to start making
    >> and selling such workstations again), IMHO they first have
    >> to get back to basics and be able to make great desktop
    >> machines for a wider audience - gamers,
    >> and professionals working in design, multimedia, visualization.

    >
    > Just what I thought: gamers... :-D


    Well, why not? For starters, we'd get publicity.
    Probably some profit as well.


    >> Competitor in this space would be for example Dell XPS 600.
    >> However, the idea is to make something that is better
    >> than this, for a slightly (10-20%) higher price. The objective
    >> would to be on the bleeding edge of computer graphics.
    >> I'm sure that the targeted customers (gamers, creatives) will
    >> take notice. Especially gamers - there is a lot of customers
    >> out there who will pay premium for such a machine.

    >
    > sgi is not at all in that market, sgi makes most of its revenue from
    > support, not selling hardware... There is close to no margin in "gamers"
    > hardware...


    I know SGI is not in that market now. I merely say
    it should. Wishful thinking, perhaps.



  5. Re: Future of SGI?

    HI,

    Tony 'Nicoya' Mantler wrote:
    > In article <443dlrF5brqU1@individual.net>,
    > Michael Laajanen wrote:
    >
    > : Now we have to hope for the Fujitsu guys to create a great Sparc so we
    > : still have a few high end CPUs in this world.
    >
    > IBM's Power is still alive and well.
    >

    Yes, but sadly they lost Apple and they have sold the embedded PPC team
    to AMCC which makes me worried about Power abit.


    /michael

  6. Re: Future of SGI?

    Le Sun, 29 Jan 2006 23:21:16 +0000, Tony 'Nicoya' Mantler a écrit*:

    >
    > This race to the bottom has driven down the price of low-end systems to the
    > point where in a lot of cases it's more cost effective to buy multiple low-end
    > systems and cluster them in an n+1 or n+2 setup than it is to buy one large,
    > ultra-reliable supersystem. The market for quality is shrinking every day.


    And you actually have to do with the fact that even low-end hardware are
    increasingly reliable.

    --
    Si non confectus non reficiat.


  7. Re: Future of SGI?

    On Sun, 29 Jan 2006 03:04:55 +0100,
    kolaric , in
    wrote:

    >+ That whitepaper that I mentioned in my OT certainly
    >+ doesn't describe such a strategy. It's a sad, apologetic document,
    >+ leading to nowhere.


    Then SGI is done for. Time to move on.

    James
    --
    Consulting Minister for Consultants, DNRC
    I can please only one person per day. Today is not your day. Tomorrow
    isn't looking good, either.
    I am BOFH. Resistance is futile. Your network will be assimilated.

  8. Re: Future of SGI?

    HI,

    Emmanuel Florac wrote:
    > Le Sun, 29 Jan 2006 23:21:16 +0000, Tony 'Nicoya' Mantler a écrit :
    >
    >
    >>This race to the bottom has driven down the price of low-end systems to the
    >>point where in a lot of cases it's more cost effective to buy multiple low-end
    >>systems and cluster them in an n+1 or n+2 setup than it is to buy one large,
    >>ultra-reliable supersystem. The market for quality is shrinking every day.

    >
    >
    > And you actually have to do with the fact that even low-end hardware are
    > increasingly reliable.
    >

    You should know what stuff that can you into a military system today
    which used to demand very high quality assurance, anyone remember the
    54xx circuits for instance, backtrack where the chips where manufactured
    and so

    I remember when the talk to replace our high end systems with cheap COTS
    systems like Sun, SGI and IBM started, Unix X11, gosh

    /michael

  9. Re: Future of SGI?

    Tony 'Nicoya' Mantler wrote:
    >"S.C.Sprong" wrote:
    >
    >: A company that would concentrate itself on high reliable _systems_ with
    >: much of the hardware and software integrated and hidden could make it,
    >: when|if reliable systems start to become popular.
    >
    >The problem is that high-end reliable is a hard sell these days. Most
    >people just don't expect anything they buy to break, so see no value in
    >buying something that's designed to break less.


    It is tragic that over the years the general public was _taught_ that
    computer systems are inherently unreliable and break down all the time.
    Therefore the cost of lost productivity, or even lives, has become quite
    acceptable, and there's hardly any protest.

    Most companies with a focus on high-end reliability are gone: Tandem/
    Himalaya, and VAX/Alpha based VMS systems. Indeed, nowadays, most people,
    even ICT folk, think that reliable computing systems never existed.

    And until the private and public sectors are _forced_ by law to take full
    responsibility for their computing mishaps the demand for reliability will
    stay low.

    >This race to the bottom has driven down the price of low-end systems to
    >the point where in a lot of cases it's more cost effective to buy multiple
    >low-end systems and cluster them in an n+1 or n+2 setup than it is to buy
    >one large, ultra-reliable supersystem. The market for quality is shrinking
    >every day.


    As I see it, the fundamental problem to that approach is that the effects
    of their accumulated unreliability are still poorly understood, and, as far
    as I know, there's hardly any new research done in that area.

    This means that they can work, or they don't. And if they don't, it is very
    difficult to discover why.

    >And don't forget that SGI already tried selling to the PC workstation
    >market with [custom systems] and failed.


    I won't. I like SGI's overall engineering prowess, but I'm not a fanboi.

    scs

  10. Re: Future of SGI?

    Emmanuel Florac wrote:
    >Le Sun, 29 Jan 2006 16:27:57 +0000, S.C.Sprong a écritÂ*:
    >>
    >> They bet on two dead horses at the same time. With choosing Linux
    >> they chose to throw away most if not all of their multi-CPU support.

    >
    >This is grossly exwagerated. There are 512-cpus linux nodes nowadays.


    Are you referring to loosely coupled systems such as Beowulf systems?
    Those evade the multi-cpu problem and concentrate on message passing.
    In some problem areas this approach works, but not in all, such as in
    large transaction database systems.

    At the moment, the continuous increase in CPU performance is stagnating.
    Perhaps it lasts only a few years, perhaps it is permanent, who knows?
    A solution is to design systems with a large number of cpu's (>> 8).
    With choosing Linux, SGI has to do a lot of that work all over again.

    scs

  11. Re: Future of SGI?

    Le Mon, 30 Jan 2006 16:43:59 +0100, Michael Laajanen a écrit*:

    >
    > You should know what stuff that can you into a military system today
    > which used to demand very high quality assurance, anyone remember the
    > 54xx circuits for instance, backtrack where the chips where manufactured
    > and so
    >


    You know, very cheap and low-end EPIA VIA systems were approved for use in
    aircrafts...

    --
    Writing about music is like dancing about architecture.
    Frank Zappa


  12. Re: Future of SGI?

    Le Mon, 30 Jan 2006 16:12:09 +0000, S.C.Sprong a écrit*:

    >
    > Are you referring to loosely coupled systems such as Beowulf systems?


    No, I'm talking about NUMA single kernel image systems.

    > At the moment, the continuous increase in CPU performance is stagnating.
    > Perhaps it lasts only a few years, perhaps it is permanent, who knows?
    > A solution is to design systems with a large number of cpu's (>> 8).
    > With choosing Linux, SGI has to do a lot of that work all over again.


    But they already did it ! They shipped 128 CPUs altix for years.

    --
    Writing about music is like dancing about architecture.
    Frank Zappa


  13. Re: Future of SGI?

    In article <446s6oFl0fpU1@individual.net>, "S.C.Sprong"
    wrote:

    : >This is grossly exwagerated. There are 512-cpus linux nodes nowadays.
    :
    : Are you referring to loosely coupled systems such as Beowulf systems?
    : Those evade the multi-cpu problem and concentrate on message passing.
    : In some problem areas this approach works, but not in all, such as in
    : large transaction database systems.

    The SGI Columbia installation is composed of a cluster of 20 512-CPU Altix
    systems for a total of 10240 CPUs.




    Cheers - Tony 'Nicoya' Mantler

    --
    Tony 'Nicoya' Mantler -- Master of Code-fu -- nicoya@ubb.ca
    -- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

  14. Re: Future of SGI?

    In article <1138747412.260552.305020@g49g2000cwa.googlegroups. com>,
    mapesdhs@yahoo.com wrote:

    : Just a shame that rather than being
    : fixed/improved in IRIX, the better ways of doing these things will end up in
    : SGI Linux.

    It depends on how you look at it. If IRIX had better prospects for the future,
    then yes it would be seen as a shame. As it is, with commercial unixes
    everywhere being marginalized and eliminated, we can at least be thankful that
    the technologies in IRIX won't be lost with the OS, even if SGI itself
    disappears.


    Cheers - Tony 'Nicoya' Mantler

    --
    Tony 'Nicoya' Mantler -- Master of Code-fu -- nicoya@ubb.ca
    -- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

  15. Re: Future of SGI?

    In article <1138746987.077118.6470@z14g2000cwz.googlegroups.co m>,
    mapesdhs@yahoo.com wrote:

    : Cool, a Tony-post, the ones with the more sensible comments!

    Flattery will get you everywhere.


    : I've heard company people say they'll just replace systems when they
    : go wrong because it's cheaper & easier than bothering with warranty cover! Sheesh...

    Some companies don't even do that. Heck, when a system dies at Google, they
    don't even bother to turn it off.


    : Meanwhile, Linux still doesn't feel particularly professional to me, and lacks the
    : integrated, ease-of-use feel that I get in IRIX. Trying to set something up in Linux
    : feels too much like I'm expected to get my hands dirty downloading all sorts of extra
    : packages and dealing with dependencies; why should I have to do that?

    I've been fairly happy with Debian. They've got pretty much every linux app in
    existence in their repository, with full automatic dependancy fetching etc. Just
    one command to install it and everything it needs. Keeping things up to date
    with the latest versions is just as easy and automatic.

    It's certainly not the same experience as IRIX, but it's certainly very workable.


    : I wish I knew a lot more about programming, so many things I'd write if
    : I could, codecs for O2, VPro accelerators for GIMP, etc. Oh well.

    No time like the present to learn.


    Cheers - Tony 'Nicoya' Mantler

    --
    Tony 'Nicoya' Mantler -- Master of Code-fu -- nicoya@ubb.ca
    -- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

  16. Re: Future of SGI?


    Tony 'Nicoya' Mantler wrote:
    > The SGI Columbia installation is composed of a cluster of 20 512-CPU Altix
    > systems for a total of 10240 CPUs.


    I was told that the original MIPS/Origin4000 plan was going to offer
    37500 CPUs
    single-image. I suppose it is a valid point that if MIPS/IRIX/Origin
    development
    had continued, this level of scalability would be out and in use by
    now. The move
    to Altix/etc. has introduced a delay, but they've been able to do the
    switchover
    pretty quickly all things considered, and it has offered SGi a chance
    to redo
    some of things which weren't done so well in IRIX. Many things in IRIX
    were
    first attempts, especially the video stuff. Just a shame that rather
    than being
    fixed/improved in IRIX, the better ways of doing these things will end
    up in
    SGI Linux.

    Still, others continue to write new code. A friend of mine who did the
    O2 'nomagic'
    capture tool says he'll get back into working on a full O2 video
    editing suite. Capture
    hours and hours of video with no frame drops? No problem. Uncompressed
    video
    included. Nobody *needs* a 3GHz CPU to do video, which is not what MS
    would
    have you believe...

    Ian.


  17. Re: Future of SGI?


    Cool, a Tony-post, the ones with the more sensible comments!

    Tony 'Nicoya' Mantler wrote:
    > bottom-line PC with a few consumer-grade IDE-Raid cards and using that. I'm sure
    > if you mentioned "fiberchannel" they'd think it's a laxative powder that you mix
    > with orange juice. You can't sell quality to these people.


    Yup, just what someone at SGI told me ages ago, the market right now is
    all
    about lowest possible cost. Managers and accountant/IT people don't
    care
    about quality & reliability. As such, there's no market for a quality,
    integrated,
    innovative workstation of the kind SGI became so famous for making.

    Of course SGI could design and make a new system which would be great,
    fast, blah blah, but nobody would buy it. Not right now. The pressures
    from
    all angles are all about lowest possible cost, even when the hardware
    being
    bought is rubbish. As an admin years ago, I once had to sit through a
    procurement
    of a load of 'high end' (ha) PCs, nearly a third of which were faulty
    on delivery, yet
    this was regarded as normal! And don't even get me started on how such
    systems
    come installed, eg. big IDEs setup as zillions of 2GB FAT32
    partitions... 8\

    I ended up spending almost all of my admin time dealing with PC virus
    and security
    crap. Meanwhile, the SGIs just ran themselves. More PC stuff coming in,
    more PC
    hw/sw nonsense, more winblows crashes, more system freezes 'cos someone
    ran
    more than one thing at once, virus attacks, etc. and a vast range of
    other ridiculous
    Wnslows issues that wasted so much time. Got sick of it, not why I
    became an
    admin, so time to leave.

    There was the occasional fun moment, like when newly bought P4 GF4 PCs
    were
    initially massively slower than VW320s because of the large texture
    datasets (180MB+)
    and the use of huge composite textures. IVC is cool...


    > This race to the bottom has driven down the price of low-end systems to the
    > point where in a lot of cases it's more cost effective to buy multiple low-end
    > systems and cluster them in an n+1 or n+2 setup than it is to buy one large,
    > ultra-reliable supersystem. The market for quality is shrinking every day.


    Exactly. I've heard company people say they'll just replace systems
    when they
    go wrong because it's cheaper & easier than bothering with warranty
    cover! Sheesh...


    > And don't forget that SGI already tried selling to the PC workstation market
    > with a custom system, the 320 and 540, and failed. They then tried selling to
    > the PC workstation market with non-custom systems like the 550 et al and again
    > failed.


    Bad marketing ruined that one IMO, though I do think the later 550/etc.
    were a bad
    idea. However, the 320/540 arch was good, ideal for many markets, but
    was
    incorrectly marketed in so many ways, and died as a result. SGI didn't
    even
    properly get across that just the basic parts used were of a superior
    quality,
    which contributed to a good chunk of the extra cost compared to a
    generic PC
    of the same on-paper spec.


    Reading through this thread, a lot of people seem to be living in a
    dream world.
    Somehow they think SGI can magic up a new design and enter into low
    margin
    markets just like that. Not going to happen. Design talent doesn't hang
    around
    when rivals can offer vastly higher salaries, even when the company is
    doing ok.
    Intel pinched the CPU people, NVIDIA grabbed the gfx people, though
    others
    went elsewhere, then ATI grabbed some of the same people from NVIDIA.

    And 'setting objectives' has no relation to whether or not a company
    can come up
    with a new, better design. Like Tony said, many companies are hitting
    physical
    limit problems, and/or bottlenecks that are elsewhere in the system (I
    doubt there's
    any PC system available today that could push a 'top-end' NVIDIA/ATI
    card
    even as much as half of the gfx card's peak possible speed for a
    real-world task).

    What makes me laugh is seeing the MHz numbers go sky high, disk
    capacities
    soar, etc., yet the main OS of choice by most companies remains so
    ruddy awful
    (Winblows I mean) and the quality of much of the hardware stays equally
    dreadful.

    Meanwhile, Linux still doesn't feel particularly professional to me,
    and lacks the
    integrated, ease-of-use feel that I get in IRIX. Trying to set
    something up in Linux
    feels too much like I'm expected to get my hands dirty downloading all
    sorts of extra
    packages and dealing with dependencies; why should I have to do that?
    Nobody
    seems to bother bundling all the essentials for a particular type of
    task together, eg.
    video processing. Ironically, those who write Windows apps seem to
    bundle things
    quite often, eg. the vast range of video tools available from
    www.videohelp.com.

    It's a pity SGI never ported the digital media tools, or just made the
    source
    code completely free, bundling it with the 320/540 and now all the
    Altix/Prism
    stuff. Ditto for the IRIX desktop stuff. At the very least, it would
    ease the transition
    that many customers are going to have go through. Hmph, ditto the
    compilers and
    all the other stuff that required licenses. Dumb...

    What I like about IRIX is the strong sense of the way things 'just
    work'. 'Out of the
    box' is a cliche, but largely true when doing so many things with IRIX.
    For the moment,
    it still does everything I need, and at a speed I'm perfectly happy
    with (R12K/400
    Octane2).

    I wish I knew a lot more about programming, so many things I'd write if
    I could, codecs
    for O2, VPro accelerators for GIMP, etc. Oh well.



    Ian.

    SGI Depot: http://www.futuretech.blinkenlights.nl/sgidepot/
    Email: mapesdhs@yahoo.com (eBay ID: mapesdhs)
    Backup email (send copy to this too): sgidepot@blueyonder.co.uk


  18. Re: Future of SGI?

    Le Wed, 01 Feb 2006 00:41:17 +0000, Tony 'Nicoya' Mantler a écrit*:

    >
    > It depends on how you look at it. If IRIX had better prospects for the future,
    > then yes it would be seen as a shame. As it is, with commercial unixes
    > everywhere being marginalized and eliminated, we can at least be thankful that
    > the technologies in IRIX won't be lost with the OS, even if SGI itself
    > disappears.


    Have a look at this :

    http://www.informationweek.com/blog/...ix_dead_i.html

    Unix still represents a $2 billion [EUR 1.7 billion] market, the largest
    operating-system market by far. Despite Windows Server recent gains, it
    still represents about $1.6 billion [EUR 1.3 billion], when you're looking
    at operating system-only revenues. And Linux in terms of revenues
    represents one-tenth of what the good, gray Unixes combined represent.
    Granted the future belongs to Linux, but as a $2 billion market, is Unix
    dead?

    --
    Quis, quid, ubi, quibus auxiliis, cur, quomodo, quando


  19. Re: Future of SGI?

    HI,

    Emmanuel Florac wrote:
    > Le Wed, 01 Feb 2006 00:41:17 +0000, Tony 'Nicoya' Mantler a écrit :
    >
    >
    >>It depends on how you look at it. If IRIX had better prospects for the future,
    >>then yes it would be seen as a shame. As it is, with commercial unixes
    >>everywhere being marginalized and eliminated, we can at least be thankful that
    >>the technologies in IRIX won't be lost with the OS, even if SGI itself
    >>disappears.

    >
    >
    > Have a look at this :
    >
    > http://www.informationweek.com/blog/...ix_dead_i.html
    >
    > Unix still represents a $2 billion [EUR 1.7 billion] market, the largest
    > operating-system market by far. Despite Windows Server recent gains, it
    > still represents about $1.6 billion [EUR 1.3 billion], when you're looking
    > at operating system-only revenues. And Linux in terms of revenues
    > represents one-tenth of what the good, gray Unixes combined represent.
    > Granted the future belongs to Linux, but as a $2 billion market, is Unix
    > dead?
    >

    Unix(and clones) is not and will never go away I think, VMS tried, NT
    tried what next Vista?

    NT now adding more and more shell scripts, Linux is growing and we still
    have Solaris(which is opensourced) trying to get on the x64 train so
    Unix will most likely be around when I retire and that ain't soon I hope

    High end CPU arch. is more a problem I think, especially in my HW Eng.
    eyes, MIPS, Alpha and HPPA was great CPUs and now they are all gone as
    high end CPUs.


    /michael

  20. Re: Future of SGI?

    Tony 'Nicoya' Mantler wrote:
    > In article <1138746987.077118.6470@z14g2000cwz.googlegroups.co m>,
    > mapesdhs@yahoo.com wrote:
    >
    >
    > : Meanwhile, Linux still doesn't feel particularly professional to me, and lacks the
    > : integrated, ease-of-use feel that I get in IRIX. Trying to set something up in Linux
    > : feels too much like I'm expected to get my hands dirty downloading all sorts of extra
    > : packages and dealing with dependencies; why should I have to do that?
    >
    > I've been fairly happy with Debian. They've got pretty much every linux app in
    > existence in their repository, with full automatic dependancy fetching etc. Just
    > one command to install it and everything it needs. Keeping things up to date
    > with the latest versions is just as easy and automatic.
    >
    > It's certainly not the same experience as IRIX, but it's certainly very workable.
    >
    >


    Solaris 'x86 is pretty good these days. Virtually identical to Solaris on
    Sparc. Excellent price (free) with development software also available
    free (Sun's compilers). I think it has a lot more of a robust and well
    planned feel to it than Linux. I still prefer IRIX, where everything seems
    to just work the way it should, but given that migration is now inevitable,
    I think Solaris is worthy of consideration.

    --
    Dr Tristram J. Scott
    Energy Consultant

+ Reply to Thread
Page 2 of 3 FirstFirst 1 2 3 LastLast