0:00:05i want my name is matt dispenser inaccurate a today to talk to about replacing
0:00:10transistors with mechanical switches
0:00:13upon hearing this
0:00:14you might rightly think why on earth would you do that
0:00:18i'm going to reply by taking a sort of roundabout route through history and pointing
0:00:22out that there's some poetry to this
0:00:24the first computers word fact mechanical the picture of data just difference engine
0:00:29which one powered by hand crank hoods all fit order polynomials
0:00:35if you best for the hundred years you get any which was the first fully
0:00:39digital electronic computer
0:00:41it way twenty tons consumed hundred fifty kilowatt power and perform the blazing five syllables
0:00:47per second of floating point operations
0:00:50now the difference between these two computers points out a tension in computer design that's
0:00:55been around since the eighteen forties which is between
0:00:58high powered high performance and lower power and lower performance
0:01:02fortunately we broke some of the design tradeoffs in a pretty significant way since the
0:01:06seventies
0:01:07but this tension actually resurfaced in a very significant way around two thousand
0:01:12this plot is a prediction of how computer power would increase from two thousand two
0:01:17thousand and ten and you lotus people expected power would increase a lot
0:01:23this obviously didn't happen as indicated by some of the annotations on the slide
0:01:29we don't have nuclear reactors in our laptops
0:01:31the question is how this happened
0:01:35then is actually prepared is very well for this the idea is that transistors have
0:01:40a property called their threshold voltage
0:01:42and if you just the threshold voltage properly you can trade off between two kinds
0:01:46of energy their dissipated in
0:01:48a very necessary energy called dynamic energy which has to do with running computer and
0:01:52weighted energy called leakage energy
0:01:55and
0:01:56by setting the threshold voltage properly you can actually find the minimum between them and
0:02:00make them operate perfectly
0:02:02this is actually what happened between two thousand two thousand ten in this why we
0:02:05have many course not computers right now uptalk three that based on the cartoon on
0:02:09the right the slide
0:02:10the idea is that if you were operating at the one x point in that
0:02:13cartoon you're consuming lots of dynamic energy above the optimum however you can slow yourself
0:02:18down in order to save energy
0:02:20and then stick to computers next to each other in order to recover your performance
0:02:23and let the software engineers figure out what to do with two computers
0:02:30you can do that again in this part you going from to x the four
0:02:32x parallelism in order to save energy but once you're cores at that point
0:02:38running it's lower won't save you any energy
0:02:41and so as a result
0:02:42you could this ceiling on parallelism which limits our ability to improve computing performance going
0:02:47forward
0:02:50now at this point my group likes it
0:02:52turns sharply in the left field and says
0:02:54the problem here is the transistor
0:02:56if we can replace that with something that doesn't have this wasted leakage energy
0:03:00then we can continue improving computing performance by scaling or voltage forever until we get
0:03:05some other physical them
0:03:07and what significant about this idea is that we've succeeded in building it
0:03:11this is a cartoon of the device that we build the ideas you have a
0:03:14piece of metal thing up near suspended by spring
0:03:17when you put a voltage on that piece of metal at home towards the surface
0:03:21and decorations on the bottom of connect different points on your chip
0:03:24the other significant things
0:03:26after we built that we measured it and found out that it has immeasurably low
0:03:29leakage as near as we can tell it has not
0:03:32so this means that we can replace the car drawing from the previous slide with
0:03:35the drawing in the bottom corner of this one where there's no leakage energy and
0:03:38we can just keep scaling forever
0:03:41now does this mean it's a good idea not necessarily
0:03:44we can call devices are big
0:03:46and their slow compared to electrons
0:03:48so there's a chance that we get a very energy efficient terribly performing computers if
0:03:53we tried use these
0:03:54however we've done a lot of very interesting work with circuit design in order to
0:03:58mitigate that
0:04:00problem in particular
0:04:02by changing our design style from stacking series of gates next to each other making
0:04:07very large distributed gates where all the input sit at the same time therefore all
0:04:11mechanical delay is incurred at the same time
0:04:14we can improve both are performance and advice count to save power energy and delay
0:04:21this is even more significant because we built some of these things we demonstrated that
0:04:24it is possible to get the functionality that we've been talking about and we don't
0:04:27a lot of extensive simulations showing that we can improve performance
0:04:32a result test case wasn't adder which we demonstrated twenty ten and simulations show that
0:04:35you can get ten x m for improvement over the absolute best transistor could do
0:04:42in terms of energy from the a ten x
0:04:44delay penalty
0:04:46we've also built a microprocessor and this other stuff coming the future including optimize memory
0:04:50structures minorities