Why do we use SuperPI as a benchmark?

C

cyclone3d

Fully [H]
2FA
Joined
Aug 16, 2004
Messages
18,454
So I just downloaded and ran PiFast on my system.

For 32M digits of PI is took just over 2 minutes total.

PiFast is an OLD program (as is SuperPI)...

I wish the sourcecode was available.... grrr.

Anyhow, on my system, the results are as follows:

Total Physical Memory Allocated = 229050 Kbytes
Computing series with index N=2256707
Starting step 5
Starting step 4 ...
Starting step 3 ...
Starting step 2 ...
Starting step 1 ...
Starting step 0 ...
Series computing time : 102.77
Division time : 10.16
InvSqrt time : 4.83
Final huge multiplication time : 3.63
---------------------------------------------------------------
Total computation time : 121.69 seconds(~ 0.03 hours)
 
Total Physical Memory Allocated = 458034 Kbytes
Computing series with index N=4513407
Starting step 5
Starting step 4 ...
Starting step 3 ...
Starting step 2 ...
Starting step 1 ...
Starting step 0 ...
Series computing time : 235.19
Division time : 18.98
InvSqrt time : 14.63
Final huge multiplication time : 8.22
---------------------------------------------------------------
Total computation time : 277.75 seconds(~ 0.08 hours)
 
A64tweaker

http://files.extremeoverclocking.com/file.php?f=106

After tweaking memory timings.


32M
Total Physical Memory Allocated = 229050 Kbytes
Computing series with index N=2256707
Starting step 5
Starting step 4 ...
Starting step 3 ...
Starting step 2 ...
Starting step 1 ...
Starting step 0 ...
Series computing time : 102.42
Division time : 8.41
InvSqrt time : 4.80
Final huge multiplication time : 3.63
---------------------------------------------------------------
Total computation time : 119.59 seconds(~ 0.03 hours)

64M
Total Physical Memory Allocated = 458034 Kbytes
Computing series with index N=4513407
Starting step 5
Starting step 4 ...
Starting step 3 ...
Starting step 2 ...
Starting step 1 ...
Starting step 0 ...
Series computing time : 232.81
Division time : 22.62
InvSqrt time : 10.67
Final huge multiplication time : 8.13
---------------------------------------------------------------
Total computation time : 274.89 seconds(~ 0.08 hours)

Also.. I gained about 300Mb/sec memory throughput in Sandra. :)

Up to 7.38Gb/sec in the Sandra memory throughput benchmark :)
 
A new alternative that replaces a perfectly viable solution, without any real benefit, gets you nowhere. Sorta like Vista :)
 
One thing to note... a couple of the settings don't save correctly in the save file.

Before you do a "load on boot" for the settings... go back and load the startup save file and edit the settings that didn't save correctly.. or else you puter will lock up when loading windows.

As for SuperPI being a viable solution vs PiFast.... PiFast is a LOT faster than SuperPI.

It shows up changes more easily than SuperPI does.... and will be affected a lot more by increased memory bandwidth.

If it comes down to it... and we actually want a "viable" benchmark based on the calculation of PI.. we need a whole new program....

If Mathmatica was free... like that will ever happen.. we could just use that.

A program that is coded for older processors that doesn't actually test very much of what the currecnt processors can do is pretty much useless for benchmarking.

That makes me want to make a benchmark program that tests the seperate types of calculations instead of just trying to rely on super synthetic benchmarks that inly give an overview.. and a poor one at that.
 
That makes me want to make a benchmark program that tests the seperate types of calculations instead of just trying to rely on super synthetic benchmarks that inly give an overview.. and a poor one at that.
Click to expand...

Please do. I'm serious.

But you can't really disown a time tested classic. The golden standard of overclocking CPU benchmarks is superpi, regardless of the reason why.
 
You must log in or register to reply here.
Back
Top