Comment Re:If one was produced with a 40nm process... (Score 1) 124
And yet, you can't propagate a signal all the way across the chip in one clock cycle.
And yet, you can't propagate a signal all the way across the chip in one clock cycle.
That is way oversimplifying what is needed to make a competitive chip. If it was that easy, there would be a lot of people doing it, giving Intel a lot more competition than they have. And it wouldn't take ~2-3 years per generation.
In order to get high performance (== high frequency, and == reasonable die size), you cannot rely completely on automated tools.
... perhaps the problem is that 10 hours and reboot is a ridiculous requirement from the get go.
Or, it could have been a lack of requirement. If there was no spec about the tolerable down time (5 nines? 4 nines? less?), it simply would not have been designed in. The "reboot every 10 hours" could be a workaround added after the design was completed.
Sorry, that's idealistic crap. Engineering is all about making tradeoffs based on the intended usage. Over-engineering costs time and money. Most of the time, there is no way to design for every IMAGINABLE use case.
For example, there is no way to build a completely earthquake-proof building. You might be able to design one (if you knew for certain that there is an upper bound for the magnitude of an earthquake), but no one would have enough money to build it.
By taking your comment to the extreme, one should be able to use pebbles fired from slingshots to shoot down ballistic missiles. Why hasn't anyone IMAGINED doing that?
"Experience has proved that some people indeed know everything." -- Russell Baker