SPCR Folds Team Blog
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Posts: 871
- Joined: Fri Jan 05, 2007 11:37 am
- Location: North Carolina
Well, we lost two more places.
We are now officially #25
Yes we did lose two places but on the bright side, we are 25th in the world! The teams around us seem to produce so many points. We've lost five places in the last four months and we're scheduled to lose three more in the next six months.
On another note I'm looking to build a small folding farm. Its purpose will be to combine the power of several old computers. Most farms just use several computers to individually fold. I'm looking for all of the computers to work on the same WU. Possible? Any ideas? (Computers are two 1GHz P3's, 1.7GHz P4, 2.6GHz Celeron, and possibly an Athlon 64 3200+.)
We are now officially #25
Yes we did lose two places but on the bright side, we are 25th in the world! The teams around us seem to produce so many points. We've lost five places in the last four months and we're scheduled to lose three more in the next six months.
On another note I'm looking to build a small folding farm. Its purpose will be to combine the power of several old computers. Most farms just use several computers to individually fold. I'm looking for all of the computers to work on the same WU. Possible? Any ideas? (Computers are two 1GHz P3's, 1.7GHz P4, 2.6GHz Celeron, and possibly an Athlon 64 3200+.)
-
- Friend of SPCR
- Posts: 1439
- Joined: Tue Dec 14, 2004 4:06 pm
- Location: New Hampshire, US
- Contact:
well, I just got my cousin all set up with two single core clients on his X2 machine. He didn't really know how to deal with SMP errors, so i figured this was the best way, it's easy to use, and he doesn't have to remember to start it up, because it will run as a service. I've got a friend who could fold with both his PS3 and Q6600 system, but he's reluctant, and only folds with the PC every so often. But, it is for our team. I almost had a quad core opteron system folding for us this summer, but it had some serious stability issues. Oh well.
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Wibla, you may be able to reach 401MHz FSB because the northbridge divider calculations will adjust themselves so that many components on the motherboard will return to their original values.
http://www.overclock.net/intel-motherbo ... -975x.html
Angelkiller, I would like to be able to monitor say 4 boxes from just one display.
People used to take pics of their farms, haven't seen any in a while.
http://www.overclock.net/intel-motherbo ... -975x.html
Angelkiller, I would like to be able to monitor say 4 boxes from just one display.
People used to take pics of their farms, haven't seen any in a while.
You can use FahMon to monitor several clients at once. You just need to mount the work dirs so that they are readable on the monitoring computer, over Windows filesharing/samba.aristide1 wrote:I would like to be able to monitor say 4 boxes from just one display.
-
- Posts: 26
- Joined: Thu Aug 03, 2006 7:06 pm
Several computers on same wu
Angelkiller,
"I'm looking for all of the computers to work on the same WU. Possible? Any ideas? (Computers are two 1GHz P3's, 1.7GHz P4, 2.6GHz Celeron, and possibly an Athlon 64 3200+.)"
Probably not easy, but might be doable with smp workunits. They result in 4 processes. In theory you ought to be able to send each process to a different machine. People do this kind of thing with clusters. You might look into setting up your machines as a Beowulf cluster. If you want to do it for fun or learning, great. Otherwise, I'd stick with 1 wu per machine.
On a more fundamental basis, some processes lend themselves to parallelization, and others don't. If you have four bakers, when you need 4 cakes each can bake a cake at the same time. If you need 1 cake fast, it won't work to have one baker measure, the second mix, the third bake, and the fourth ice the cake at the same time. these must be done sequentially.
"I'm looking for all of the computers to work on the same WU. Possible? Any ideas? (Computers are two 1GHz P3's, 1.7GHz P4, 2.6GHz Celeron, and possibly an Athlon 64 3200+.)"
Probably not easy, but might be doable with smp workunits. They result in 4 processes. In theory you ought to be able to send each process to a different machine. People do this kind of thing with clusters. You might look into setting up your machines as a Beowulf cluster. If you want to do it for fun or learning, great. Otherwise, I'd stick with 1 wu per machine.
On a more fundamental basis, some processes lend themselves to parallelization, and others don't. If you have four bakers, when you need 4 cakes each can bake a cake at the same time. If you need 1 cake fast, it won't work to have one baker measure, the second mix, the third bake, and the fourth ice the cake at the same time. these must be done sequentially.
You can set what FahMon should base the ETA calculations on under Preferences -> Monitoring: All frames from project, last frame or last 3 frames. The next version will also have effective rate, which is % you are at now / time since you downloaded the WU, good for those who don't keep the computer on 24/7.
I'm not sure how it does the calculations, but it might take a few frames for it to find a good average.
I'm not sure how it does the calculations, but it might take a few frames for it to find a good average.
-
- Posts: 871
- Joined: Fri Jan 05, 2007 11:37 am
- Location: North Carolina
Thanks for the response and sorry for my late one. I seriously don't remember getting an email about this....Probably not easy, but might be doable with smp workunits. They result in 4 processes. In theory you ought to be able to send each process to a different machine. People do this kind of thing with clusters. You might look into setting up your machines as a Beowulf cluster. If you want to do it for fun or learning, great. Otherwise, I'd stick with 1 wu per machine.
On a more fundamental basis, some processes lend themselves to parallelization, and others don't. If you have four bakers, when you need 4 cakes each can bake a cake at the same time. If you need 1 cake fast, it won't work to have one baker measure, the second mix, the third bake, and the fourth ice the cake at the same time. these must be done sequentially.
Anyways, your theory makes sense. As for a Beowulf cluster, it looks 10x too advanced for me. I'm not too good with software.... I was hoping I could get more power out of combining the computers then having them operate alone. I think the points produced when working alone would be so low that it won't be worth the electricity. Great analogy btw.
-
- Friend of SPCR
- Posts: 381
- Joined: Fri Aug 01, 2003 12:13 pm
- Location: Rochester, MN
I just wanted to express my appreciation of your efforts in folding for our little team here at SPCR. I'm sure I speak for all of us when I say that we are very grateful to have had you folding beside us, and hope that you will be able to return in the future, whether with a big farm or solo desktop.
We'll really miss your contribution!
Sincerely,
--KansaKilla
We'll really miss your contribution!
Sincerely,
--KansaKilla
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Greetings,
As an aside, when Stanford posts a newer Beta of the Windows SMP client, can someone notify us, please?
As KansaKilla wrote -- thank you very much, and we definitely will miss your efforts, so come back as soon as you can!vg30et wrote:Just wanted to stop by and wish everyone luck with their continued folding efforts. Some recent changes have forced me to shut down the folding farm so I will no longer be able to contribute points to the team/project. It's been a fun ride and I hope to be able to return again someday.
As an aside, when Stanford posts a newer Beta of the Windows SMP client, can someone notify us, please?
-
- *Lifetime Patron*
- Posts: 4284
- Joined: Fri Apr 04, 2003 6:21 pm
- Location: Undisclosed but sober in US
Sorry to see you go and thanks.vg30et wrote:Just wanted to stop by and wish everyone luck with their continued folding efforts. Some recent changes have forced me to shut down the folding farm so I will no longer be able to contribute points to the team/project. It's been a fun ride and I hope to be able to return again someday.
You know what to ask for for Christmas, right?
-
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Hello,
We are missing the big producers, but OTOH, the #20 producer is averaging almost 1,200PPD, whereas #20 used to be ~600PPD. In fact, the overall average (of the active Folders) is ~547PPD! This is better than TSC! Russia's team average of ~517PPD
We will be better off getting more people to Fold, with single machines, I think.
We are missing the big producers, but OTOH, the #20 producer is averaging almost 1,200PPD, whereas #20 used to be ~600PPD. In fact, the overall average (of the active Folders) is ~547PPD! This is better than TSC! Russia's team average of ~517PPD
We will be better off getting more people to Fold, with single machines, I think.
-
- *Lifetime Patron*
- Posts: 1740
- Joined: Sun Dec 21, 2003 11:24 am
- Location: 'Sunny' Cornwall U.K.
Hi djkest, the answers, or most of, can be found here: http://folding.stanford.edu/FAQ-SMP.htmldjkest wrote:what exactly is SMP, and how do I get it?
Shorter answer, it's the folding client for multi core CPU's. Why bother? well on my XP2400+ single core I was pleased to get 100 PPD (points per day), on my current A64 4400+ I am averaging ~950PPD !!! (Mind you points awarded has changed a lot since I started in Dec 03, my first WU took a week to finish and was worth a whole 40, yes forty points. )
What OS are you using? Lots of forum help to be found either here or over at: http://forum.folding-community.org/forums.html
Regards
Pete