Comparision of the speed of computers

I have been measuring for several years the speed of various types of computers ranging from i8086 based PC's to vector super computers by running several simple fortran programs.

Click here to see the table of the results of my bench mark tests.

The source codes of the bench mark tests:

• legval.f multiplies a 1024*1024 matrix of real*8 to a 1024-dimensional vector of real*8 many times.
• intosc.f repeats arithmetic operations to integer*4 variables. Since these operations cannot be parallelized efficiently, this test is expected to reflect the CPU clock.
• osc.f replaces the integers in intosc.f with real*8. Integer divisions are replaced with real number multiplications.
• intpi.f contains a IF block in a DO loop. Only integer*4 variables are used.
• pi.f replaces interger*4 in intpi.f with real*8.
• pemute.f for 4KB chooses randomly two elements from an integer*4 array of 4KB and swap the contents of them.
• permute.f for 128KB uses a 128KB array instead of 4KB.
• permute.f for 4MB uses a 4MB array instead of 4KB.

Back to: fortran benchmark tests (ver.2)