In the case of Mac applications, RB2006R1 is only slightly slower
than RB5.5.5 as shown in the following benchmarks run on a G5,2.7GHz.
Please note that the numbers should be interpreted as operations per
time. They have no absolute meaning except that they scale linearly
with throughput (higher is faster).
Data for G5,2.7GHz.
benchmark speed 5.5.5 speed 2006R1
Math (long integer) , 1354.5 1197.0
Math (double precision) , 1395.8 1197.0
Math (transcendental) , 1898.8 1559.6
Math (scientific mix) , 1351.3 1467.0
Math (composite score) , 1500.1 1355.3
String Manipulation , 1010.6 1017.8
Graphics (mixed) , 670.5 953.0
Disk Read/Write , 321.2 344.2
Composite Benchmark , 1126.7 1094.2
Thus, 2006R1 is slower for most math (particularly transcendentals)
and faster for graphics. The good news regarding math is that thanks
to Mars, two-dimensional array elements are accessed much faster in
2005/2006 so scientific math routines, which use lots of arrays, run
faster.
Unfortunately, the performance of 2005/2006 is significantly slower
for math on Pentium 4s but much faster for graphics (which are
slower on the Mac because of buffering). I found no significant
differences for 2005 versus 2006 so the results are averaged. I have
no clue why math has taken such a hit on PCs. Here is the data from
out computer lab. (Use a non-proportional font to display the
following table):
RB version 5.5.5 2005/2006 5.5.5 2005/2006
System G5(2GHz) G5(2GHz) P4(3.6GHz) P4(3.6GHz)
Math (long integer) 1003.4 886.5 1145.9 1011.1
Math (double precision) 1022.8 884.2 1033.0 449.4
Math (transcendental) 1419.8 1173.0 1386.1 792.1
Math (scientific mix) 1004.9 1084.9 1131.2 800.6
Math (composite score) 1112.7 1007.2 1174.1 763.3
String Manipulation 745.5 756.3 896.5 884.0
Graphics (mixed) 501.2 761.2 1482.0 2119.7
Disk Read/Write 481.9 469.7 785.7 807.4
Composite Benchmark 867.2 846.6 1096.8 970.7
Finally, I note that all of my applications are larger in 2005/2006
than in 5.5.5, but the size difference is highly variable. This is a
subject worth further discussion.
If anyone would like the benchmark program, please email me.
Cheers,
Bob Birge
--
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
