When BYTE published its first IBM-specific special issue (early '80s sometime), I tried another benchmark in the article on the '286/287. This was a benchmark where the value 1.0 (a float) was added to every element of a FORTRAN array of floats. On my Stride 440 (10MHZ 68000, *no FPU* -- that is *software* floating point), the runtime for a C coded version was just as fast as a stock IBM-PC AT w/80278 (I think 8MHZ). The 80287 is probably great IF you don't have to fetch or store floating point numbers
in memory.
More recently, BYTE did some benchmarks of MACs and 80386 boxes (late '80s) and found that for some things the '386s were faster and for some things ther MACs (68020s and 68030s) were faster. Overall they were about the same (more or less). The question is not which is "faster", but what it is you are trying to do, exactly, and which processor features will help you do what you want to do better. The m68k chips have more registers and don't mess you up with a segmented address space. The ix86 chips a