Reply With QuoteWe lost one position.
Carlos
39th place and going up. The server and the client are very stable. Lots of work to crunch.
Carlos
We lost one position.
Carlos
Shouldn't Free-DC be in TOP 20? Who wants to help?
CC
only one DC project Free Dc would be number 1.
The more projects the more spread out our little band of brothers and sisters get.
There is a push on FAD, EON, DIMES and Folding @ Home as far as i know. There was a push on BOINC LHC but that ran out of work.
There is talk of a gauntlet in october over at FAD but TBH at the moment we are all doing our own "thing".
Best thing to do if you want to drum up interest is pop a post with all the salient bits about the project and see who fancies popping some GHZ on it.
Some people don't care for BOINC.
Good luck drumming up firepower.
I'm running it a little as I try to work out whether Einstein is doing enough work on my linux boxen. Only have a single opteron running it right now, though i had the X2 on it for a short period doing 2 wu's every 20mins
Bok
The only problem I see with Predictor is.
I ran it for a week with all my systems. I tried to get enough units downloaded for at least 24hours of work. But their time aloted for returning wu's just doesnt seem to work for the fastest machines I have.
They only downloaded enough work for 18 - 20 hours...... But of course I am on dialup, and can only be online in the evenings when I get home for work.
Since you can read this, Thank a Teacher.
Since you are reading this in English, Thank a Veteran.
Personal Web Page
http://www.maqs.net/~jmerlin/
Short term goals:
Team BOINC@AUSTRALIA
CC
Since LHC is out of work currently, I've thrown this on the BOINC farm.
Let's do some stompin'
If I may suggest, having been with P@H for a long time, that we download approximately 4 times what we think we will need. The actual numbers vary, but typical is a 3-4x overestimation by P@H due to the *potential* for a very complex WU (which you will get from time to time).
For those of us on dialup, I suggest a simple download all WU's, then disable network access until you get home / can enable the upload-download session.
To prevent yourself from getting WU's that timeout during this first batch....
download 2x what you need before bed, let it run overnight with network disabled... In the morning, get online, update the profile to 3x, and then let it upload and download again, disabling network access when done again....
Repeat this process of increasing by 1x until you achieve balance and a nice load of WU's that are staged 1 day apart. The optimum is for a nice 'spread' that will keep your machine busy while you are away and won't expire on you in the typical '1 week' window they give us.
It will take time to build up to the appropriate queue level as we each have an initial quota of 165 WUs permitted to be 'in progress'. Use your best judgement in what fits your needs / speeds.
I personally have my machines set for 2 days worth of work each, but know that I can churn through that in about 10-12 hours if the DSL goes down. Having DSL, I get a single WU each time I upload one. If I am planning maintenance, I will grab an extra 2 days worth.... obtained before work, after work, and before bed... it's not the best but it works.
Hope this helps and hope I'm not too 'foggy'... it's been a long day for me...
PS: Nice to see us kicking some tail and doing some stomping. Looks like Combustion's goals are pretty easily reachable.
G'day and G'nite
Chuck
What happened to my invitation to the beta then.
I notice that you made sure TPR members got on it.
There was a management and subsequent policy change at P@H. The # of users was cut off. After that, (Michaela's departure), I left DC entirely until just a few days ago.
I do apologize for not getting the authorization for you to run as yourself. The best I could have offered you was to run under my username. I didn't think that would have been acceptable to deny you your own points.
As you have noticed, this is my first return to the DC scene since P@H changed.
It's still not like it was, and never will be quite the same again. I submitted a bug report directly (with screen capture) and it was completely misunderstood.
If my return now is not desired, I will return to my other work.
We can discuss in pvt if you wish.
Chuck
Chuck
I noticed you hadn't been around on the TPR forums.
I must admit to not being best pleased at what I perceived to be an attempt to exclude certain folks from the testing.
But as you say the decision ultimately was not yours so no hard feelings.
Are you going to join the team here or continue with TPR ?
I noticed you used the term 'we' in reference to the FDC team.
Your welcome here.
Thank you for the welcome. It's nice to be back and contributing to science & humanity. I do plan to stay here at FDC unless otherwise requested/stated. I will not be very vocal as work is very demanding... The machines are running P@H now, but I will have to switch over to running my thesis work again soon for a week or so. I simply took this time while doing some white paper and code writing, including the new algorithm development, to give this a 'test drive' and am glad I did, for which I thank you and all. It did take some convincing to get me back into the forums & DC arena. I admit to feeling like a neophyte all over again.
I used the term 'we' in reference to those of us running P@H, but I am sure you will notice that anyone with an X86 machine with BOINC has to make adjustments for the 'cobblestone' calculations. Mac's seem to be the only machines immune (perhaps their code was fixed but never included).
I thank you again for the welcome.
To you and everyone... please don't ever hesitate to ask me a question or give me a pokeif I need it.
Thanks,
Chuck
PS: I must get some sleep now.. it will be dawn soon and 88F is as low as it gets around here right now, so I will get my sleep in while I can.
Until later.
Cheers PCZ.
PCZ,
Chuck is already crunching for Free-DC. Check the stats.
Try crunching, just for this time, a few Predictor wu's. Come on, for the team.
Chuck,
Welcome to Free-DC Team.
CC
Combustion
PCZ
95,679.45
I think thats enough cobblestones
Besides 5 mins after I get set up the server will run out of WU's or crash.
Probably both
I see, you're afraid of my 200 machines.Originally posted by PCZ
Combustion
PCZ
95,679.45
I think thats enough cobblestones
Besides 5 mins after I get set up the server will run out of WU's or crash.
Probably both
CC
No computers don't scare me. users doI see, you're afraid of my 200 machines.
It's always the same excuse. Bah.Originally posted by PCZ
No computers don't scare me. users do
CC
Which version of Boinc is predictor running ?
Do you still need to compile the linux version yourself to get realistic CPU ratings or is it fixed now ?
Sorry for the delay.
Yes, you need to roll-your-own... you also need to fix the code because it tests 1 cpu then reports 'results / ncpus'. That one caught me until I fixed it. With that fixed, it's not so bad. Just do the standard 'adjustments' we always do.
I am having trouble getting a good 'roll' on FC4 in x86_64 mode though. I am going to pull up VMware tonight and see what happens... should be a nice clean layer of win98 on FC4 in 64 bit. If that works, I'll go from there, but I still want the native 64 bit version.
If you have any advice / success, I am all ears.
I wish to thank Helix, PCZ, Condor, and DevZero for successfully sucking up the available WU's. I did expect it to take less time than it actually did, but no matter. Good work gents!!!
Time to take care of the competition
I will spend a bit of time trying to get a 64 bit client running. My last attempt was unsuccessfull, but the development environment is far more stable now.
I'll keep running, but don't be surprised as my stats drop. until the server catches up.
Last edited by Chuck; 08-25-2005 at 02:03 AM.
Hey Combustion!!!!!!
You got the first part of your wish!
28 Poland 105 1,815.44 478,349.89 Poland
29 Picard 37 2,705.42 462,181.30 International
30 US-Distributed 56 1,713.30 456,578.81 United States
31 Free-DC 40 5,366.43 444,455.19 United States
32 BOINC@AUSTRALIA 58 3,199.35 443,883.97 Australia
Gonna take one heck of a blast for them to over take FDC now with all these Ghz cookin.
Last edited by Chuck; 08-25-2005 at 05:26 AM.
One down, two to go.
CC
I noticed on a dual MP machine. It would start out with 2 instances, then after a while only one instance would be running. I just attached to einstein in addition to Predictor on that patricular machine.Originally posted by Chuck
Yes, you need to roll-your-own... you also need to fix the code because it tests 1 cpu then reports 'results / ncpus'.
Bok
you can set your max to two in your general prefs for any project.
on duals, there is a handshake bug until it gets into 'round-robin ops',
then all is fine.
when it gets behind, because of the lack of affinity locking, it drops to 1
per project for a few hours, then once caught up, round-robin mode returns.
round-robin ops is visibile in the log and a good sign you are there is 'reschedule_cpus: project op"
i am looking at locking cpu affinity at process exec and submitting the change,
but it indeed does work nice for two projects with single instances of each.
Bok
I observed the same thing on my duals.
Seems to be a problem with the Einstein nix client.
Strangly boincview was saying that both clients were running so took me a while to spot it.
Detached Einstein and haven't seen the problem since.
Another thing is return_results_immediately doesn't seem to be working properly.
Some times the clients are sending in results shortly after completion but mostly they build up waiting for me to update.
PCZ,
You can check in the source for the call to request_resched_cpus
when a process terminates. sounds like it's not being executed
under all conditions. I am having the same problems as you
with the duals but only occasionally...
I'll take a look, but two or more sets of eyes are always best.
Chuck
30th Position.
New short-term goals:
CC
2 boxen online, not much but it all helps.
We are up one place. Team Picard is down.
CC
Poland today
DPRGI tomorrow
Canada and Erster Kontakt Tuesday
Another 2 weeks of this pace and we'll be in 20th.
I see PCZ going for 1st place. Watch out Helix Von Smelix.
Poland is down.
CC
Within 10 days we will enter TOP20.
CC
I bet you we do it in less than 7.Originally posted by Combustion
Within 10 days we will enter TOP20.
CC
As for the next team to go down, I have but one friendly
thing to say:
21st most probably in 10 days but Anandtech don't go down quietly, usually.
It's a pride thing, even though we usually right royally
So here's to thewith our "buddies" at Anandtech.
Posting Permissions
