BOINC - Rosetta @ Home New Thread

[Chilly]

Originally Posted by Inik

I have an error on my main rig


01/12/2008 10:28:55 PM|rosetta@home|Message from server: Server error: can't attach shared memory


I'm trying to report 9 completed tasks!

I'm getting the same error on my secondary rig, but my main rig is reporting, getting new tasks and crunching fine. I've tried everything short of "reseting" (ie deleting all the completed work) it.

Grrr.

[Inik]

I lost 9 tasks
I reformated after getting GPU crashes with every freaking nvidia drivers!
Working good now

I still don't know why, but my rivatuner installation disappear after a reboot ..strange!

[misterd]

Originally Posted by Chilly

I'm getting the same error on my secondary rig, but my main rig is reporting, getting new tasks and crunching fine. I've tried everything short of "reseting" (ie deleting all the completed work) it.

Grrr.

I was getting it as well on one machine that i've noticed. Grrrr X2.

[misterd]

Anyone notice their RAC go way down when they set it the default time to 24 hours? Mine seems to be dropping a lot. I forgot i set it to 6 hours way before all the problems started, then changed it back cause it was falling down. Now that i've changed it to 24 hours, I'm having this issue again.

Now i'm on page 2..

[cwestwell]

Yeah I noticed on my Cwestwell Account I set it to 2 days with 6 hour CPU time and my RAC has not recovered from the outage I assumed it was because of the CPU's I had moved over to Aurora but after a closer look I realized that it was becuase my task were taking 4-6 hours with little to no extra points for the longer jobs.

[Jon_di2]

I find that setting my target CPU run time at 2 hours gives me the best RAC... I know they want us to set it higher, but I find that setting your runtime longer gives you crappier RAC.

But having my additional work buffer at 1.5 doesn't seem to make a difference

[cwestwell]

Yeah at the expense of catching up to Dalton the best results I have had over the last couple days have been with 1hour CPU task's with 2 day buffer. however the 2 hour task gives a lot more tasks to run. there dont seem to be as many 1 hour tasks but the points seem just as good from the short tasks as the long ones.

[LCB001]

Originally Posted by misterd

Anyone notice their RAC go way down when they set it the default time to 24 hours? Mine seems to be dropping a lot. I forgot i set it to 6 hours way before all the problems started, then changed it back cause it was falling down. Now that i've changed it to 24 hours, I'm having this issue again.

Now i'm on page 2..

The RAC seems to be heavily backend weighted, my quad hasn't been crunching Boinc since sunday yet my RAC has hardly dropped. Going by previous experience it won't for a couple of days, or until I restart the client. Your drop in RAC is probably a combination of the recent server outage and the lull in tasks uploaded while switching to the longer runtime. Judging from when I was switching between F@H and Boinc it will take at least a week+ for the RAC's to recover...
I've found a combination of 8 hr CPU runtime with a 0.5 day work buffer worked best for me once my RAC stabilized. A longer runtime means less time lost to switching between tasks and the starting and stopping of the CPU. I think I read somewhere on the Rosetta forums that the Tasks were optimized for an 8 hour runtime but can't find the link...

[cwestwell]

Wow Ok so over the last 4 hours my RAC has not moved but the Big D jumped about 1200 points ......what did you change.

[LCB001]

Originally Posted by cwestwell

Wow Ok so over the last 4 hours my RAC has not moved but the Big D jumped about 1200 points ......what did you change.

With a long runtime he might have had a few tasks finish and upload together, because of the server outage they would all start and finish around the same time. Having a longer runtime they would be worth big points and have a big effect when updated...