boinc stuff

[Brucifer]

So I noticed a recent post on the milky way thread regarding the gpu stuff, and it had to do with issues getting gpu's running on win7 and stuff like that. Not being a boinc fanatic to say the least, I will have to admit that there are a couple of times in the not to distant past where I have even almost remotely contemplated the possibility of installing boinc on a system to fire off on a gpu. So off I go to the various places and before long stuff is inundated with problems, ie no work, connection issues, gpu and boinc version issues, projects either being better on stream or cuda, etc. The long and short of it being that I just figure to heck with it and go back crunching on rc5.

I wouldn't mind giving a boinc project or two a go on a gpu on windows. However there are a lot of projects out there. Milkyway sounds interesting after going and reading about it. What gpus are best on it with the least hassle both setting up and continued operation?

What about the old eternal standby of seti@home. Is work solid on it for gpu's? If so, cuda or stream? Double or single precision gpu's?

What is the team really pushing on boinc, besides everything. Lots of pretty girls out there in the world, but you can't chase them all, DC projects are rather like that. LOL.

So if there is a gpu inclined person on the team that has the good fortune to have been able to figure out all this stuff it would be nice if he/she would post up a how to for team members. And with relatively reasonably priced gpu's too, not the latest and greatest things.

Once upon a time when I was younger, it seemed like various Universities were well known for their specific schools that they excelled in. All the schools want all their fingers in everything now it seems and there is too much mediocrity rather than being truly great in a few areas. DC projects are like that to me. Some people seem to go absolutely nuts in having to participate in every boinc project that comes along. So what's this team really excelling in?

[Nflight]

Since no one answered you and I find myself up early on a Saturday morning no less, here is where we are it seems. This is a view from my mind to yours!

This would be nice if it worked all the time: http://stats.free-dc.org/stats.php?p...team=AMD+Users

IN Case it does not work for you; Collatz conjecture is #1, Milkyway is #2, and FreeHal is #3, after that there are really no serious projects running gargantuan outputs at this time!

As for the GPU competition it seems it is either Collatz Conjecture or Milkyway, no other project has consistency of work units available like these two. Aqua@home and Seti@home and the beta version have work units but they also have times of no work units which leaves one feeling left out or lost with out a paddle!

I have been reading the Distributed.net forums about rc5 and find they are not quite ready for my participation if they can't get there act together it may be a long wait. I hate wasting electrons if I don't have to !

[vaughan]

Collatz = points per day from a GPU are amazingly high. Good project for ATI cards. 5850 is the current sweet spot (thanks to mitchellds for the heads-up on this card).

Milkyway = points per day are very good too. You need a high powered nVidia card compute version 1.3 for this project so that means a GTX260 or higher.

Put your 9800/8800GT cards on PrimeGrid AP26 or Seti or SetiBeta. Both the Seti projects have downtimes of over a day occasionally so don't rely on them as your main project.

GPUgrid = quite large downloads, work and points are pretty consistent and pay well.

Aqua does not support GPU cards anymore, CPUs with at least 2 cores only on this project now.

What is the team focussed on? Not sure as we haven't set a target. Collatz and Milkyway are popular.

[Brucifer]

Thanks to both of you for taking the time to answer my questions.

I'll have to admit that the distributed.net folks get a bit laggard at times. Like the present read only database stuff, as that stops people from joining teams which isn't a good thing for engendering crunching support.

As for the current state of the boinc world, it isn't that impressive either. Basically you are saying that there are two reliable projects if you have certain gpu cards. I think overall that the whole dc thing is rather in the doldrums as there just doesn't seem to be much enthusiasm for stuff like there used to be. Sorta goes hand in hand with the significant slowdown in forum posting across the teams spectrum. A lot of apathy out there in dc land I fear.

[Nflight]

Your Welcome ObiWan..!!!

[Brucifer]

So I fired off boinc on collatz. Crunched out a WU on the gpu. However in the process I was reacquainted with why I don't like boinc, ie it decides to talk whenever it wants to... just gives you a bunch of buttons to click, but still only does what it wants to. It loaded up 4 processors, even though the default was set for 2.. Didn't want to send the completed units when the update button was pushed, etc. So I just removed it from the system,.... that showed it who owns the computer. :-) Don't need no steenking boinc....

edit: so aborted, cleaned out, re-installed, sucked new work units, it's following the defaults I wanted now, so it is off and running. Might even get a work unit or two completed at this rate. Have 2 gpu's on it, nothing fast, just the slow ones. Keeping the fast ones on rc5 for now at least anyway.

[vaughan]

...I think overall that the whole dc thing is rather in the doldrums as there just doesn't seem to be much enthusiasm for stuff like there used to be. Sorta goes hand in hand with the significant slowdown in forum posting across the teams spectrum. A lot of apathy out there in dc land I fear.

That is true. Recently I put some CPU Ghz on Poem@Home only to have the "inevitable" happen - the project ran out of work. It appeared to be quite reliable until I decided to crunch it In the official forum for the project when another user and myself posted that there was no work and the server gave http errors the admin eventually replied with words to the effect of "why is there a sudden increase in requests for work"? What two additional users broke his project! Sheesh.

I was stunned to say the least.

[Brucifer]

heh................ yup, that takes the cake Vaughan.

Makes you wonder what goes through the heads of these folks that put up a boinc project and then get ticked when they actually have to start holding up their end of the project... That part is the down side to boinc cause there are projects out there that shouldn't be there. Just a waste of everyones time.

[Nflight]

Brucifier, I also have run into the sse CPU driven work units on the Collatz Conjecture project, but with some manual labor I abort these work units and only run the GPU ones. Each time I download the 150 maximum workunits I always get a few CPU Wu's and abort them. Every BOINC project has some flaw to it, every Distributed Project must have flaws too, or we would not be human. Wait wait wait; is mitchell human or is he a god?

[vaughan]

OK whilst we are on the topic of BOINC I have one problem PC. This unit used to behave itself until BOINC Virtual Prairie cam out of hibernation last week. Its an Intel Q6600 running at stock 2.4GHz with Win XP 64 SP2 installed. On startup from booting it runs NPLB (4 instances) which is set for no additional work), Folding@Home on the nVidia 9800GT GPU and FAHSpy with RivaTuner set to make the GPU fan run at 100 percent, as well as Dimes and BOINC. Good old eOn starts but of course there is no work for it anymore so I close that application.

BOINC version is 6.10.18 and it persistently displays a message box:
Communicating with BOINC client. Please wait ...
Exit BOINC Manager or Cancel

How do I stop this annoying message from popping up and interrupting anything I try to do?