Van Neumann?
Who is he? Wakipedia keeps giving me von Neuman instead.
(CPU's with a separate cache for each core are a step past von Neumann architecture, which assumes one memory device sharing code and data for a computer with only one core.)
The United States' Defense Advanced Research Projects Agency (DARPA) wants to find the electronics industry's next iteration of Moore's Law and has loaded up a US$75 million defibrillator to jolt industry into making it happen. The moolah comes by way of an expansion of the blue-sky agency's Electronics Resurgence Initiative ( …
When you consider the amount of money the semiconductor industry needs to produce new technologies.. 75 million is peanuts.
Instead of buying a new, unneeded carrier, they should spend the 10400 million dollars each cost doing things that are needed in the us, like research, repairing bridges, etc.
My thoughts exactly. The kind of exponential increases seen in the last decades are obvioulsy not sustainable, and however much improvement is made there's going to be a limit at which the speed of an individual processor tops out, and a physical limit beyond which they cannot get any smaller.
keep in mind what the moon program did for tech, and there's no harm in buying something that needs to be small, fast, and multi-core.
But the problem with Moore's Law isn't the width of the bus, the number of cores, or even the Ghz clock speed. It's the general LACK of decent software that takes advantage of multi-core SMP design.
And Micro-shaft's move towards a "phone-like" "App" instead of multiple applications sharing multiple desktops and simultaneously doing things YOU want done [not slurping nor advertising nor sharing your personal data over 'teh intarwebs' "in the background"], definitely is NOT moving in the right direction.
True multi-thread SMP solutions would make the UI run a lot faster, or at least be PERCEIVED as faster. The obvious things like MPEG decoding (and encoding) are already being addressed, as well as high performance gaming. But as long as Micro-shaft (and others) continue to excrete BLOATWARE instead of software (and ".Not" and UWP are a big part of THAT problem) you're not going to see that technological push that makes people believe that "the new computer" is actually BETTER than what they already own...
Contrary to popular belief defibrillators don't start things, they are designed to stop the heart dead when it is fibrillating - that is, when the AV node is firing randomly and the muscle is contracting and relaxing spasmodically.
So to use defibrillate as a metaphor for starting something is incorrect. As a metaphor for giving something / someone a massive shock it is probably admissible.
it's getting harder and harder to keep Moore's Law tick-tocking over
Is it really? Go back to when Gordon Moore first stated his observation, at that time the main driver was shrinking the silicon. We take that for granted, but it wasn't a walk in the park, it was cutting edge research and cutting edge manufacturing by some of the cleverest people on the planet. Just because we're now approaching the limits to shrinking silicon, there's plenty of mileage in other areas of research.
It's ALWAYS been hard to keep aligned with Moore's Law, I'm not convinced it is getting either harder or easier. And I'm with the commentard above who observed that $75m is gnats piss. In 2016, the top 10 semiconductor companies spent over $35 billion on R&D (and there were companies outside the top 10 individually spending over $1.5bn on R&D).
Cause that's what makes the world go round.
Mines the one with the old "Cabaret" CD in the pocket.
But seriously. Here's the thing.
The current road map says 2017 is the 10nm node
10nm is 48 Silicon atoms. And that's proving to be b**ls achingly difficult to produce, and historically the oxide layer is 1/10 the line width. Currently it takes essentially a re-mapping and re-factoring of the minimum geometry into easier to image geometry that the exposure tools can deal with, multiplying the number of masks needed.