Would you linux guys bother to retest SuperPi?

I find a very interesting OLD thread from here:
http://www.ocforums.com/showthread.php?t=405925

It basically provide you source code for SuperPi and let you test it under Linux. Their result is really surprise me. 1M number on 64bit machine is about 25.58 sec. Of course, the score is under overclock(2.7GHz) machine. But I do believe Linux takes the advantage of 64 bit computation.

It is a bit of unfair of comparing 64 bit AMD under 32 bit OS vs. 32 bit Intel under 32 bit OS.

Would you guys take effort to re-test it? The result maybe surprise you. or maybe not.

Here is the so called AMD64 patched SuperPi
http://xtremesystems.org/forums/attachment.php?attachmentid=23762
I think it is nothing new except using SSE2
and
http://www.overclock.net/attachment...uper-pi-thread-super-pi-sse3.zip?d=1118123380
Which is optimized for SSE3.

But all these modified versions are Windows based. can't find more on linux version.

 

Other forums show Core Duo will perform worse(4s slower at 1M number) if you load two identical SuperPi task into two cores(intel). Only one post show AMD X2 will perform the same whether if you load single task into two core or load two tasks into two cores.