Classic Computer Magazine Archive ANTIC VOL. 4, NO. 6 / OCTOBER 1985


Mac eats the ST's dust

by PATRICK BASS, Antic ST Program Editor

Perhaps it's in our blood...
   Seems as if people just need to rate and compare things to see which is the fastest or most powerful. This certainly holds true when it comes to rating different computers against each other.
   One standard that's used for comparing computer performances is speed of program execution. The same program, in the same language, is run on different computers.
   Since 1981, Byte Magazine's "Prime Number Benchmark" has been the most widely accepted test of computer speed. This benchmark program is a simple procedure for finding all the prime numbers betwen 3 and 16,381. It adapts the Sieve of Eratosthenes, which has been around since the third century B.C.

How does the Atari 520ST stack up?

Figure 1
Computer Operating System Language Run Time (Sec.)
68000 Atari 520ST TOS C Digital CP/M 68K 3.8
68000 Apple Macintosh - C Manx 7
68000 Apple Macintosh - C Hippo L2 13
Z8001 5.5 MHZ Unix C 1.97
Z8000 Z-Lab Zeus Unix C 4.8
Z80 CP/M Digital Basic 15.7
Z80 CP/M MicroSoft COBOL 5115
6502 Atari 800 OS Rev.B ACTION!  display off:
                  display on:
6502 Atari 800 OS Rev.B BASIC 389
6502 Atari 800 OS Rev.B BASIC XL 214

Antic thanks Craig LaGrow of Computer Language magazine for providing comparative C run times, and Bill Wilkinson of Optimized Software Systems for the Atari 8-bit language run times.

   We typed the "Prime Number Benchmark" into our 520ST using C language and timed how long it took to run.
   The 520ST turned in a time of 3.8 seconds. This speed puts it right up there with minicomputers running the UNIX operating system! (See Figure 1.) A Z8000 Z-Lab Zeus UNIX minicomputer running C took 4.8 seconds, and a Z8001 5.5 MHz mini running C on UNIX took 1.97 seconds.
   At the low end of the scale, microcomputers running the benchmark took from 15.7 seconds (Digital BASIC on a Z80 microprocessor) to an astounding 5115 seconds (1 hour, 25 minutes) on a CP/M Z80 running COBOL.

The Macintosh is the closest relative of the Atari 520ST because both machines use the 68000 microprocessor. Fastest Macintosh time for running the Sieve program in C was 7 seconds and the slowest was 13 seconds. The 520ST left Mac in the dust!
   Note that this test does not depend on any I/O devices. Some computers might run a program blindingly fast, and then take the rest of the afternoon to write the results to disk. Other computers might take longer with the same program but write to disk quicker, thus completing the entire task in less time.
   Suppose we had selected a benchmark test that included sorting a number of disk files. The speed of the disk drive hardware and software could be as important as the actual speed of the computer. Printers also tend to slow computers down a lot. An Atari 520ST can execute more than 10 million instructions in the time it takes a printer to perform a carriage return!
   We can also speed up the way a benchmark program performs its job. One common technique for doing this to take advantage of special hardware features found on your particular computer. A familiar example for Atari 8-bit computer users would be to turn off the screen and speed up the program by 30 percent.

Just how fast is fast, anyway? Sometimes it seems funny to praise one computer for performing 8 million operations per second and then scorn another computer that merely performs 2 million operations per second. How long would 2 million operations take you with a pencil and paper?