Monday was the first time in a week that I had any time to read, and so I read Skin Game by Jim Butcher, which is the latest entry in his Dresden Files series...and I ended up finishing the thing in one sitting, just about. I'd read some of it Sunday evening--a few chapters--but otherwise I read the whole thing at one go. As usual.
It's a fine adventure yarn, just like every other book in the series is. I really, really enjoyed this book...again, just like every other book in the series.
Today I have a few errands to run, and because I stayed up all night reading I guess I'm going to go run them and go to bed afterwards. Well, a new Dresden book comes but once per year (more or less) and I have to geek out when it does. Whee!
* * *
JayG writes about his Pentium laptop running Win 95. I recall musing about current technology letting us put an entire Pentium-based computer on the faces of a grain of salt.
...which makes me wonder: how many C-64s can dance on the head of a pin? The present state-of-the-art in nanolithography probably would let one put dozens of complete C-64 computers in that much real estate. (Power and I/O are still left as an exercise for the student, of course.) I mean, the C-64 was designed in the early 1980s, back when they could still design integrated circuits using colored tape and large sheets of acetate. The processor for the C-64 (the 6510a, which was a variant of the venerable 6502) clocked in somewhere around 5,000 or so transistors. Compare that to the processor in Cephiro--new in 2007--which had something like 200,000,000 transistors.
It's hard to get a definitive answer on how many transistors there were in a 6510, so I am estimating based on the position of the 6502 in a chart of Moore generations, but even if I'm off by a factor of two and it was 10,000 transistors, that's still about 20,000 6502 processors on the die occupied by a single Core 2 Duo.
Given modern manufacturing techniques, an entire C-64 on a single IC is possible. Packaging and I/O are still a problem, though not a big one since we have techniques which allow us to make ICs with huge number of connecting pins. You'd have to have all sorts of external buffer ICs to interface the damned thing with the outside world, but the cost of those chips are equally as trivial as the cost of manufacturing something like a C-64 on a chip in bulk. (You probably would not need state-of-the-art techniques to do it; technology about five or ten years old would be more than sufficient to the task.)
...except, of course, there is virtually no market for 8-bit computer hardware, single IC or not. Certainly there's not a mass market for such things.
When I get to thinking about it, though, I am struck with just how much computer technology has advanced since I first started noodling around on the old C-64. Because it's more convenient to build computers as assemblies of generic ICs, it's really not all that obvious just how much smaller the feature size is in 2014 than it was in 1984 (or 1994!) until you start thinking about putting whole Pentium computers on grains of salt, and dozens of C-64s (or Apple ][s if that's your bag) on the heads of pins. But that's what modern nanolithography can do--processor, memory, I/O, everything, even if you can't really connect them to anything. (And I wouldn't make bets on that.)
This is why the IC manufacturers are experimenting with multicore CPUs with staggering numbers of cores. 256 cores will be typical for entry-level computers much sooner than we think; even considering how the performance curve has flattened in the past half-decade I'd be surprised if it took much longer than another ten years for that to become commonplace.
* * *
I have just about gotten to the point that I can no longer think coherently about complicated subjects. I think what I'm going to do is go run my errands, and then go to bed.