Posted on 08/02/2006 5:16:19 PM PDT by texas booster
Time for a new FreeRepublic folding@home thread.
Our FreeRepublic team of 358 members comprised primarily of Free Republic members in good standing have banded together to donate their excess CPU cycles to a worthy cause. Via distributed computing, millions of computers around the world, contribute directly to scientific research, in the quest for a greater understanding of diseases such as Alzheimer's, Cancer, and Mad Cow (BSE).
Currently, the team is in 75th place (with 1009 active CPUs - 55,700 completed Work Units and 9.75 million points).
This is an entirely voluntary program, and if you want to learn more, please see the links posted below (or read one of the previous 20 folding threads)
I've been looking for info on dual core optimization as well.
So that's how you stay a step ahead of me. ;-)
Shurato's in the top 200! Wohoo! My big brother was over at my place last night, and although he's a liberal, he said he always liked Reagan, so he doesn't mind helping us out. :D
I have a dual-processor Ubuntu Linux machine folding. I actually had to download (well, my brother, the Linux genius did) a specific build of the Linux core to enable dual processor functionality.
After that, I think you can assign each folding process to the correct processor.
Dual core CPUs can show a gain greater than 30%, but there are limitations. From another FAQ:
Q: Should I use the Console or GUI version client?
A: I highly recommend the Console version only! It uses less overhead and is capable of running two clients at once. Single-core and non Hyper-Threaded systems should only run one console client. Hyper Threaded and dual-core system should run two console clients at once in seperate folders such as CPU1 and CPU2. GUI clients have reported problems while running other applications and games.
Q: Are there any configuration settings that you could recommend?
A: Yes. On a single running client the memory usage should be no more then the maximum amount available, and on a multi-CPU dual-client setup each should be configured for half the amount of available RAM. Of course, these are maximum amounts, so you can also further reduce them. The core priority should always be idle unless this is going on a system that is very seldom used. The CPU usage should be kept at 100 unless your system is overclocked and usually runs hot, in which case you should lower this amount slightly.
Q: Should I really run two F@H clients at once on my Pentium 4 HyperThreaded CPU?
A: If it is a 1MB cache or higher P4 HT CPU, you would see some benefit from running two clients at once. If it is not, I advise against it since it is not true dual core CPU. It is possible, but the benefits are not as great as a true dual core CPU.
Q: How do I set up two clients to run at the same time?
A: You must use the console version of the client for both instances. Create a directory (folder) for each client named CPU1 and CPU2 and run each from their own directory.
I also found insight to hyper-threaded processors at F@HWiki:
http://fahwiki.net/index.php/FAH_&_SMP
I saw your first points post today.
Thanks for joining the folding team.
WELCOME. It's great to have you "folding" with us.
http://folding.extremeoverclocking.com/user_list.php?s=&a=2&t=36120
Gotcha!
Power is back on. I expect over a thousand points soon (today/tommorow).
I run both. My cad workstation Folds, and the home machine searches for ET. The home machine isn't hoss enough to run the Folding software.
Good luck to you. I knew it would be short lived. I lost one of those MONSTER brodskys at 90% complete two days ago and lost all hope. Seems that your power problems allowed a brief pass anyway...WOOHOO!
Thanks Swamp Fox.
My break is over - full speed ahead!
I've added some filtering to the project histories that effects how project stats (as well as machine stats) are calculated.
I'm looking at project completion times, and filtering out anything under 100 minutes, and anything over 120,000 minutes (1000 hours) as invalid. Although rudimentary, it should get rid of instances where a project aborts right away, and machines that have stalled for long periods.
Filtered history is still shown (in red), to allow for machine-level analysis (i.e. why does the @#$% machine keep throwing up??), but the stats aren't being figured into calculations.
This isn't an ideal situation, though, as there are still instances where a project has aborted over the threshold. If anyone can point me to a better algorithm, that would be great.
Right now, I'm thinking of a percentage-outlier approach: If a project completion is less than 50% of the average, it's probably invalid. Same on the flip side: if a project completion is 200% of the average, it's probably invalid.
Thoughts?
Hey, Egon. Thanks for all your hard work.
Scatter-shooting while pondering machine stats:
In my download of your script file, I've used the USER_MACHINE_SECTION field to designate either AMD or Intel processors and the USER_MACHINE field to designate the motherboard of that machine.
Perhaps we could have a few more fields in the script file that would give better info on stats. I'm thinking about specific fields for CPU, Motherboard, Desktop or Laptop or Server, Total RAM, RAM allocated, % CPU allocated, Avg. time powered up & folding, main use (dedicated to FAH, web browsing, business, research, gaming, etc.), other distributed programs running.
I know that's a lot of info, but knowing that a laptop is turnded off 75% of the time or that a system is a 486 or P-1 (or a gaming system)could help explain those very long crunching times. And if the system is the latest greatest server with raid arrays and gigs of RAM and dedicated CPU, then that could explain those really fast times.
Also, is there any way to parse info from the FAHlog to determine when a WU couldn't be completely folded and the partial unit sent and credited, or whether it was abandoned?
Just some random thoughts for your consideration.
RT
Someone brought that up awhile back-- might've been you. I can't remember. Some sort of notes field(s) is definitely a good idea, and it's on my list of things to add. The upsides are what you've mentioned. The downsides:
Also, is there any way to parse info from the FAHlog to determine when a WU couldn't be completely folded and the partial unit sent and credited, or whether it was abandoned?
I looked at this early on, and dismissed it for these reasons:
Thanks for the feedback!
"Thanks for the feedback!"
You're welcome.
"I'm fundamentally lazy"
U2? LOL
"can't parse for comparisons because people will enter similar information differently-- it'd just be for human comparison"
Perhaps a limited number of choices (prompts) set by you indicating how you want to see them, with the last one being 'other', might work.
"only some of us will likely take the time to fill the information out-- again, making it useful only anecdotally"
hmmm - can't argue with that, probably more than just a few other lazy dudes out there, LOL. (And I sure wouldn't want to tick off Klutz with all the systems he would need to reconfigure.)
;^D
Add me to your list. I'm a system administrator and we have a few servers/workstations that can run this program along with my own workstation.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.