Graphene Chip Reaches the Level of a 1979 Zilog Z8000

Max Shulaker and a group of researchers he leads has developed a working 16-bit microprocessor built from over 14,000 carbon nanotube transistors that is the most complex ever demonstrated. The techniques they have come up with can be implemented with equipment used for making conventional silicon chips, which means chipmakers won’t have to invest in expensive new gear if they want to make nanotube processors.

They got around the intermixing problem. They discovered that some kinds of logic gates, which are fundamental building blocks of digital circuits, were more resistant to problems triggered by metallic-like nanotubes than others. That led them to develop a new circuit design that prioritizes these gates, while minimizing the use of more sensitive metallic ones.

They solved the bundling problem. They coated a wafer in a polymer and then carefully washed it off in stages. This stripped off the nanotube clumps, leaving behind the monolayer needed to make the chip work most efficiently.

If they can rapidly scale up to the billions of transistors on current silicon chips then the graphene chips could be faster and use ten times less energy.

Nature – Modern microprocessor built from complementary carbon nanotube transistors

Written By Isaac Wang,

17 thoughts on “Graphene Chip Reaches the Level of a 1979 Zilog Z8000”

  1. I would like to learn more about holographic hypercube processors and hypercube algorithm programming. Photon geometry advancing past electron, flat, 2D topography. Silicon is a more complex structure, trigonal, than carbon’s tetrahedral geometry. Aqueous silicon computers are within people. Maybe we could explore water computers? For Plasma applications wrt holography.

  2. graphene rolled up into a ball = fullerine

    If you’re going to call anything with aromatic elemental carbon graphene you might as well admit that graphene was just a rebranding of graphite anyway and nothing to write home about in the first place.

  3. I suppose.  

    The more I’m thinking on it, the more it seems that the picayune current gating ability of each junction … and distributed trace capacitance, must be at the center of the low-clock-rate phenom. From what I’ve heard, the actual junctions switch at blazingly fast rates. However, in the analogy of fire engine hoses and tiny kiddie plastic squirt guns, each CNT is like the squirt gun. The stream of electrons is really, really small.  

    This in turn presents a probable cap on upside switching speed. Probably not 500,000× faster. Maybe 1,000× faster. I’ll be really (pleasantly) surprised if it rises above that in the next 10 years. However, it doesn’t rule out having really good uses for the things someday. Perhaps when conventional logic shrinks to the 2 nm (10 atom wide!) level, itsy-bitsy short CNTs will be able to switch better than the conventional gate semiconductor topologies.  

    In which case, “a win”. Just not even close, yet. 

    I really do like the fact tho that a whole, complete, working microprocessor was ginned up by these researchers. A lot of chutzpah, that. Big.

    Just saying,
    GoatGuy ✓

  4. The 2nd paragraph suggests they don’t know how to separate metallic from semiconducting nanotubes. In which case, they can’t make good conductors out of them yet.

    The 10 KHz, while 500000 times slower than today’s processors, is only ~1000 times slower than Zilog Z8000, to which it’s being compared here. For what is essentially an experimental proof-of-concept, that’s probably not too bad.

    They’re still trying to figure out the manufacturing process. Archtecture optimization comes after that. And of course the process affects the speed too.

  5. Quantum and digital solve different problems. They are both excellent at different things.

    Digital computers aren’t going away. Even if quantum totally surpasses digital then you will need digital to manage data and to make data for quantum.

  6. It needs to beat Quantum to still be of interest. Lots more money going into polynomial time then linear at the moment.

  7. It took a while to go from the transistor to the 4004, much less the latest microprocessors. It’ll be interesting to see where this goes – or if it ever gets beyond “Hey, this is something cool MIT did once upon a time.”

  8. Cool, and quite an achievement. However, from another reporting of this, I learned the chip only clocks at about 10 kilohertz.  That’d be ¹⁄₅₀₀₀₀₀ the speed of a modern Intel processor. In clock speed terms. Its not clear how much power was measured feeding the chip at-speed. What is clear is that 10 kilohertz is Really low. It begs wonder.

    The other thing of note is that the manufacturing process doesn’t exploit nanotubes’ conductivity to actually carry the digital ‘data’ around. It relies on conventional metal wires for everything data-and-power oriented. The nanotubes only play a role in being the ‘active channel’ in the (I don’t know what you’d call them, CNTFETs?) gates.

    Now you might object along the lines of “but Goat, silicon only plays the same role in conventional chips. Metal traces do all the data-and-power moving!” This is true. However, CNTs are supposed to have excellent conductivity powers. So, …

    Thinking further, the 10 kHz clock rate is particularly strange. I wonder if it is related to capacitance, and the ultra-low current that individual nanotubes are limited in handling. Perhaps this tech is awaiting the same kind of nearly-incredible speedup that happened between the “10 kilohertz scientific calculator” 1970s era, and today’s 12 core, 5 Gigahertz wonders.  

    Just saying,
    GoatGuy ✓

  9. The last paragraph explains that – if the technology can eventually scale up to near the number of transistors used in typical chips today, they could be faster and use much less power.

    Also, last I’d read of this technology, it was still a struggle to make more than a couple of gates – so this is far better. There has been some other progress using nanotubes for memory chips, but I haven’t heard much about using them for logic. So…impressive progress, with a LONG way to go yet.

    Though really they still seem to be using them as a sort of aggregate material, rather than (somehow) placing a few nanotubes precisely where needed. That could potentially limit how far they can scale it down. But again, this is progress.

  10. From the standpoint of their electronic properties and how they are processed at the wafer scale, they are completely different.

Comments are closed.