A lot has happened in the 44 years since I wrote my first program on a PDP-8...
To illustrate the development: In 1989 I wrote a C program which computes a perfect minimal hash table given a fixed set of keywords. (Based on an article by Richard J. Cichelli in CACM January 1980.) The interesting point with this is that I still have the program, and I have run the same test suite on most machines that have come in my way, starting with a Vax 750 and a DEC 2060 (which were old, but still around when I wrote the program). The test case which took the longest to run still takes measurable time. A few examples of running times (minutes:seconds) over the years:
Vax 11/750 | Vax | 32:12.6 |
DECSYSTEM 2060 | PDP-10 | 16:20.1 |
Sun 3/160 | 68020 | 11:22.2 |
Apollo DN3500 | 68030 | 6:27.9 |
HP 9000/820 | HPPA | 3:33.3 |
SPARCstation 2 | sparc | 1:46.4 |
SPARCsystem 630/M41 | sparc | 50.2 |
SPARCstation 5 | sparc | 25.6 |
SGI Indigo | MIPS | 12.9 |
PC | Pentium II 450MHz | 4.3 |
iBook | PPC G3 700MHz | 1.8 |
MacBook Pro | Intel Core 2 Duo 2.2GHz | 0.5 |
MacBook Pro | Intel Core i7, 2.5GHz | 0.2 |
Note that this is a CPU bound program which uses very little memory and does no disk I/O. The timings do not necessarily show the different systems' overall performance. For example, the DECSYSTEM 2060 could handle 50-100 users, or more, a magnitude more than the faster (in this benchmark) Sun 3/160 could handle.
If I try to estimate the year of different machines (not exactly the ones above - the full list is about 30 machines) and plot a graph, you get something like this:
From half an hour to half a second in about 25 years, not bad. :)