The missing switch: High-performance monolithic graphene transistors created(extremetech.com)
extremetech.com
The missing switch: High-performance monolithic graphene transistors created
http://www.extremetech.com/computing/132988-the-missing-switch-high-performance-monolithic-graphene-transistors-created
3 comments
Say, we end up with 100x faster but 100x bigger transistors. A core that's 100x smaller (transistor-count) is not necessarily 100x slower. With a 100x clock speed-up, however - propagation delays will be the main limiting factor for speed.
So, if this tech ends up being used, should I expect small, highly-pipelined, single-core CPUs?
So, if this tech ends up being used, should I expect small, highly-pipelined, single-core CPUs?
One of the more interesting things here is the production process.
One of the reasons companies work so hard to make small scale features is that smaller scale means smaller die, smaller die means more devices per wafer, cost per wafer is fixed, more devices at a fixed defect rate means more yield (good devices), more yield means more 'value' per wafer.
Got that? The chain of events is predicated on the high cost of creating a semiconductor wafer.
Now if instead you can 'print' the circuitry, and you do mixed signal stuff. Then you can make your entire system as one board sized integrated circuit. Think an LCD television (transistors on glass) where instead of each transistor being the same you put a CPU and various other peripherals there. Size is not as important but yield still is.
So for things like phones/tablets, if you can do LED pixels (vs LCD) then you don't need the glass to be transparent (backlit) so you put down a layer of display, then a layer of compute, then a layer of memory, and finally a layer of radio. And voila, a device which looks like a piece of glass with black scum on the back that lights up and makes phone calls if you put power to the ends :-).
One of the reasons companies work so hard to make small scale features is that smaller scale means smaller die, smaller die means more devices per wafer, cost per wafer is fixed, more devices at a fixed defect rate means more yield (good devices), more yield means more 'value' per wafer.
Got that? The chain of events is predicated on the high cost of creating a semiconductor wafer.
Now if instead you can 'print' the circuitry, and you do mixed signal stuff. Then you can make your entire system as one board sized integrated circuit. Think an LCD television (transistors on glass) where instead of each transistor being the same you put a CPU and various other peripherals there. Size is not as important but yield still is.
So for things like phones/tablets, if you can do LED pixels (vs LCD) then you don't need the glass to be transparent (backlit) so you put down a layer of display, then a layer of compute, then a layer of memory, and finally a layer of radio. And voila, a device which looks like a piece of glass with black scum on the back that lights up and makes phone calls if you put power to the ends :-).
I might not understand the article correctly but only the contacts seem to be made of graphene. The canal is still made of SiC with far lower conductivity than graphene so it seems to me that the speed improvements wouldn't be as tremendous as what the article claims. Don't expect 100 GHz CPUs landing any time soon.
Yeah I was thinking about that too. Though if the limiting factor for switching speed is the impedance of the interconnects (I'm guessing it is), makes sense.
[1] http://www.nature.com/ncomms/journal/v3/n7/full/ncomms1955.h...