I wouldn't say this if it were BS. So Please look here: http://www.gpugrid.net/hosts_user.php?userid=54651

ID: 61666 Avg. credit 1,580.03
ID: 62706 Avg. credit 3,610.32

They are identical. ID: 61666 ran Win7 64bit Ultimate the 8800GT was OC & the Average already maxed out.

ID: 62706 runs Mint Linux 64bit the GPU isn't OC'ed & the Average hasn't yet maxed out.

Sabayon (Easy Alternative to Gentoo) promises even better performance.

I'm just saying...

If I can be so naughty as to promote the Team I'm on (as a member NOT Captain), please do join "Team Gentoo Linux Users Everywhere". BTW I guess I must be a point whore! ;-)
____________

Just an added note. ID: 61666 was working 24/7 from 07/01 to 22/01 & reached a Total credit of 72,236 in that period. ID: 62706 has been running from 22/01 & as of 02/02 has reached a Total credit of 60,450. So after 15 days it should be up to over 90,000.

The World Wide Grid was running on ID: 61666 but has not been able to run on ID: 62706 for some reason. But ID: 62706 is running ibercivs ID: 61666 which ran from 17/01 to 22/01, & both ran rosetta@home. But still, not running The World Wide Grid on ID: 62706 & running ibercivis for 5 days on ID: 61666, can't account for a 25% on ID: 62706 improvement that wasn't OC'ed like ID: 61666 was by Core Clock 700MHz (vs. 600MHz standard) Shader Clock 1728MHz (vs. 1500MHz standard) & 2000MHz (vs. 1800MHz standard), can it???
____________

As of 07/02 ID: 62706 had an Average Credit of 4,523.30 & a Total Credit of 91,779. After 15 days of 24/7 running BOINC Manager on Mint Linux 64bit VS a Total Credit of 72,236 on ID: 61666 running BOINC Manager on Windows 7 64bit also having run 24/7 for 15 days on the exact same PC that had the GPU OC'ed. That's roughly a 27% improvement on a GenuineIntel
Intel(R) Pentium(R) Dual CPU E2180 @ 2.00GHz [Family 6 Model 15 Stepping 13]
(2 processors) with a NVIDIA GeForce 8800 GT (255MB) Just by using a different OS.




____________

(Thanks for your PM)

Nice shot liveonc,
But did you use this computer during your test ? Or was it just a crunchbox ?
Did you have the same Wu ?
I think the best way to run such a comparative will be to run XX days on Linux, XX days on windows, without touching the computer, then take one type of WU, average them, and compare it with the other OS. (And without CPU Projet)
No ?

I know some people who are trying to optimize the Linux kernel for computing purposes, but I don't know if there is a significant improvement.

On some projects Windows 64b is quicker than Linux 64b (I guess it links to how they compile the app)

In my case, I got a crunch box without a proper GPU to crunch on, and as I'm using this one only to host my website and others stuff, I didn't install the graphic interface. (I have a significant gain on this machine).

Some ppl did a trick to crunch on GPUGrid without launching the graphic interface. (Thanks to some nodes). But it is not very convenient to do so.

Thanks liveonc for the test

It was used as a crunchbox, the only times I used it, was to check up on BOINIC & if nothing bad was happening fx PC hang or crash, etc. It did do the occasional browsing, but nothing else. As I said, it was only an estimate. I already explained that it was not a perfect comparison, only an estimate. Precise comparisons are useless, to my opinion, & would require that I had exactly the same WU's to reuse. So it's not even possible. Anyway, even if it was. What use would a precise test be for anyone who doesn't plan to have the exact same hardware?
____________

Yeah sure !
[Partial Joke ON] And why do we need to prove that GNU/Linux is the best ? [Partial Joke OFF]

But it's still a good comparison. Thanks.

PS : BOINC ! (not BOINIC ^^)

Sadly the times Wii live in "requires" more than "pretty words & white lies".

Just because I read, or heard, or saw. That something was better, unless there is more than just the claim, I can't really trust that it's for real, or just part of the BS mountain that is piling up faster than the speed of light.

Wii have so much BS, that Wii don't even need oil any more! All Wii need is a car that can run on BS & Wii can use that instead!
____________

Some ppl did a trick to crunch on GPUGrid without launching the graphic interface. (Thanks to some nodes).

Actually, one member of my team ([AF>Linux] Elgrande) did it. And he wrote a "how to":
http://linux.boinc-af.org/2009072676/Linux/Specifique-Distribution/Gpugrid-en-ligne-de-commande-Gentoo.html

Unfortunatly it is in french :p

The point is, Boinc has to recognize your cuda device without launching Xorg.
Then you have to create some nodes. (You can modify and use his script to do so)
But this seems to work only with GTX2xx cards.

And before shutting down your computer, you have to add a delay of 5s in order to avoid GPUGrid error.

Actually, one member of my team ([AF>Linux] Elgrande) did it. And he wrote a "how to":
http://linux.boinc-af.org/2009072676/Linux/Specifique-Distribution/Gpugrid-en-ligne-de-commande-Gentoo.html

Unfortunatly it is in french :p

The point is, Boinc has to recognize your cuda device without launching Xorg.
Then you have to create some nodes. (You can modify and use his script to do so)
But this seems to work only with GTX2xx cards.

And before shutting down your computer, you have to add a delay of 5s in order to avoid GPUGrid error.

Nice job !
Thanks for the translation.
And thank you Elgrande for this guide !

I did a rewrite of the runscript above to make it more generic. The message board eats the indentation, though.

Please pay attention, that if you use the script, and restart the X server (later), the X server can hang. Without starting X, everything is OK...

One last trick.

In addition to the big advantage you get by NOT running X, you get better performance from the card if it is not your primary display. I have the onboard video as my main display and do not run X. This dedicates the GPU 100% for CUDA. I recall reading you can do the same with a cheap PCI video card if your motherboard does not have onboard video.

I should have specified. The script is meant to be run instead of X. I not sure how either script could cause X to hang, though. They only create device nodes. X servers can be finicky at times, so maybe.
PS I missed something when I wrote the script. It shouldn't say "Aborting" because it doesn't. It tries to create the nodes even when modprobe fails.