LITTLEBLACKDOG.COM

Posted: Sun Mar 04, 2001 9:37 pm Post subject:

Now that the seti server is back up and running we can all start crunching away again... OH JOY!!!

During this outtage, I was dreaming up ways to get even greater speed out of the seti client.
What about running the command line client using dos on a P4 or T-bird. Without all the windows overhead... would this or would it not just scream through wu's?

Edited by - geminione on 03/05/2001 00:49:02

Big Dog Joined: 20 Oct 2000 Posts: 3689 Location: USA

it might work. . it might not

my guess is you'd need 32 bit DOS (DOS 7 or better) to get it to work . . the downside is that you wouldn't be able to cache any WU's because all of the driver programs I've seen are win32 native only ( aka DOS is not a multitasking environment

I think you'd have better luck running a Linux Command line client, and not loading Xwindows . . if you wanted the best throughput possible ?

Blame is for people who need a reason to be pissed off !
-Brain

Posted: Mon Mar 05, 2001 9:07 am Post subject:

My question to this, if my comp is just sitting there doing nothing but seti, is the windows overhead really that high. I have never seen seti use more than 16MB of RAM, I almost always have some 64MB of ram free with seti running, so that isn't a limitation. Is windows(Or XWindows) really using my processor very much when I'm not doing anything?

Eagles may fly, but Weasels don't get sucked into jet engines!!

Big Dog Joined: 20 Oct 2000 Posts: 3689 Location: USA

Quote:

Is windows(Or XWindows) really using my processor very much when I'm not doing anything?

Seti uses less than a meg of memory (or it should be using less than a meg anyway)

as for the processor intensivity wasted by windows .etc. . I guess it depends on how you try to measure it

the only 2 ways I can can think of are overall output (how many units linux can put out from the command prompt. . versus linux running X-windows would put out . . . take that over like a 30 day period, and average the total down to a daily output

and responsiveness . . which is much harder to measure.

such as finding a specific task that can be run from command, or by gui, and seeing how fast this task can be completed under CL or GUI (something like a script file I'm guessing)

This is something you'd have to test for yourself (since I'm not a Linux guy)
I do know you get alot more work out of a low end machine running the CL version than you can with the same machine running X-win(even after doubling the ram for the gui)

I think the biggest place you would notice this would be under an SMP setup.
or if you were taking something like a celery overclocked at 100mhz FSB running the CL and comparing it to a P2 that was of equal multiplier but running gui . run the same process on both, and in theory the one with more cache should pull ahead (even if only marginally so for processes bigger than the cache of the p2)

long story made short . . it's hard to measure the difference real-time, but it would be noticable over multiple processes or when measured over a long enough period of time.

Blame is for people who need a reason to be pissed off !
-Brain

Posted: Fri Mar 09, 2001 3:37 pm Post subject:

If you want to eliminate overhead from windows, I would HIGHLY recommend you try running a stripped linux distro (like redhat, immunix, mandrake etc). Do a small install and kill off alot of the services (see inetd.conf). It would be alot less hassle than DOS. I think the cmd line client for windows versions requires a winsock to work, which requires windows. Don't quote me on that.