Up to now, “turn on” and “boot up” have been synonomous terms, but that may not be for much longer.
In an article about carbon nanotube memory, the folks developing it makes two promises:
1) You’ll see some sometime next year, at roughly the same price at current RAM, and
2) You’ll have “a machine that boots up instantly.”
For the first point, well, I’ll believe it when I see it.
For the second, well, that’s where the difference between “booting up” and “turning on” becomes a real difference.
You cannot “boot up” instantly, by definition. “Booting up” means setting up the CPU to work through an operating system from scratch. For the foreseeable future, that will take some time, simply because any CPU we’ll see anytime soon can’t do everything at once. It’s like getting dressed in the morning, first you put on this, then that. You can’t put everything on at once.
“Turning on” is something quite different. The closest equivalent we have today is “deep sleep” mode, where the OS/applications are already set up in memory, and the only action taken is to wake the CPU up.
The key to an “instant on” machine is to take “deep sleep” mode the final step, and wake the CPU up not from a coma, but from being dead/off. Can that be done with current CPUs? Don’t know, maybe they can, maybe they can’t, though even if they can’t, it probably wouldn’t take too much to fix that (provided “instant” gets refined as “a second or two or three.” Once you do that, though, you can play that trick with any “live” memory: carbon nanotubes, iRAM-like devices, even flash memory (though the last would be a lot slower and a non-starter).
This is like sleeping with your clothes on, once you wake up and get out of bed, you’re ready for action.
This would change how an OS operates. For instance, you would have to have two kinds of shutdowns: one in which the system isn’t flushed, just put in suspense, and another where it is.
The first would let you resume work where you left off, which most people would rather like. However, you have to retain a true shutdown and restart simply because computers malfunction. If a computer crashes, restoring the crash obviously does you no good, just like sleeping in a shirt you ripped in bed the other night doesn’t mend it.
What if you change a part (or one goes bad during a “turn on.”)? Depending on the part, and how it operates, the image sitting in RAM may or may not be 100% valid any longer, and you’ll find out which it is when you first use that part.
None of these points are a stopper, and fast non-volatile memory would be an improvement in any “instant on” system over fast voliatile memory dependent on one form of life support or another. It just wouldn’t work that way 100% of the time.
However, if carbon nanotube memory were around today, it would face most of the same problems and bottlenecks any solid state RAM faces today. This is yet another new technology that needs new support to function, and it won’t quite turn your computer into a TV.