disappointed...

I've come to lose all respect for linux today. I did a little benchmarking on my various systems today and it seems that XP is tromping Gentoo and Debian on the same boxes by several hours on each of the workunits. It's slightly better under wine running the windows client, but the 'doze boot is still on top..

What's the secret guys? What's the best OS/client to do some serious crunching? I've cut my farm and I'm only running 2 dedicated machines right now for a pathetic 35ish units a day. (clocked1) I'm down in 88th place, and I'm going to start losing ground if I can't come up with a few more units a day and the wife won't let me buy anymore toys. lol

Did you play around with the thread priority "nice". Maybe it was just not getting enough "oomphf" while running. What else was running on the system while SETI was running?

Cy

Did you use the same work unit and bench the same hardware? I've never found more than a minute or two difference at most between various OS's on the same hardware using the same work unit.

Conclusive results are hard to arrive at, but I will admit that in my experience the native windows client running in windows is a bit faster.

Running the win32 client using wine in Linux does get all or most of it back for me though, and occasionally it seems a bit more.

I usually just run the native linux client though.....

I have done a small amount of testing, but there are so many variables that exist I get overwhelmed:
client version
work units
kernel version
os tweaks

All of the above seem to have major affects on the results, and I have just never been able to narrow evrything down well.

My initial statements about relative speeds are born out by my personal SetiQueue running over long time periods though- I'm talking months.

I did the same 2 units on all of the machines with fresh/naked installs of gentoo, debian, xp pro, and 2000 server all through multiboot environments so the hardware and units are identical.

In all instances there were no unnecessary services running and the machines were left to crunch untouched without so much as a screensaver interfering. Process priority didn't even come into play because there was absolutely nothing else going. The 'doze loads had services stripped down, and in the linux loads there was absolutely nothing running. There wasn't even a syslog, cron, X, etc...

My primary box is going to stay in gentoo, but the other machine is now going to be strictly windows. Also, one side note I hooked up my old ultrasparc 10 to see how a 64 bit 333mhz processor with 2meg cache would crunch. Sadly it looks like about 22hours/unit in gentoo. I might load Solaris later tonight and see if the sparc-solaris client works better than the generic sparc-linux client.

SETI is probably the most OS agnostic software I've ever seen. We use it to compare server performance without regard to OS.

You sure your HDDs are setup for DMA and that they are not going to sleep under the Linux installs?

Cy

35 wu a day and your complaining

as i recall you just past us (puffy) not 2 days ago....


well technically you haven't really past us yet cause i got 350 completed wu sitting on my rig at home just waiting for me to dump them


in lighter news I got a new cruncher comming arriving ~20th

pentium m 1.6
512 ddr
40gig hd
geforce "go" graphics 4x agp
<--will be upgraded to mobility 9700 in the near future.
802.11g card
15.4" wide flat screen.
6 hour battery life

for the first time in several years this will be my first rig to own without actually building it myself.

the cruncher will only crunch part time but it's better than nothing.

now to upgrade the 1800+ to a 2400+