BIG work units..

[veebee]

I didn't actually see the download size (I know Wng was inyerested in that), but I am beginning to get worried about the actuyal time to CRUNCH these monsters !


after 12 odd hours, the WU is at 10.x% ..... so a crunch time of 120 hours "seems" to be where its headed..(although I know, WU's can fly through the last "whatever" percentage in some projects)..


Anyone else finished any of the D-waves Adiabatic QUantum Algorithm 2.15 WU's ?? if so, are they fixed credit ? and how much per WU..?

Cheers

vb

[Agnew]

Haven't had any of the 2.15 wu here.  The 2.35 wu's that haven't "errored out" seem to be paying in the range of 6,389.41 (71.6k sec) to 13,281.94 (103.8k sec's) on my i7, if it's of any 'reference' help to you.

I seem to be crashing a few,,, most at under 7%, then they seem stable until in the 90% range,,, It's quite annoying when they die with less than 30 minutes to go, so I've backed off to 2.8GHz to see if thing improve :) 

[veebee]

yes, the 2.35 Multi - Thread_ (MT) WUS are great... good credit , especially on WIN 7 64 bit !!!

[WikiWill]

yes, the 2.35 Multi - Thread_ (MT) WUS are great... good credit , especially on WIN 7 64 bit !!!

Well, you've convinced me... I just slapped win7 x64 onto my ubuntu box and will give it a go.

[veebee]

Well, you've convinced me... I just slapped win7 x64 onto my ubuntu box and will give it a go.

have a look at the last few days on HERE ... climbing beautifully since swapping my i7 and 1 q600 over to win7 x64 and aqua (they also both run gpugrid)

[BF]

That's a beautiful chart.. Up, up and awayyyyyy! 

[veebee]

My quads happened to do a "160 -6M" WU (over night I guess) and it payed GREAT credit !

now I have 2 x 200-8M WU's in each of my Win7x64 machines - can't WAIT to se what they pay...

EDIT: just checked...it was the ONLY 160-6M wu I got...

[kashi]

PoorBoy has reported his first completed 200-8M has granted 27,911 credits.
His i7 is at 3.6 GHz and he is using the 8 of 9 cores method. I am at 2.67 GHz and am using the 7 of 8 cores method for faster ATI performance, so my AQUA processing times will be 30-40% greater than his. My second 200-8M task is going to take at least 2 hours longer than my first which was granted 23,776.

[AusTerror]

So are the work units working correctly now?

I connected all my PCs to this project a couple of months ago when the multi threaded and GPU stuff was new.  All I got was a bunch of WUs trying to use all CPUs at the same time and basically getting no where.  then they would crash out and I would get zero credit.

So I went back to Nqueens.  Is the project more stable now?

[Furlozza]

Yes.

TheGnat is on 200 series Wus and they take... like fur ages.... but when mixed in with time off and on for MW turn in a decent return and no crashes to date. iGnatious gets through the same Wus in about 24 hrs and last one returned..... hm over to his box...brb... 26050 crs for the roughly 26 hrs actual time on all four cores, which works out at roughly 250 crs per hr.

I am sorta waiting with bated breath for when I finish or rather TheGnat finishes one of his to do a calc and see how it compares.

[kashi]

AQUA has been very stable for me. The current 200-8M tasks take 18-19 hours each on my computer running on 7 cores. AQUA multi threaded tasks still use all available cores if run in the default state.

It is possible to specify the number of cores to allocate to AQUA by reducing your BOINC CPUs in "Computing preferences" before you download AQUA tasks. The AQUA tasks will then run with the number of cores you had set before you downloaded the tasks. So I switch to 7 cores by setting 87.5%, update project, then download AQUA tasks. Then I switch back to 8 cores and update project again. This means AQUA runs on 7 cores and leaves one core free to support MilkyWay ATI. You could possibly do the same thing by using the BOINC Manager local preferences file, but I do it the online way.

I can also use 8 out of 9 cores for AQUA by setting <ncpus>9</ncpus> option in a cc_config.xml file but 7 of 8 cores gives better overall efficiency for my needs. It is necessary to suspend the second AQUA task if you configure AQUA tasks to run on less than the number of available cores otherwise 2 AQUA tasks will run at the same time. I think that may also happen sometimes if you use all the available cores to run AQUA, not sure about that though because I have always run AQUA with 7 of 8 or 8 of 9 cores.

[AusTerror]

Hmm... the problem I had was that each WU (in my case 2) would be using 2 (two) cores.  So thats 2 x 2 = 4.  When my understanding is that 1 WU should run at any 1 time and use 2 cores.

Anyway I'm busy with GPUgrid and Nqueens for now :)

[kashi]

Yes I believe in the past each AQUA task used less than the total number of cores. But currently a task uses all available CPU cores. I think it should now only run 1 task at a time when it is using all available cores, not certain. Veebee reported that after he configured AQUA to run on 7 cores that it started running 2 tasks at once, so from that it seems that it ran only 1 task at a time when he was using the default AQUA configuration which used all 8 cores.