200GHz iMac?

Posted:
in Future Apple Hardware edited January 2014
From: <a href="http://www-916.ibm.com/press/prnews.nsf/jan/BEBB30C4F4637CCC85256A7600498511"; target="_blank">http://www-916.ibm.com/press/prnews.nsf/jan/BEBB30C4F4637CCC85256A7600498511</a>;



[quote]IBM Announces World's Fastest Silicon-Based Transistor



New Chip Technology Breaks Speed Barrier at Ultra-Low Power

EAST FISHKILL, N.Y.--June 25, 2001--IBM today announced it has built the world's fastest silicon-based transistor, a basic building block used to make microchips.



IBM expects the new transistor will drive communications chips to speeds of 100GHz within two years -- five times faster and four years sooner than recently-announced competitive approaches.



The transistor uses a modified design and IBM's proven silicon germanium (SiGe) technology to reach speeds of 210 GigaHertz (GHz) while drawing just a milliamp of electrical current. This represents an 80 percent performance improvement and a 50 percent reduction in power consumption over current designs.<hr></blockquote>



Will this make its way into the PowerPC architecture? Oh yeah. I'd say IBM is definately the way to go. Die Motorola, Die!



[ 08-11-2002: Message edited by: Faeylyn ]</p>

Comments

  • Reply 1 of 11
    amorphamorph Posts: 7,112member
    This is not all that much faster than the speed of a current transistor - not as fast as you think.



    A transistor is one little tiny piece of a processor. You need tens of millions of them working together, and most of them work in chains, so that it takes time for the current to settle down to a final value. By the time you've got them all arrayed, the whole chip will run at a much lower speed than any of the individual transistors are capable of. This is good news, but don't expect IBM to ship whole processors running in the hundreds or even tens of gigahertz any time too soon.
  • Reply 2 of 11
    sc_marktsc_markt Posts: 1,401member
    [quote]Originally posted by Amorph:

    <strong>



    A transistor is one little tiny piece of a processor. You need tens of millions of them working together, and most of them work in chains, so that it takes time for the current to settle down to a final value. By the time you've got them all arrayed, the whole chip will run at a much lower speed than any of the individual transistors are capable of. This is good news, but don't expect IBM to ship whole processors running in the hundreds or even tens of gigahertz any time too soon.</strong><hr></blockquote>



    Did you see the time frame from IBM?



    "IBM expects the new transistor will drive communications chips to speeds of 100GHz within two years -- five times faster and four years sooner than recently-announced competitive approaches."
  • Reply 3 of 11
    myahmacmyahmac Posts: 222member
    note: they did not say CPU's communications chips are very diferent
  • Reply 4 of 11
    amorphamorph Posts: 7,112member
    [quote]Originally posted by sc_markt:

    <strong>



    Did you see the time frame from IBM?



    "IBM expects the new transistor will drive communications chips to speeds of 100GHz within two years -- five times faster and four years sooner than recently-announced competitive approaches."</strong><hr></blockquote>



    Yup. Communications chips are made with a different, more expensive process that yields much higher speeds. They're also really simple and highly specialized.



    No 100GHz CPUs out of this, sorry.
  • Reply 5 of 11
    You guys are boring.



    Okay, how's this? It will help get Firewire2 (communications) scaling up to the spec'd 3.2Gbps a lot quicker and will be a key component of Firewire3?
  • Reply 6 of 11
    sc_marktsc_markt Posts: 1,401member
    [quote]Originally posted by Amorph:

    <strong>



    Yup. Communications chips are made with a different, more expensive process that yields much higher speeds. They're also really simple and highly specialized.



    No 100GHz CPUs out of this, sorry.</strong><hr></blockquote>





    Fair enough. But isn't the G4 also a communication chip?



    Note that a while ago, I read an article from a technicial publication on the internet about an IBM process improvement that was going to speed up transistors to 200GHz. I distinctly remember reading a passage in it where somebody from IBM noted that it'll speed up PowerPC chips such as the ones that go into Apple computers (I'm paraphrasing here). He also then made a comment to the effect that, if I remember correctly, was something like 'the turtle was going to blow past the hare' (again, I'm paraphrasing but this was the basic gist of this part of the article).



    I've been trying to find that article for a while. I'll try again now that I have a good time frame of when IBM posted it.
  • Reply 7 of 11
    amorphamorph Posts: 7,112member
    [quote]Originally posted by sc_markt:

    <strong>Fair enough. But isn't the G4 also a communication chip?</strong><hr></blockquote>



    Not the same class. It's not made with Gallium Arsenide, and it's more useful in high-end routers than in high-end phone switches.



    [quote]<strong>Note that a while ago, I read an article from a technicial publication on the internet about an IBM process improvement that was going to speed up transistors to 200GHz. I distinctly remember reading a passage in it where somebody from IBM noted that it'll speed up PowerPC chips such as the ones that go into Apple computers (I'm paraphrasing here). He also then made a comment to the effect that, if I remember correctly, was something like 'the turtle was going to blow past the hare' (again, I'm paraphrasing but this was the basic gist of this part of the article).



    I've been trying to find that article for a while. I'll try again now that I have a good time frame of when IBM posted it.</strong><hr></blockquote>



    Oh, it'll speed up any processor that IBM applies the technology to, I don't doubt. But CPUs won't suddenly leap into the tens or hundreds of GHz.
  • Reply 8 of 11
    rick1138rick1138 Posts: 938member
    The NSA won't allow it.
  • Reply 9 of 11
    ebbyebby Posts: 3,110member
    It's time for GaAs processors!



    Gimmie gimmie gimmie :eek:
  • Reply 10 of 11
    [quote]Originally posted by Rick1138:

    <strong>The NSA won't allow it.</strong><hr></blockquote>



    Overheard:

    "Jimmy, I've told you a hundred times! Stop reading the President's mail!"



    256bit (minimum) encryption bites the big one.



  • Reply 11 of 11
    mmicistmmicist Posts: 214member
    [quote]Originally posted by Faeylyn:

    <strong>





    Will this make its way into the PowerPC architecture? Oh yeah. I'd say IBM is definately the way to go. Die Motorola, Die!



    [ 08-11-2002: Message edited by: Faeylyn ]</strong><hr></blockquote>



    This is a bipolar transistor, not an MOS transistor, it is poorly suited to logic as currently envisioned, and has an enormous power draw per transistor, which is also pretty much independent of switching frequency. A single transistor draws 1mA, a 30 million transistor chip would draw 30,000 Amps, and dissipate 30,000 W if running at 1 V. (Yes, I know it could be scaled down, but not by very much without losing significant speed)

    This is a marvellous achievement by IBM, and in principle, a few such devices might find their way into critical sections of logic chips, but I would'nt want to be involved in the process development for merging the two processes. Motorola have a nice process for merging GaAs and Si devices on one wafer, which seems nearer reality than this.



    michael
Sign In or Register to comment.