Super Pi Under Linux

I am curious to know if Super PI values done under WINE are meaningless. I got some awesome values here under Ubuntu 7.04. I know that Super Pi is kind of a meaningless benchmark, but still it's being used in reviews.

I have a 2.0ghz Pentium M processor.

I'd like non-Linux people to comment as well. Thanks.

 

No, they aren't meaningless. They still calculate the same things. Wine still lets the application run on the same processor without emulation, it just provides libraries that allow the application to think it's talking to the Windows graphical API. They're completely valid numbers, as SuperPI calculates the same thing under each OS. It's a valid comparison of how much the OS "gets in the way" of running applications. I'd be curious to see the SuperPi results for your same hardware under Windows.

 

Thanks, Pita. I knew you would know about the technical details of this.

 

Just ran it on my M90. I got 1:12 for 2 million digits, which is exactly the same as I got on Windows as noted in my initial Precision M90 review. This could either mean that there is no difference between operating systems, or the fact that it is still running on a Windows API is causing it to be the same speed... after all, as Pita said, it thinks it's running in Windows.

 

In fact you can run native SuperPI in linux, why bother use wine?

 

Good to know. Thanks for looking into this, ftw!

I didn't see a Linux tarball or .deb package on the SuperPI site...can you give me a link to this?

Thanks.

 

ftp://pi.super-computing.org/Linux

no idea if it gives you results that you can actually compare to the one for windows.

actually, just poking around a bit, I think this might be the same program.

final edit:

ok, the program is put out by the same people...

http://forum.notebookreview.com/showthread.php?t=41097

still don't know if the results are truly compatible, as apparently the program was translated from FORTRAN to C and then optimized. (link is in spanish, scroll down 'til you see an inline frame with text output.)

 

I got ~36 seconds to calculate pi to a little over 1 million digits (./super_pi 20) my CPU is in my sig, how does that compare to the results in Windows?

2097152 digits took 1:26... (./super_pi 21)

well, looking over the super_pi thread here on NBR, this is significantly faster than times listed for windows runs with my CPU/RAM combo. (Pentium-M 740, 1GB 533MHz) I'm guessing that this is due to code optimizations, but if we don't show them this, we can just claim its the inherent superiority of Linux!

 

My results are quite similar. Only <0.5 sec difference

 

Took me 9 minutes to calculate to 2097152 digits.

That's on a 500MHz Intel Celeron by the way.

 

I could grow a beard in that time.

 

Yep, oh yes...

LOL. What was your SuperPI time(s), hotshot?

 

Just ran it again on my M90. Went in 54.803 seconds (./super_pi 21). That's nearly 20 seconds faster than in Windows, and closer to a 2.16 Core 2 Duo in Windows.

 

Boy, I should say there is noticeable difference.

Mine Athlon 3500+(2.2G) use 34.578 sec to run 1M, and 78.769 sec to run 2M. But I am using DDR400, not you guys DDR2 stuff. I think it is faster than Win, but my ubuntu is 64 bit. I think 64 bit may count more weight.

Actually SuperPI is not that accurate all the time. I compared SSE3 optimized SuperPI, 32bit win SuperPI, and 64 bit Linux SuperPI.

Here is the result.

# bit SSE3 64bit Linux 32bit WinXP
1M 32 Sec 34 Sec 43 Sec
2M 1m 14 Sec 1m 18 Sec 1m 40 Sec

So SuperPI is pretty biased to me. All above tests are on same machine.

 

"LOL. What was your SuperPI time(s), hotshot?"

Me? Oh... er ... you don't wanna know!

 

i use vmware to play this game....
the first two is vm only expreme programming

the last two is with 8.04 and extreme programming

but amazing last two is faster than the first


and vmware's time is not right , please see the host mashing, around 30s~~