Thursday, July 28, 2016

End of Moore's Law

Richard Chirgwin at The Register reports that the Semiconductor Industry Association has issued their roadmap for chip technology, the ITRS:
The group suggests that the industry is approaching a point where economics, rather than physics, becomes the Moore's Law roadblock. The further below 10 nanometres transistors go, the harder it is to make them economically. That will put a post-2020 premium on stacking transistors in three dimensions without gathering too much heat for them to survive.
This is about logic, such as CPUs, but it is related to the issues that have forced flash memories to use 3D.

Energy demand of computing
There are other problems than the difficulty of making transistors smaller:
The biggest is electricity. The world's computing infrastructure already uses a significant slice of the world's power, and the ITRS says the current trajectory is self-limiting: by 2040, ... computing will need more electricity than the world can produce.
So we're looking at limits both on the affordability of the amounts of data that can be stored and the computations that can be performed on it.

The ITRS points to the wide range of different applications that the computations will be needed for, and the resulting:
research areas a confab of industry, government and academia see as critical: cyber-physical systems; intelligent storage; realtime communication; multi-level and scalable security; manufacturing; “insight” computing; and the Internet of Things.
We can see the end of the era of data and computation abundance. Dealing with an era of constrained resources will be very different.In particular, enthusiasm for blockchain technology as A Solution To Everything will need to be tempered by its voracious demand for energy. An estimate of the 2020 energy demand of the bitcoin blockchain alone ranges from optimistically the output of a major power station to pessimistically the output of Denmark. Deploying technologies that, like blockchains, deliberately waste vast amounts of computation will no longer be economically feasible.

No comments: