Perhaps if the writer understood what Moore's law was actually about.
Supercomputers in 2030: Lots of exaflops and LOTS of DRAM
Supercomputers will overcome the predicted crumbling of Moore’s Law over the next few years to show massive leaps in performance between now and 2030 – but will still look pretty familiar to today’s compute power junkies. In a series of sessions at ISC this morning covering exascale architectures, speakers predicted a rapid …
COMMENTS
-
-
-
This post has been deleted by its author
-
-
-
Wednesday 22nd June 2016 06:56 GMT John Smith 19
What's the current estimate of the computing power of the human brain?
Maybe 10 Pefaflops?
In a 2Kg package dissipating about 400W?
The end point for Moore Observation is a 1 atom wide gate with a 1 atom thick insulator on top.
And that's it.
You want more processing power per unit area? Start stacking wafers.
Not impressed.
-
Wednesday 22nd June 2016 08:18 GMT Nigel 11
Re: What's the current estimate of the computing power of the human brain?
The end point for Moore Observation is a 1 atom wide gate with a 1 atom thick insulator on top.
Actually it's several atoms thick. Below that quantum effects come into play and your insulator becomes intolerably leaky. Alternatively, you might be able to work with the quantum effects, but the result will not be a field-effect transistor as we know it, nor (AFAIK) is anything of this nature working at the VLSI level in R&D labs.
-
-
Wednesday 22nd June 2016 09:01 GMT Charlie Clark
Your crystal ball is cloudy
Both Intel and Micron are heavily invested in a DRAM world and it's clouding their ideas.
As systems get larger the power draw directly attributable to DRAM gets more important and should encourage research into less power hungry alternatives. It's not my field but I can imagine that people might start mulling over alternatives like a 2 ExaFlops DRAM system drawing 100 MW versus a 1.5 ExaFlops xRAM system drawing 10 MW and opting for the slower but cooler and cheaper to run system. MRAM looks good but Flash is also getting faster and faster.
Time to throw some dollars at that kind of research rather than yet another Silicon Valley service of minimal utility.