SETI@home News Vol. 115 (12/09/2010)

[TheTross]

Toxic, don't forget that each GPU task for Einstein needs one CPU core as well... not sure if that makes it easier or harder for your decision making though

Speaking of Einstein, all downloads have finally finished and I am proud to report that a measly 94 WUs currently occupy ~3.7GB of disk space...

 

[ToxicTBag]

Aye i noticed that TT, it's also using the whole gpu at the moment but its not drawing much power temps on gpu dont go above 55 nice change from mid 70's on seti.
Am going to pursue the app_info thing on einstein see what trouble i can get into. As far as milkyway goes i think a second hand ati might help...shop time

 

[BaBa]

Ok so in my sig EAH = Einstein , SAH = Seti any idea what it is for collatz?

Not a great project but i use it as my Milkyway ATI backup and MW is down so much ive got quite a total going.

 

[edgedemon]

Yep, you just need to alter the y value. Y = 10 puts the Rank & Credit text right at the top, which leaves enough room at the bottom for a fourth project, although setting y to 13 aligns everything properly.

Think I will need Mr Halz help as 'madmac' gets in the way if I move the text up to get EAH in..

 

[TheTross]

Am going to pursue the app_info thing on einstein see what trouble i can get into.

In that case I strongly recommend that you disable network activity and back the entire BOINC folder up before doing that - I tried writing an app_info for Einstein a few days ago and it borked my client_state.xml, which caused BOINC to lose track of all my Einstein WUs. Thankfully I got them back as resends after I'd detached and reattached, but you might not be so lucky... You'll need to include references to the GPU app and at least the Global Correlations S5 Search #1, if not all available CPU apps. Good luck though, and if you manage to get a working one then please spill the beans as I'd really, really want to use it

Ok so in my sig EAH = Einstein , SAH = Seti any idea what it is for collatz?

Not a great project but i use it as my Milkyway ATI backup and MW is down so much ive got quite a total going.

Unfortunately you won't be able to use Collatz in your sig as the stats that the sig uses are taken from Berserker's stats site (www.teamocuk.co.uk), which doesn't include Collatz.

 

[BaBa]

Thanks for the relpy i will just have to add my rarther poor LHC total into the sig(wish they would give us more work).

If anyone gets Einstein to run more than one task per GPU without reseting please let us all know how it was done, 12% GPU activity really is a waste.

 

[ToxicTBag]

If anyone gets Einstein to run more than one task per GPU without reseting please let us all know how it was done, 12% GPU activity realy is a waste.

I am not that knowledgeable but have a basic grasp of the app_info so if anything good comes out of it from my part it will be by trial and error rather than expertise.
This place is bristling with brain power am sure an einstein multiple instances gpu app is not impossible....just need to get people interested

 

[ReaVerUK]

Just a silly thought could you not make the app_info file read only stopping the count being altered when it communicates with the server?
Thats probly to simple to work but it may be worth a go. Just don't hold me responsible if anything goes wrong

 

[ToxicTBag]

Lol will be sure to come a looking for you when i blow it up Mr Reaver sah.
Why not just suspend network activity on einstein (assuming you have say 24 hours work cached) just go routing through the client_state.xml with the search feature in notepad and change every <count>0.500000<count> to .330000 etc saves all the time and effort of doing an app_info and its working for me....Einstein-a-go-go



Next to no effort and voila increased rac

Ok that made it look like it was going to take forever having more than one run at a time, but if you work it out it's about 10 mins approx extra per task and your gpu is still running cool

 

[TheTross]

Three Einstein WUs running simultaneously, nice! I can only get two to run on my GTS250

Also, the Astropulse WUs you have are legendary . Must be something going on though, as every AP WU I've crunched is completed in less than 10 hours...

 

[ToxicTBag]

There's a shedload of those darn astropulse, they're probably trying to p me off well it's not going to work i will crunch every last one of em.
Am just having a dry run on q6600 and if it works out will try it on t'others, think i have it just about right now



The Einstein gpu tasks do need a minimum of .25 cuda so i cant think for the life of me why i can't get 4 running....that's next job.

Be interesting to get some thoughts on this, apart from seti which will allways be my 1st what other crunching project best utilises fermi/nvidia cards at the moment.....i spy a gap on bottom of my sig

 

[edgedemon]

Lol will be sure to come a looking for you when i blow it up Mr Reaver sah.
Why not just suspend network activity on einstein (assuming you have say 24 hours work cached) just go routing through the client_state.xml with the search feature in notepad and change every <count>0.500000<count> to .330000 etc saves all the time and effort of doing an app_info and its working for me....Einstein-a-go-go



Next to no effort and voila increased rac

Ok that made it look like it was going to take forever having more than one run at a time, but if you work it out it's about 10 mins approx extra per task and your gpu is still running cool

Is this with a machine with one card or has it got multiple cards Toxic? When I tried, I had 3 per card, but had to divide my quad up to service the cards and it wasn't worth it in the end - they were taking 2-3 hrs per task rather than 1 1/4

 

[ToxicTBag]

Morning Mr Edge sah, its a q6600 with 1x gtx470 works fine so far fingers crossed

 

[edgedemon]

Ahh, that explains it, if you try that with multiple cards, you will get <epic fail> as the cpu will become the bottleneck...

 

[ToxicTBag]

Oh poop ....bang goes that idea