IBM's new 110 GHz chip

rupertrupert
Posted:
in Future Apple Hardware edited January 2014
No joking!



Too bad this thing won't be in any Apple computer for less than $500 million!



Anyway, it's too bad IBM can't do this well on the desktop market...



<a href="http://story.news.yahoo.com/news?tmpl=story&u=/nm/20020225/wr_nm/tech_ibm_dc_25"; target="_blank">http://story.news.yahoo.com/news?tmpl=story&u=/nm/20020225/wr_nm/tech_ibm_dc_25</a>;
«1

Comments

  • Reply 1 of 31
    sameoldshtsameoldsht Posts: 66member
    Nice for IBM.



    This has absolutely on the PowerPC chip though therefor no baring on the mac community whatsoever.
  • Reply 2 of 31
    eskimoeskimo Posts: 474member
    That's a SiGe HBT (Heterojunction Bipolar Transistor). Basically by using varying amounts of Germanium mixed with the silicon you can narrow the band gap of the base region and greatly increase your base transport factor. Bipolar Junction Transistors (of which this is a subset) have been running at very high frequency for awhile now even at larger process nodes. Far in advance of the CMOS technology which all microprocessors are made with.



    But the downside to BJTs is two fold. One, the design of BJTs prevents dense packing of transistors meaning it takes considerably more silicon to put down the same amount of transistors. In the semiconductor industry silicon area IS money. Also their is no nice easy way to combine bipolars to form basic building blocks of logic (such as inverters) without considerable off state power drain. CMOS has a lot of headroom left in it before we have to consider alternatives like SiGe. But for communications devices this is certainly a great technology, only took them 20 years .



    [ 02-25-2002: Message edited by: Eskimo ]</p>
  • Reply 3 of 31
    pookjppookjp Posts: 280member
    Steve will demonstrate that a Dual 1GHz will kick its ass in the Gaussian blur.
  • Reply 4 of 31
    baumanbauman Posts: 1,248member
    The thing this chip demonstrates is the process. Couldn't it be applied to other chips (i.e. PPC)?



    ~bauman
  • Reply 5 of 31
    willoughbywilloughby Posts: 1,457member
    [quote]Originally posted by Eskimo:

    <strong>That's a SiGe HBT (Heterojunction Bipolar Transistor).</strong><hr></blockquote>



    Dude, my mom had that same problem. <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
  • Reply 6 of 31
    crusadercrusader Posts: 1,129member
    [quote] Some IBM competitors are also working on developing faster microchips based on silicon germanium technology. Conexant Systems Inc. , a small chip company based in Newport Beach, California, said at a December conference that it is working on a circuit that runs at 200 GigaHertz. <hr></blockquote>

    :eek:



    wow...



    [ 02-27-2002: Message edited by: Crusader ]</p>
  • Reply 7 of 31
    eskimoeskimo Posts: 474member
    [quote]Originally posted by bauman:

    <strong>The thing this chip demonstrates is the process. Couldn't it be applied to other chips (i.e. PPC)?



    ~bauman</strong><hr></blockquote>



    No, it demonstrates an advanced process for a bipolar transistor. Not a MOS transistor which all of today's microprocessors are made with. IBM has demonstrated advanced MOS transistors in the past, I believe at 30nm in size switching at 10-20GHz.
  • Reply 8 of 31
    g-newsg-news Posts: 1,107member
    As far as I know the fastest conventional transistors (that's transistors, not chips made of millions of transistors) currently clock around 70GHz.



    G-News
  • Reply 9 of 31
    bogiebogie Posts: 407member
    the chip is for telecommunications networks, not PCs.
  • Reply 10 of 31
    bodhibodhi Posts: 1,424member
    IBM has the R&D dept to beat all R&D depts. Love that company!
  • Reply 11 of 31
    outsideroutsider Posts: 6,008member
    Exponential wanted to make a 604 variant made on a bipolar process running at 466-600MHz (back when a 250MHz 604 was the king). Must have been their old fabs but it was 600nm! At the time the regular 604e was being manufactured on 250nm... But there were samples running up to 500MHz I think. I wonder if a bipolar processor can be made now with advanced technology..
  • Reply 12 of 31
    g-newsg-news Posts: 1,107member
    Problem was that a 450MHz prototype Exponential chip that was tested by a German Mac mag back then performed worse than a 300MHz 604e.

    The world, and Exponential themselves weren't ready for the tech.



    G-News
  • Reply 13 of 31
    eskimoeskimo Posts: 474member
    [quote]Originally posted by Outsider:

    <strong>Exponential wanted to make a 604 variant made on a bipolar process running at 466-600MHz (back when a 250MHz 604 was the king). Must have been their old fabs but it was 600nm! At the time the regular 604e was being manufactured on 250nm... But there were samples running up to 500MHz I think. I wonder if a bipolar processor can be made now with advanced technology..</strong><hr></blockquote>



    Yes Exponential's chip was a BiCMOS device utilizing both bipolar and CMOS technology in the same device. Bipolar is very nice since it can operate at high frequencies at much larger features sizes. The problem as I stated earlier in this thread is that the power consumption and heat produced are enormous for the complexity required for a modern microprocessor. In order to make it use a reasonable amount of power they had to scale back on the number of transistors and thus as you stated were left with an underperforming part. To use that technology to compete with today's multi-issue, mult-execution superscalar processors would be futile.
  • Reply 14 of 31
    tigerwoods99tigerwoods99 Posts: 2,633member
    That's pretty interesting.



    Does anyone know the last time Macs had a MHz lead on x86? I'd say probably when the 350 MHz Mach Vs were out, or am I wrong?
  • Reply 15 of 31
    amorphamorph Posts: 7,112member
    That's about right.



    Of course, the 604ev's needed lots of bandwidth to perform well, and they were hopelessly starved on the 50MHz busses Apple used.



    Had Apple bumped up the bus/memory speed, those machines would have been much, much faster. Of course, they would also have been much, much more expensive...
  • Reply 16 of 31
    powerdocpowerdoc Posts: 8,123member
    110 ghz is not enough to stay competitive versus intel



    Just kidding, but to your wize advice, will it be true in 20 years ?.

    We will check this in time if AI will be still here. Em an will have 100 000 post at this time
  • Reply 17 of 31
    macaddictmacaddict Posts: 1,055member
    steve jobs



    "No, see, our new 1GHz Dual Processor can easily beat this 100GHz IBM crap. See it speed up the rendering of this Toy Story 3 poster! Wow, look at those lickable buttons!"



    /steve jobs
  • Reply 18 of 31
    jasonppjasonpp Posts: 308member
    bump
  • Reply 19 of 31
    tarbashtarbash Posts: 278member
    If IBM can implement this technology into the PowerPC chip soon, the days of having higher MHz are going to be back....
  • Reply 20 of 31
    amyklaiamyklai Posts: 29member
    [quote]Originally posted by Tarbash:

    <strong>If IBM can implement this technology into the PowerPC chip soon, the days of having higher MHz are going to be back....</strong><hr></blockquote>



    Have fun waiting for that...
Sign In or Register to comment.