Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
AMD

AMD Makes 10-Nanometer Transistor 180

Yorrike writes: "Reuters is reporting that AMD are about to reveal their smallest double-gate transistor to date. From the article: 'The gate of the transistor, across which electrical current flows to turn the switch on, measures 10 nanometers, or 10 billionths of a meter.' The article goes on to suggest that this may lead to a 1 billion transistor chip."
This discussion has been archived. No new comments can be posted.

AMD Makes 10-Nanometer Transistor

Comments Filter:
  • by Syncdata ( 596941 )
    But how many of these transistors would fit on a single human hair?
    • Re:I'm confused (Score:2, Informative)

      by Raul654 ( 453029 )
      IIRC, a human hair is .25 mm in diameter. (.25 thousanths of a meter) The transistors are 10 nanometers (billionths of a meter)

      (.25*10^-3)/(10*10^-9)=25,000 transistors would fit into the diameter of a human hair.
      • "25,000 transistors would fit into the diameter of a human hair."

        They run all over, so be VERY CAREFUL not to spill them!
      • IIRC, a human hair is .25 mm in diameter. (.25 thousanths of a meter)

        No, it's around 0.05mm. That would be 5000 transistors then...

        Maybe you should try a pubic hair instead, they're slightly thicker.. :)

    • Certain newspapers always ask, 'how many of these would equal the height of a double-decker bus?'.
    • And how many libraries of congress can it process per second?
    • Silly rabbit, you don't put transistors on human hairs- you put them on chicken feathers.
  • by Anonymous Coward on Wednesday September 11, 2002 @02:38PM (#4239709)
    Dr. Evil: "And I will design a chip with transistors... 1 MILLION transistors!"

    #2: "Uh, Dr. Evil, we've already designed a chip with 1 million transistors, quite a few years ago in fact."

    Dr. Evil: "How many is a lot then?"

    #2: "1 billion would be very impressive. We're working on that right now."

    Dr. Evil: "Alright then... 1 BILLION transistors."
  • The gate of the transistor, across which electrical current flows to turn the switch on...

    You mean there's another type of transistor?
    • I may be going of topic, but there actually are two types of transistors: bipolor and FET.

      Bipolor is fast, but very difficult to manufacture on the small scales required for microprocessors.

      FET is slower, but is relativly easy to manufature on small scales. This is manufatured by over laping two thin "strips" of silicon over eachother like a cross. The width of the silicon "strip" is the path size. For the new chip this would be (correct me if I'm wrong) 10nm.

      Most chips use FET transistors because of the difficulties of manufaturing bipolor transistors.

      (I did have cool ascii-art diagrams of the two types of transistors, but aparently trying to explaing things well is considered lame on slashdot, and the lameness filter tried to stop me. ARGHHHH!!!!)
  • Same. (Score:5, Informative)

    by Perdo ( 151843 ) on Wednesday September 11, 2002 @02:42PM (#4239743) Homepage Journal
    This is the same transistor that IBM was talking about. They are sharing (or stealing) this technology. Electron microscope pictures are of the same transistor.
    • Or stealing? Was that comment even necessary? It's no secret that AMD and IBM have worked together in the past.
    • Re:Same. (Score:2, Informative)

      by unicron ( 20286 )
      Intel employee: You built a 10nm transistor! We we're doing that! You must've stolen it!

      AMD employee: You are aware we make the same product, right? Any technological advances either one of us makes, rest assured the other isn't far behind. Surely you don't think you are the only cpu manufacturer interested in reducing transistor size, right?

      Intel employee: I'll shut up now.
      • So how come Intel is demonstrating 4.1GHz processors (even overclocking them to 4.7GHz!) while AMD hasn't even broken the 2GHz mark?

        • Because AMD isn't like that. They're a very "show me the money" type company: look when the new Thoroughbread Rev B came out. Little fanfare, it just happened.

          Now if Intel were actually shipping 4.1GHZ CPUs, I might be impressed. Remember that they can probably only maintain this frequency with hand-picked CPUs and components, watercooling (posibly peltier as well), large fans, etc. It's just not reasonable so far.
        • AMD's never really played the bandwidth game. The issue with Pentium IV's is that the high gigahertz readings they have are quite misleading. For one, they can do far less work during each cycle. They may have more cycles, but the effect is lessened in that the AMDs can do far much more work per individual cycle. Second, the cache is much worse in the Pentium IV's. They have inclusive caches, which because of their design cause the chip to have to go to much-slower memory more often than the exclusive cache designs of the AMD Athlon processors.

          Plus, you've got to remember that these "fast" Pentium IV's are months away from even being introduced on paper. By that time the comparably-performing Hammers should be out, running both 32 and 64-bit applications and from what I hear quite fast as well. Basically, the difference is this: Intel caters to the average Joe who thinks bigger numbers are better, while AMD takes much more care to ensure that their chips run at a fast, reliable speed that rivals that of higher-rated Pentium IV's.
  • by chris_mahan ( 256577 ) <chris.mahan@gmail.com> on Wednesday September 11, 2002 @02:44PM (#4239756) Homepage
    What, where's not going to benchmark on Mhz cycles anymore?

    I can already picture this:

    "Buy Intel's new Pentra Plus 2004, the only processor with a billion transistors!!!" (yay, men in blue dance yadda yadda)

  • Faster invalidation of digital media than you ever thought possible!
    GMFTatsujin
  • by Snarfvs Maximvs ( 28022 ) on Wednesday September 11, 2002 @02:49PM (#4239794)
    Great, but AMD needs to get their .13 micron process stable first.
  • That's not a billionth, that's only a thousand millionth.

    A billionth would be 10^-12

    Silly americans.....
    • It's probably not the best reason for my first /. post but it was a billionth, 10^-12 is a trillionth. It would be more logical for a million millions to be a billion, because a thousand thousands is a million, but by that reasoning a thousand should be one hundred hundreds and not ten hundreds. Anyway forgive the pedantic first user post but there's too much misinformation on /. already. Oh and the "silly americans" comment makes me fearful you might be british, say it ain't so!
      • nah, I'm not British, I'm a New Zealander!

        We do in NZ however follow the British definition for a billion, which IS one million million....

        Welcome to slashdot :)
        • Thanks for the welcome :)

          I'm actually a brit myself, I never knew we used a different definition of billion here (and there too it seems), in fact when I was at school doing physics and economics A-levels we were taught to use the 10^9 version.

          Oh well, I expect it came over with McDonalds or something. Thanks for educating me on my own country's mathematical scales, heh.
        • Um. No.

          I'm not sure which New Zealand YOU live in, but the New Zealand standard for 1 billion is 1*10^9 not 1*10^12

          This is one of the rare occasions where I agree with the American definition, since it's useful. Who the hell would want to say 1 thousound million when they can just say 1 billion?

          • I know, replying to my own post, but to avoid further embarrassment, I will confirm that no one would ever say "thousound", but rather "thousand".

            ....... still a billion is a thousand million.

          • So you obviously live in a lazy incorrect part of NZ.

            My guess is Auckland.
            • No, I'm not JAFA.

              And I'm not lazy. As I said, it's one of the few American definitions I agree with. I still spell colour with a "u", night with a "ight" and I know the difference between "then" and "than".

              • > I know the diffence between "then" and "than"

                And the difference between "insure" and "ensure" - I hate that one.

                How do Americans spell "light"? I bet they have some silly way of spelling "hiccough" too.
        • The two naming conventions were in fact devised at different times by the Frenchmen Chuquet (1B=10e12) and De la Roche (1B=10e9). The De la Roche system was in vogue at the time of the revolutionary war (or the colonial rebellion take your pick) and that usage has remained current in America ever since.

          This is similar to the changed spelling and pronounciation of aluminum (=> aluminium) in an attempt to systematize its naming. This only occured after the industrial revolution made the once rare metal commonplace and hence the original naming has been preserved in America.
  • by rot26 ( 240034 ) on Wednesday September 11, 2002 @03:00PM (#4239877) Homepage Journal
    Just think about all the DRM they can pack into that baby!
  • Press Release Slants (Score:5, Informative)

    by FuzzyDaddy ( 584528 ) on Wednesday September 11, 2002 @03:05PM (#4239908) Journal
    It's amazing how two press reports on the same breakthrough claim it's significant for entirely different reasons.

    The earlier press release talked, in a half assed way, about the performance benefits. This one talks, in a half assed way, about the reduced size.

    Of course, this happens because whoever digests these things for us unwashed masses doesn't understand what the hell they're talking about.

    And the gate length has much more to do with the performance, not the transistor density, as the transistor density is dominated by the via sizes and interconnect sizes.

  • by echucker ( 570962 ) on Wednesday September 11, 2002 @03:06PM (#4239914) Homepage
    .... this would be the "wee" size. "Not so wee" and "friggin' huge!" are now obsolete.
  • "So small that they power the palm pilot used by the angel dancing on the head of a pin?"
  • Heat Sink? (Score:1, Funny)

    by Anonymous Coward
    Imagine the size of the heatsink you'd need for a chip with 1 billion transistors.
    • The HP-PA8800 is already a thrid of the way there with over 300 million transistors. For more info see here [lostcircuits.com] . At the rate of Moore's Law we will be to a billion transistors on the biggest chips in less than 3 years.
  • ...and with the new DRM capabilities, it'll be able to do so much less.
  • >For its part, Intel INTC.O plans to detail its development and manufacture of triple-gate transistors in a paper it is delivering at a technical conference next week in Japan, Intel President and Chief Operating Officer Paul Otellini told an Intel conference on Monday.

    I do research on double-gate transistors. I have no clue what they mean by a triple-gate transistor. I think that this is another lame intel PR attempt. It's probably like when they made up a new name for FDSOI and called it depleted-substrate or some BS like that and got laughed at. They must want to be one gate better.
  • What pisses me off is that al this is well and good..intel finds this..amd finds that... but i hate WAITING for the tech. to actually be used to some benefit to the end user.

    Yea. this means smaller and faster chips for amd..but when...10 years from now?
  • When will AMD realize that people don't care how *fast* their PC is, but that it has a life affirming, penis enlarging number attached to it!

    How does this sound; the AMD 10 GHZ noipcon. Sure, it has the IPC of an 8088, but 10 GHZ -- lets see Intel beat *that*!
  • by GeneralEmergency ( 240687 ) on Wednesday September 11, 2002 @03:37PM (#4240122) Journal


    Once the RIAA gets news of this I'm sure they'll find some hack think-tank that will make the proclamation that according to their data, there is a direct correlation over time between "Transistors Per CPU" and "Music Piracy". Lawyer parties will follow.



    (When I started typing this, I was making a joke. But now that I've read it back a few times, this looks more probable than improbable.)

  • ... (snaps suspenders)

    I've got a 9 nm transmitter right here in my pocket I just whipped up the other day.

  • CMOS Technology ??? (Score:4, Interesting)

    by Proudrooster ( 580120 ) on Wednesday September 11, 2002 @03:42PM (#4240167) Homepage
    "AMD said its laboratory demonstration of 10 nanometer
    Complementary Metal Oxide Semiconductor Fin Field Effect Transistor (CMOS FinFET) is the outcome of collaborative research between AMD and the University of California, Berkeley"
    CMOS technology with a silicon "fin" to prevent electron leakage?
    Where on earth am I going to find a motherboard for a CMOS processor?
    First CMOS, then Silicon, then Gallium (what ever happened GaAs anyway?), then Coppermine, and now back to CMOS with a silicon fin.....
    Mmmrrrppphh, full circle you have come young Jedi.
    -Yoda
    • First CMOS, then Silicon, then Gallium (what ever happened GaAs anyway?), then Coppermine, and now back to CMOS with a silicon fin....

      You must not have watched Sesame Street as a child.

      Which of these things doesn't belong?
    • WTF are you talking about? CMOS is used by everyone to make chips. It's a circuit technology, not a material. Silicon is a material, CMOS is the way you put 2 complementary transistors MADE OUT OF SILICON (and other stuff) together to create an inverter (or whatever other logic you want).
  • Now I'm wondering if I should concern myself with getting a hold of a clawhammer or wait until they re-engineer them on a 0.01 micron process :) And since AMD acquired Alchemy Corp. imagine what kind of power you could pack in a PDA... say a really low power duron at 1GHz with some DDR chips (not modules of course) soldered on the PDA's board and maybe a mobile GeForce chipset...

    Ah but that's probably 10 years down the road anyways... Sometimes it's painful to know about future technologies, but not be able to get em yet :)

  • FinFET Technology (Score:2, Insightful)

    by taylor ( 11728 )

    It seems the real question of relevance rests in the new technology they're using to build these devices. The FinFETs have a nice writeup here [compoundse...ductor.net]. They can be built just with the defects from plateaus in normal photo-lithographic processes, thus using the nicely developed techniques usually limited to 125 nm structures to build 10 nm structures. This still means the overall transistor size will be on the order of a a few hundred nm, to deal with contacts, etc, but it is a sight better than standard 0.13 micron transistors, and much easier to use in mass production than e-beam lithography. (Just think about those old vector displays -- that's ebeam lithography for you). Seems like a fine idea for nanoscale structure building, and not one of these technologies may have impacts far beyond just standard IC circuit technology; with 10 nm devices, all sorts of quantum coherent processes become accessible, if you work for them.

  • ... I would have expected tail fins measuring a little bit more than a measly 10nm !
  • 10 Nanometer you say? Whats that in Thickness of Human Hair?
  • Oh wow! So three gates on a transistor! Now we can move past binary onto trinary. This will invalidate lots of old code, but I'm sure they can emulate! Of course Microsoft has been running trinary in Windows for years. The three states are ON, OFF, and CHOKE.

    The ultimate system, in my opinion, would use a multistate transistor with ON, OFF, and a dimmer function for everything in between.

    • There is, it&amp;apos;s called the transcapaciter. I heard about this a few years ago on Art Bells radio program. There might be a patent on it. Basically it was a device that measured 1 and 0 and then the signals inbetween. Neet stuff but I don&amp;apos;t know how it could be used. - ShawnX Posted from my blackberry =)
  • The article goes on to suggest that this may lead to a 1 billion transistor chip.

    Does it require a radiator to stay cool?
  • It's rather funny to watch AMD and Intel battle it out in the R&D departments, arguing over whose silicon dick is smaller.
  • Insanely Large Scale Intergrated Circuits?

Sendmail may be safely run set-user-id to root. -- Eric Allman, "Sendmail Installation Guide"

Working...