1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

When will 90nm technology be "ready for prime time"?

Discussion in 'Intel' started by AJ, Dec 18, 2004.

  1. AJ

    AJ Guest

    Intel admits that 90nm technology has problems (heat production from
    current leakage). Will 90nm chips ever run as cool as 130nm chips?
    Do you think the marketing folks at Intel sat around and decided that
    more heat at this juncture is an asset since it is the only way to sell
    the otherwise unneeded BTX case designs?

    AJ
     
    AJ, Dec 18, 2004
    #1
    1. Advertisements

  2. AJ

    KMS Guest

    90nm chips dissipates significantly less power and also run cooler when
    compared to 130nm chips. Look the new AMD Athlon64 chips power consumption
    and temperature figures.

    http://www.hothardware.com/viewarticle.cfm?articleid=613&cid=1

    Too bad Intel did not make their homework before releasing their 90nm
    chips... ; )
     
    KMS, Dec 19, 2004
    #2
    1. Advertisements

  3. AJ

    Yousuf Khan Guest

    AMD's 90nm is already ready for primetime. In fact they're already into
    their second season of prime-time. They're on their second generation
    chips built using the 90nm technology.

    Yousuf Khan
     
    Yousuf Khan, Dec 19, 2004
    #3
  4. AJ

    AJ Guest

    Well, as you could have reasoned logically since this is an INTEL group, I
    was asking about Intel chips and when _they_ would be ready for prime time.
    My concern being whether I'm going to have to build "hot" PCs in the interim
    before Intel "fixes" their technology.

    (Aside: everyone may assume that unless I specifically state otherwise, that'
    I'm not interested in brands other than Intel when I post in here: I'll post in the
    other groups for info on other brands if I need it).

    AJ
     
    AJ, Dec 20, 2004
    #4
  5. AJ

    AJ Guest

    Not Intel chips!

    AJ
     
    AJ, Dec 20, 2004
    #5
  6. AJ

    Alex Johnson Guest

    My guess is that intel screwed up somewhere in the process numbers and
    90nm is a permanent and unfortunate bump in the road. I expect 65nm
    adoption to be accellerated by intel to move back into the flight path,
    cover up the bungled 90nm work, and claim "one generation ahead".

    In the mean time, P4 90nm is a really REALLY bad choice. Pentium-M on
    the other hand consumes 21W at 90nm and performs comparably with P4s
    consuming 100-110W. Buy smart.

    Alex
     
    Alex Johnson, Dec 20, 2004
    #6
  7. AJ

    Yousuf Khan Guest

    Well, then I'd say the answer is no, they very likely won't be able to
    fix their current 90nm generation. They made too many wrong decisions
    with it. It would require buying new equipment and if you're going to
    buy new equipment, you might as well get prepared for the next shrink.
    Pentium 4 is a lost generation on 90nm.

    However, their Pentium M might be quite salvageable at 90nm. It's not
    nearly as close to wall as Pentium 4 was.
    It's always useful to know what other manufacturers are doing.

    Yousuf Khan
     
    Yousuf Khan, Dec 20, 2004
    #7
  8. AJ

    AJ Guest

    Slim pickens for desktop motherboards, and expensive too. I'll wait for Intel
    to produce one.

    AJ
     
    AJ, Dec 21, 2004
    #8
  9. AJ

    AJ Guest


    I read an article today that said that Pentium M runs "cool" though, with
    a max power dissipation of about 21 watts. Is the 90nm technology different
    than in P4's (it must be!)? How?
    That's for the other groups though.

    AJ
     
    AJ, Dec 21, 2004
    #9
  10. AJ

    Yousuf Khan Guest

    No, it's the same process as used on P4. However, Pentium M was running
    at much slower frequencies than P4 (i.e. "farther from the wall"). PM
    is running cool, but not quite as cool as its old 130nm version -- at
    idle it still consumes more power than 130nm PM. The fortunate thing
    about PM is that it's got all of that dynamic Speedstep power
    management circuitry which can be tuned up or tuned down to overcome
    manufacturing deficiencies; with a more aggressive power-cutting
    schedule, PM can be kept within its old overall power-consumption
    envelope.

    Yousuf Khan
     
    Yousuf Khan, Dec 21, 2004
    #10
  11. AJ

    Alex Johnson Guest

    I wish independents would manufacture more good desktop Pentium-M
    boards. Intel has said the desktop future is Pentium-M, benchmarks have
    shown it to be a competitor even against the P4-EE in gaming. But it's
    plainly evident intel is having some kind of internal struggle because
    despite announcing that the future hinges on Pentium-M, it won't sell
    them into the desktop market until end of 2006 to early 2007 with
    "Merom". In the mean time they keep pushing P4 crap at us like we're
    going to forget that it sucks.

    Alex
     
    Alex Johnson, Dec 21, 2004
    #11
  12. AJ

    Grumble Guest

    On a related note, the PM is fairly strong in SPECint2000
    benchmarks, but gets its ass kicked in SPECfp2000.

    I suppose the FP computation is done in the GPU these days? Thus
    integer performance would be what matters for games?
     
    Grumble, Dec 21, 2004
    #12
  13. AJ

    AJ Guest

    But the PM at 2 GHz performs like the P4 at 3 GHz. So why bother with a design
    like P4 when PM is apparently better at everything?

    AJ
     
    AJ, Dec 22, 2004
    #13
  14. Hyperthreading. *IF* you can take advantage of it it's a moderate win.
    If you want to be able to run/test SMP code, HT is vary valuable. For
    most Windows applications that's not the case, just pointing out that PM
    is not better at everything.
     
    Bill Davidsen, Dec 22, 2004
    #14
  15. AJ

    Yousuf Khan Guest

    Ask Intel marketing. :)

    My speculation is that Intel still believes that raw Mhz numbers sell
    processors. There may be some truth in that still, it's very hard to
    convince people otherwise once they've been brainwashed to think in a
    certain way.

    Right from the beginning of the Pentium 4 series, people have been
    accusing Intel of subverting the whole Mhz rating system to the point
    where it was a joke. From the 8086 right on upto the Pentium 3, Intel's
    processors basically got progressively faster both in terms of Mhz and
    actual performance. So it became a standard colloquial way of comparing
    performance by simply comparing Mhz, and for the most part they were
    correlated. The Pentium 4 was the first processor in the family which
    increased its Mhz, but fell behind in its actual performance in a big
    noticeable way. I'm not sure what image Intel is concerned about, (a)
    convincing people that going back in Mhz doesn't mean going back in
    actual performance, or (b) convincing people that going back towards
    the Pentium 3 core would be better for performance.

    Yousuf Khan
     
    Yousuf Khan, Dec 22, 2004
    #15
  16. AJ

    AJ Guest

    But certainly PMs could be made with HT also (?).

    AJ
     
    AJ, Dec 23, 2004
    #16
  17. AJ

    AJ Guest

    Nah. That's why they changed to processor numbers that supposedly
    indicate many of the features and not just clock speed.
    Personally, I stopped noticing when the industry started talking in GHz.
    I think the thrust is on features now and performance is taking a back seat
    (as it should be for the mainstream user).

    AJ
     
    AJ, Dec 23, 2004
    #17
  18. The chip was not designed to have that feature AFAIK, so it would take
    design of a whole new chip with features of P4 to do so. I believe all
    P4 CPUs had a status bit for HT enabled, three bits for number of
    siblings, and three for sibling ID, so HT was a design goal from the
    start. Someone with the manuals handy feel free to clarify or correct.

    The new IBM Power-5 chips have both multicore and SMT capability, and
    according to InfoWorld shared L2 and dedicated L1 and L3 cache.

    I have no idea if Intel could or would provide HT on future CPUs, or if
    they will settle for dual core and concentrate on getting the chips out
    the door.
     
    Bill Davidsen, Dec 23, 2004
    #18
  19. AJ

    Yousuf Khan Guest

    Doesn't that sound a little freaky? I mean that they have the mid-level
    cache shared but not the last level cache? You'd assume that if they
    made one shared cache, that it would be the last level one.
    The HT identification flags allow for upto 256 virtual processors to be
    identified per chip. Some of them could be full cores, some of them
    could be SMT cores. I don't know if either Intel or AMD will push their
    designs out to that far so quickly. I'd expect we'll see higher levels
    of virtualization in two or three generations from now.

    Yousuf Khan
     
    Yousuf Khan, Dec 23, 2004
    #19
  20. AJ

    AJ Guest

    Indeed it seems like the thing to pursue at this point: multi-core rather than HT.
    So Perhaps a good thing that it DOESN'T have HT.

    AJ
     
    AJ, Dec 24, 2004
    #20
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.