What is the future of Computer Architecture
With the diminishing returns of Moore's Law, the traditional approach to Computer Architecture seems to have stop giving any meaningful returns. Do the new branches like hardware accelerators, approximate computing and neuromorphic computing have any scope? Or do we live with the diminishing gains associated with the traditional approach to Computer Architecture.
(
3 comments
| 3 points
| on HN
)
If you didn't have to worry about clock signals, had a simple circuit topology (for example laminar) and thermal issues were minimal because of having a small number of gates transitioning at any one time could you design a programmable logic chip that was more competitive with ASICS than current FPGAs are ?
However, even these will be one time gains. As Bob Colwell says, it's very difficult to replace an exponential.
This will have profound effects not just on Silicon Valley, but the global economy. We've been riding this miracle for 40 years and it will end - soon.