SuperPI 32M - Windows vs Linux

[tiborrr]

Testbed:
- Intel C2D E6600 @ 465x9 = 4185MHz
- 2x512MB D9-GMH @ 465 4-4-4-12-5-25
- ASUS P5B Deluxe
- WinXP SP2, some finetuning & tweaking done
- Gentoo Linux (2.6.18 kernel) - zero tweaks, for 24/7 usage

All in all, setup wasn't overclocked to the limits and the latencies weren't tightened at all.
Let's check how did the 32M SuperPI test ran on both Windows, where it ran in native mode, and Linux, where it ran in emulated mode via 'Wine' emulator. The results are quite interesting as you can see from the screenshots below:

Windows XP SP2:


Linux 2.6.18 - emulation:


P.S.: I have already repeated the benchmark twice in Linux and got better result (cca. 13min 24.8s). What's interesting is the first and the very last pass, which vary greatly from OS to OS. I'll try to tweak both Wine and SuperPI and try to get the last pass run as fast as any other. I think it's the problem with writing the results, so if i can speed this up the results will be fenomenal

[massman]

Strange ... Linux used to be a lot faster. Can someone test with an older cpu? I'm pretty sure the result will be completely different.

[SLi_dog]

Strange ... Linux used to be a lot faster. Can someone test with an older cpu? I'm pretty sure the result will be completely different.

hmm...that's what I seem to remember too???

Massman, is linux allowed for the hwbot SuperPi competition?

[stealth17]

Try the native linux version.

ftp://pi.super-computing.org/Linux/super_pi.tar.gz

For a 1m run the command would be:

./super_pi 20

Also I've seen gains of 2sec+ when running without X.

[kiwi]

Sorry but running appz through wine is NOT linux

Too bad there is no SP mod client on linux. You can only try what stealth17 suggested

[Ten]

Try the native linux version.

ftp://pi.super-computing.org/Linux/super_pi.tar.gz

Can you upload it on speedyshare or smthing because i think ftp is down atm.

[kemo]

IMO super pi shouldn't be allowed on Linux
1) we all use windows so we need fair comparison
2) it is a lot easier to fool the pi @ Linux

[dinos22]

you should see how fast linux is in 1M

show me how many Windows XP machines can do this with E6600 conroe B1 stepping

i did 12.7xx on last reboot at 3GHz

took a screenshot of this one







32M times are not that great but 1M oh boy

[stealth17]

IMO super pi shouldn't be allowed on Linux
1) we all use windows so we need fair comparison
2) it is a lot easier to fool the pi @ Linux

I disagree with your number first reason. It is no different then people using Server 03 or Windows 2000 just because they are faster in one type of Pi then another.

As for number two, yes, that is a problem. I wish it was like the Super Pi Mod so we didn't have to worry about cheaters.

[stealth17]

Can you upload it on speedyshare or smthing because i think ftp is down atm.

Sure thing

http://www.theoverclocked.com/hosting/custom

[GTeng]

IMO super pi shouldn't be allowed on Linux
1) we all use windows so we need fair comparison
2) it is a lot easier to fool the pi @ Linux

LOL just WOW. I am having a hard time deciding which of these two statements is the dumbest. You can go back to your kindergarten OS now

ANYWAY.... native linux superpi 1M benchies are about 3 seconds faster than Windows XP in my box

[dinos22]

LOL just WOW. I am having a hard time deciding which of these two statements is the dumbest. You can go back to your kindergarten OS now

you comment makes you no different

[Vapor]

LOL just WOW. I am having a hard time deciding which of these two statements is the dumbest. You can go back to your kindergarten OS now

ANYWAY.... native linux superpi 1M benchies are about 3 seconds faster than Windows XP in my box

1) Welcome to XS
2) YGPM
3) both his points are valid....A) the results from Linux are so different that it's essentially not even the same benchmark, thus can't be used for competitive benching. B) it's a lot easier to cheat with [the appearance of] Linux....whether that be via VM or other means.

[dinos22]

1)
2)
3)

[MuffinFlavored]

This is interesting.

I was considering just throwing Linux onto a hard drive and booting into it for this reason.

I think my overclocks are squandered by the fact that alot of my overclocks POST, but they do not boot into Windows.

So, this would be considered 'unstable' in my terms. As, I can not even get into my operating system.

I want to see if I can boot at higher speeds into Linux.

I would expect Linux to be slower because, I get that sluggish feel with the operating system, with my rather, non-sluggish system. And yes, I did have video drivers installed. :P

I used Linux for quite a while, but, for just Firefox and Pidgin, Windows is just faster. In terms of Firefox. Maybe not statically, but definitely in overall feel and general page loading. And a lot of plugins aren't out for Linux in terms of Firefox.

EDIT: I would just put my voltage at 1.5v, boot into Windows at stock, and then overclock at operating-system level, but nTune returns a 1608 installation error (which I googled and could not solve), ClockGen increase my RAM with FSB, and my RAM is so terrible it is funny. With my FSB at 487, I just leave my RAM at 900 5-5-5-15. And SetFSB doesn't support the nForce chipsets. If there is a way to un-link the RAM from the FSB in terms of increasing in ClockGen with the 650i SLI chipset, please inform me.

[dinos22]

the trouble is that everyone is benching Spi XS mod 1.5 on windows and it makes for easy comparison

once you start benching uncommon OSes it's hard to know how much is hardware and how much software related and most people will not bother to figure out linux just to bench one app

[nfm]

This is just silly comparison, you are emulating dude. We can take some sort of pi algorithm (other than superpi and one that is open), compile it with gcc under linux and mingw under windows with same cpu flags and optimizations, pair it with properly configured kernel and there you go.

[kemo]

LOL just WOW. I am having a hard time deciding which of these two statements is the dumbest. You can go back to your kindergarten OS now

ANYWAY.... native linux superpi 1M benchies are about 3 seconds faster than Windows XP in my box

Thank you

[massman]

hmm...that's what I seem to remember too???

Massman, is linux allowed for the hwbot SuperPi competition?

No, it wouldn't be fair at all

[GTeng]

1) Welcome to XS
2) YGPM
3) both his points are valid....A) the results from Linux are so different that it's essentially not even the same benchmark, thus can't be used for competitive benching. B) it's a lot easier to cheat with [the appearance of] Linux....whether that be via VM or other means.

What you call cheating I call more efficient computing just because you can't compare it to a Winblows benchmark doesn't mean any of his point make any f-ing sense. For us Linux users who actually give a damn about number crunching performance for work/research this does mean a lot. In fact at the same clock speed I get about 20% improved performance in Linux over all in number crunching.

[GTeng]

you comment makes you no different

You got a little brown something in your nose, pull your head out of it.....

[Vapor]

What you call cheating I call more efficient computing just because you can't compare it to a Winblows benchmark doesn't mean any of his point make any f-ing sense. For us Linux users who actually give a damn about number crunching performance for work/research this does mean a lot. In fact at the same clock speed I get about 20% improved performance in Linux over all in number crunching.

Yes, Linux is faster in SuperPi....that's not the cheating part. We're just saying that they're incomparable.

The cheating comment about the Linux SuperPi is that it's just EASIER to cheat in the Linux version than it is in the Windows version....just another reason why we don't acknowledge those scores as official.

If people want to have a Linux SuperPi competition...by all means go ahead, it'd probably be fun with the challenges in in-OS OCing that Linux presents. BUT, under no condition will those results be considered for the overall SuperPi WR.

You got a little brown something in your nose, pull your head out of it.....

Ugh, YGPM, again

[kemo]

You got a little brown something in your nose, pull your head out of it.....

Still it makes no difference

[afireinside]

What you call cheating I call more efficient computing just because you can't compare it to a Winblows benchmark doesn't mean any of his point make any f-ing sense. For us Linux users who actually give a damn about number crunching performance for work/research this does mean a lot. In fact at the same clock speed I get about 20% improved performance in Linux over all in number crunching.

We're not doing "efficient computing" here, it's a competition. We all know windows. It's been the standard used since this hobby began. If linux is faster for your work that's great, but this isn't a job. We're far to late into the game to learn a new OS and have every bench recoded. The only person here who's not making "any f-ing sense" is you

Think about it, we have enough problems with cheating accusations as it is on an OS we already understand inside and out. Imagine what would happen when we use an open source OS with TONS of different distros and tons of tweaks we don't know about. How would you tell a highly tweaked version of a fast distro from a flat out cheat?

[MarlboroMan]

Why linux is easier to cheat?? Nobody gives an example. No concrete evidence on it.


-if you compare superpi (the windows binary) running on linux and on windows , that is total fair! You still have the checksum to validate it.

-Linux is easier to tweak not to cheat. Like optimizing your kernel to run on your processor only.

- If someone there knows about programing a lil bit, that person would know that to hack a compiled binary without having a header or the source is quite difficult on windows. On linux is MORE difficult, because you are "emulating" windows to run this app, so you have to hack the "emulator" first than start to hack the app. 2x more difficult

-You have 10x more hacking tutorial on windows than on linux.

-If linux is easier to cheat, so is strange the fact that linux don't have a single spyware, adware, trojan or "bad boy virus". This stuff are all HACKS, CHEATS. Windows have tons of it. Your anti-virus download patch every single day, why?

-If linux can run the same binary faster than windows, its not linux fault or cheat. Its windows that is bloatware and can't even run well even an app that is compiled to run natively on windows.


So if you are claiming that linux is easier to cheat, first show me something concrete.