Einstein@Home

The Rocketry Forum

Help Support The Rocketry Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Right... I will try and get this up and running on my linux box over the weekend :) .


Phil
 
...And moving up fast on the outside...we are up to 139th place! Not bad when you consider some of the other teams ahead of us have 40-50 participants!
 
Originally posted by RocketmanTM
we are up to 139th place!

Rocketman, where are you reading that? I hope you are right, but I see us being in 197th place.
 
Hi

are you folks running seti@home at the same time as einstein@home?

Pete
 
I am running Seti, Einstien, and the Protein Predictor.
 
Just had a look at the stats - Illini has just taken the lead from Andy, and we're at #159 in the team rankings :D
 
I have a *huge* number of pending credits. Every so often one pops out to keep my recent average high.
 
Been reading E@H forums. Seems that Linux client is massively unoptimised, so I'm experimenting running windows client under Wine. Some people have reported speed ups >80%!
Thread here...
https://einstein.phys.uwm.edu/forum_thread.php?id=1208

So far everything is installed, I've attached to project, got some WU, but top CPU% is 1% so far :(
At least the AMD64 is still chugging along nicely.
Now if only E@H would release the einstein source, we could get some real AMD optimisation going
 
Wine! :eek:

Wine wins the Cydermaster award for being the linux app I have repeatedly had the most trouble getting to work!

Ah well - Its Sunday, and I've got nothing else to do (appart from housework etc, but I cba to do that :D ) .... wine here I come.
 
It is by no means perfect, gui is seriously flaky, but boinc_cli looks like it is doing its job.
CPU temp has gone up 1-2deg compared with before, and the xml file is updating, suggesting it is working, despite CPU usage at 1%. Currently at 18% of 1st WU after 1.4hrs, which extrapolates to around 8hrs/WU
Previously with linux client, was taking 13-14hrs/WU.
I'll update when it tries to return result & claim credit.

Edit: FloatingPt & Integer benchmarks are up from 877 + 1998 to 1517 + 3598, so that's gotta be good too surely?
 
Ok - I've got the windoze cli working too - the gui "worked" for setting up and initialising of the project (d'loading needed files etc), but then crashed & refused to run :rolleyes:

I've yet to see if its gonna make any differance.

For those interested, I've written a couple of simple scripts:

1) To start Boinc-ing: (winboinc)
Code:
cd '/home/dan/.wine/drive_c/Program Files/BOINC'
export WINEDEBUG="err-all,warn-all,fixme-all,trace-all"
nice wine boinc_cli.exe -return_results_immediately 2> /dev/null &
2) To see progress: (winboincstat)
Code:
grep fraction_done '/home/dan/.wine/drive_c/Program Files/BOINC/'*.xml

Does seem rather daft to be running Winblows code on a *nix machine; after all, I installed Linux to get away from the **** stuff! ;) , but needs must etc............
 
My progress status checker seems to render into a screen full of smilies, so point your browser at:
andy's boinc status check

It is linux only, and requires xsltproc installed (and associated libs).

Just run the .sh from bash

You'll need to edit the path to the BOINC directory obviously. You should be able to point the path at any boinc instance... native linux, windows client under wine, any client via nfs/smb mount etc. (same should be true for cyder's thing too)

Edit:
Found a bug in mine... need to read up on XSL again... It only seems to work if you have 1 project. If you have 2 (eg S@H & E@H) then it only lists one of them.
Also you'll probably spot that my script was based on Cyder's and infact Cyder's code is just commented out.
Will update tonight.
 
Originally posted by andysrockets
(same should be true for cyder's thing too)
But Andy's is much neater, and miles more useful. I've ditched my stat script in favour of Andy's, and learned a little about xslt in the process. :D
 
Check this out....

The Rocketry Forum team is #82, in terms of daily credit, according to the latest stats at 'boincstats.com':

https://www.boincstats.com/stats/team_stats.php?pr=einstein&co=&st=0&to=100&or=12

...so this means we ought to keep moving up the ladder, in terms of total credit. Not bad for a small team!

I'll be adding another Athlon XP 3200 system in another week or so, which ought to help.
 
Update on E@H Win client under Wine on Linux...

28000 sec compute time instead of 47000sec! Credit is pending, waiting on 1 more host for validation.

For some reason it won't start the next WU though :(
 
Current rank is 137! Duke University is within striking distance. A little voice inside me is demanding that we stomp Duke.
 
Originally posted by illini
Current rank is 137! Duke University is within striking distance. A little voice inside me is demanding that we stomp Duke.


Well, our current rank is 129!!!!

And Illini got his wish, Duke University is 133.

We still have 3 members on our team that are not showing up.....
 
To all Linux users...

So far I have returned 4 WU which were crunched with the Win client under Wine. 28600s seems to be average, but was 48000s before. Still awaiting credit.

As an experiment, I did Seti win client under Wine too. I estimate about 1000seconds or so being wiped off the crunch times, so not so brilliant there. Oh and all WU for seti have validated :)

Everything seems to be chugging away nicely now. WU are being returned as expected and next one starts without waiting. Not sure why there was problems on the first day or 2. And still not sure why CPU load indicates 1-2% max.

PS this is on AMD Athlon architecture... maybe the speed up is less pronounced on Intel? Just goes to show how badly implemented the E@H code is from a compilation & performance point of view.
 
We're at 123, and Purdue is right above us in striking distance. Purdue??? Purdon't!!
 
Couldn't see at a glance but any ideas what the absolute minimum spec machine that can usefully be used is?

Processor
memory
H/D

I ask as I have a few very old Pentiums sitting around begging to be used for something.

It will be a while before I can get any processors on line as I haven't made the move to broadband yet and am still getting the home network set up.

My XP MCE box is up and running 24/7 but mostly idle (P4 3.0MHz Prescott + 1GB + 40Gb + 250GB) so a reasonable bit of power there (may not be suitable as it is in the lounge and I don't want it making a racket all evening, as a media machine it hardly has the fans above idle). I am building the server at the moment (just a little VIA C3 800 MHz) + a few (very) old machines knocking about.
 
Originally posted by Daedalus
any ideas what the absolute minimum spec machine that can usefully be used is?

I could be wrong, but I don't believe there is a minimum. A slower (by today's standard) computer will just "work" slower. I would say, fire them up!!!!
 
I'm was wondering about resurecting an old 'Attic-ware' machine (an AMD350) to help a little, running a non-X linux distro but with the optimisation problems the linux client suffers from I'll shelf the idea and save some electricity.
 
We're ranked #108, have passed Purdue, and are now within striking distance of the US Army!
 
I was going to wait until we hit top 100, before saying anything, but that might juts be sooner rather than later.

Illini... I think your linux opterons could use a little optimisation. I was looking at your stats for one machine compared to my measly little XP2000+. It appears that the linux client is so pants that my measly Athlon running the Windows client under wine is crunching faster than your opterons :)

So you may want to give it a try... install wine & windows client. Not sure how well that will cope with SMP though.
 
Originally posted by andysrockets

Illini... I think your linux opterons could use a little optimisation. I was looking at your stats for one machine compared to my measly little XP2000+. It appears that the linux client is so pants that my measly Athlon running the Windows client under wine is crunching faster than your opterons :)

So you may want to give it a try... install wine & windows client. Not sure how well that will cope with SMP though.

Of that there is no question. I'm operating purely on brute force at the moment! :)
Actually, all of my opterons are offline doing real work at the moment. But will hope to get them back online over the weekend. Will look into wine before then.
 
Originally posted by andysrockets
Illini... I think your linux opterons could use a little optimisation. I was looking at your stats for one machine compared to my measly little XP2000+. It appears that the linux client is so pants that my measly Athlon running the Windows client under wine is crunching faster than your opterons :)

His Opterons were kicking butt there for a while!

The Windows code must be optimized for Athlons, that or the Athlon's big 512KB cache is really helping. My Athlon XP 2500 "Barton", which is overclocked to an Athlon XP 3200, is running WinXP Pro, and cranking out WU's in under 6 hours. Not bad for an $82 processor!

BTW, if anyone besides me has upgraded to BOINC 4.25 or 4.26, I noticed my benchmarks went way down from their original values (especially the integer speed value). Anyone else see this? I don't think my WU throughput has gone down, though.
 
Originally posted by vjp

The Windows code must be optimized for Athlons, that or the Athlon's big 512KB cache is really helping.
That, I seriously doubt. Windows code will be optimised for fairly generic i486/586/686. 512kb cache is a bit weedy these days, but Athlon/Opteron does have a superior pipeline.
Pentium-M and I think most recent P4's are 1Mb cache.


My Athlon XP 2500 "Barton", which is overclocked to an Athlon XP 3200, is running WinXP Pro, and cranking out WU's in under 6 hours. Not bad for an $82 processor!
That is some overclock!
 
Originally posted by andysrockets

That is some overclock!
Overclocking the Athlon XP 2500 to a 3200 is about the easiest OC'ing you can do. The chips are identical in every respect, the only difference being that the FSB is 400MHz on the 3200 vs. 333MHz on the 2500. And the quality of the chips is so good that virtually every 2500 can run at 400MHz, rock solid and no thermal issues. So you plug a 2500 into any nVidia nForce2 motherboard with good PC3200 memory, set the FSB to 400MHz (actually 200Mhz x 2), and you're set.

I'll have another 3200 up and running (hopefully) by tomorrow night, so I'll be crunching some serious WU's then - MWU ha ha ha ha:D
 
Back
Top