Einstein stability?

[Brucifer]

Hi,

I've been dinking around with boinc again. Seems like I do this about once a year to see how the stability issues are going. I was tied in with the initial closed beta in the beginning, but drifted away once things went in to production.

Anyway, how is the einstein project? Does it run out of work or have plenty of work units? Just for gits and shiggles I put four systems on it a short while ago just to see how they do.

What is the best setting for computers talking to the server? I'm currently set at 2.0 days. Too much/too little? I'm using the boinc_4.19 linux client. Anyone else doing the linux thing here?

[Empty_5oul]

i cant comment on the linux side of it.

generally i have found 1.0 to be best. But communicate too frequently and you store no work so if a project dies or maintainance or whatever you are left with nothing. Then again communicate too slowly and you store a load (filling your HD) and then when you finally comunicate send loads in one go (kill bandwidth)

while a project is stable i would go around 1-2 days. Very instable or few work units go for 0.1 so you request them more frequently.

[carlos]

I thought you hated BOINC!?!? That pending crap system, blablabla, etc

[NeoGen]

Empty_5oul I don't know... but Beerknurd I'm very sure he hates BOINC. :lol:


...although I've seen him running SZTAKI some time ago. :P :lol:

[Brucifer]

I thought you hated BOINC!?!? That pending crap system, blablabla, etc

That's right Carlos, it sucks with a capital "S" but so does a bunch of the rest of the stuff. EON requires a ton of babysitting, or else the computers end up wasting a bunch of time with the zzz. Then there are also the issues with D2OL and all the lack of communication with staff, etc., same for TSC. And rumour has it that TSC is going boinc. So things are sort of sitting with the only stable non-boinc project being distributed.net, or SOB which I'm not a fan of, or Rieselsieve which I get my fill of every now and then but get bored with in a hurry.

I fired seti off to crunch a bunch of them, but then it has issues too like going off line (like they are now, 3 hours that has grown into an overnighter and so far today, and, and, and.

So one can either just turn off the computers and chuck it all down the tubes, or look around and try to find something. Well so far, about the only boinc thing that I'm crunching with a degree of success that I like the project is einstein@home. That also doubles as we have some local people working on that, so one should support the local LIGO site. And wonder of wonders, I'm even getting credit granted from einstein, where I haven't received squat from seti even though I put several days into crunching and have a ton of pending.

And who knows, if einstein keeps on running, and I keep on getting some points, maybe I'll even join a team with einstein. Presently I'm just running it solo.

[Empty_5oul]

i dont have a problem with BOINC. As long as the project is stable and works (e.g. stats update and work is available) its fine.
I cant say i particually like it either but i think this is down to a bad project or two.

:lol: we have a team at einstein

[carlos]

My friend 72CJ5,

I ran for a few times BOINC Seti and if I am not mistaken I got credit for the wu's two weeks after they were crunched. Never tried it again.

You can shutdown the machines and try to learn something else like me. I am studying Adobe Premiere Pro 1.5.1. I found in the internet a "Lynda.com - Learning Adobe Premiere Pro 1.5" collection movie and bought some books.

Sometimes I help Lars at Prime Sierpinski Project (http://www.mersenneforum.org/forumdisplay.php?f=48) and your help would be useful.

Can't say much to a stats whore.

Carlos

[Brucifer]

Carlos --

So far I'm happy with einstein@home, and will continue crunching it. If seti ever gets back online I may crunch it some more, will just have to wait and see. As for shutting down stuff, :( that would be a real tear-jerker for me. It's really hard giving up being a stats ho! :) I would imagine that even the staunchest boinc haters would give boinc a whirl if there wasn't anything else they were really happy with.... or even close to that. IF a boinc project runs along fine with plenty of work available, then I can stomach it. The pending stuff sucks, but then there isn't much choice. On linux, the 4.19 boinc client seems to be pretty solid and stable, and for that matter I don't have any problem with the 4.43 client either.

In looking at the team stats for einstein, AMD_Users is doing pretty well. So I joined the team. At least there are some people that are really active in it, and that is good to see!

[Brucifer]

i

:lol: we have a team at einstein


Yes, I notice that it is one of the better producing ones too. And yes, I joined it. :)

[vaughan]

:lol: we have a team at Prime Sierpinski Project too :!:

[NeoGen]

And on some 30 other projects... :P :lol:

You could even run one different project a day, each day of the month. :lol:

[hallen]

I've been running E@H a little over a month and its been solid for me.

Quick to, on dial-up, as a package of 7mb lasts about a week. I think this package/configuration file reduces the servers work load so I never get the dreaded No schedulers responded

I suspect this set up would allow E@H to handle large increases in traffic.

Pending is running from 1/2 to 1 1/2 days production.

E@H is great with AMD cpu's too. 64k L2 Duron 1.6Ghz @ 2.0Ghz outperforms 512k L2 P4 3.0Ghz @ 3.6Ghz.

[Brucifer]

E@H is great with AMD cpu's too. 64k L2 Duron 1.6Ghz @ 2.0Ghz outperforms 512k L2 P4 3.0Ghz @ 3.6Ghz.


This kind of information is alway nice to hear. I haven't seen anything related to the various cpu performance. Perchance is this documented anywhere?

[hallen]

Perchance is this documented anywhere?

No I haven't seen any. You just have to make your own.

My conclusions so far


Duron 1.6GHz 64k L2 @ 12 X 166 = 2000MHz 23,149 sec cpu 6:25:49 wu
BOINC Benchmark calculation 5:56:21

http://members.localnet.com/~hallen/applebred1600.JPG

T-Bred 2600+ 256k L2 @ 12 X 166 = 2000MHz 22,974 sec cpu 6:23:00 wu

T-Bred 2600+ 256k L2 @ 12.5 X 184 = 2304MHz 20,009 sec cpu 5:34:13 wu
BOINC Benchmark calculation 5:08:56

Conclusion: Memory and cache not important. Run that cpu as fast as you can.

I have also found that true for the Socket 754/939 boards. Socket 754 single channel memory 128k cache Semprons will perform about the same as Socket 939 dual channel memory 512k cache A64 Athlons at the same clock speed.

Here are some numbers for the BOINC benchmark

A64 3000+ 512k L2 8.5 X 318 = 2703MHz benchmark = 4:20:12

Sempron 2600+ 128k L2 8.0 X 326 = 2608MHz benchmark = 4:27:42

Sempron 2800+ 256k L2 8.0 X 330 = 2640MHz benchmark = 4:25:25