Work Available

[Mike Mitchell]

Simap usually releases work units at the beginning of the month but today they seem to have released a large amount without warning. So top if you are into the project.

[CairnsByTheSea]

Hello All   

There is still some work at SIMAP available for this month - grab some quick!

:aus1:

:sg1:

[kashi]

Yes I've been doing some SIMAP for a few days. :)

CPU was overheating and has now throttled down to 1.6 GHz, something is awry but it's still limping along. Doesn't affect ATI crunching fortunately. I've bought a Mega Shadow cooler and some Shin Etsu goo and will whack them on tomorrow.

[Dingo]

Work available:

New Workunits in the morning (UTC) of March 31
New Workunits: the calculation of similarities and features of approx. 1,200,000 new sequences, that were imported from protein databases and metagenomes into SIMAP until the end of March, will start in the morning (UTC) of March 31. 30 Mar 2010 5:53:38 UTC · Comment

[dyeman]

Looks like a big month for work in SIMAP - new work still coming strongly.  Team is at position 32 and struggling to maintain even that so please feel free to grab some work while it's there..

[LawryB]

On it.  watch it dry out!!!!!

[dyeman]

Thanks LawryB !

[kashi]

Yeah heaps of work this time, I've added 30K to my total and the work is still flowing. There are still 6 of 10 batches of metagenomes that haven't been distributed yet, so it should go for at least a week more. I decided to try and increase my total on this project while there was much more work available than usual, plus it is another project "in the red".

[veebee]

I hadn't crunched Simap for a cpl years.... since there is so much work available I have put all 16 "cores" of the i7's on it until the work runs out.

... pretty big downloads though ..... luckily I have "unlimited" downloads because of the muck ups lately with my ADSL.

[clownius]

Ok ill fall into peer pressure and throw in 8 i7 cores

Edit: Make that 7 cores lol GPU grid is using one core itself due to the 4 WU's at 0.32 CPU's each.

[kashi]

Yeh, need to allocate sufficient CPU resources for full speed GPU processing. When I had a single GPU I only needed to leave a core free for MilkyWay or Collatz. Now with a single and a dual GPU I need to allocate 2 CPU cores for Collatz and 3 CPU cores for MilkyWay to avoid a slowdown.

That's the good thing about DNETC ATI, the one task is spread over all 3 GPU cores and suffers almost no slowdown when crunching CPU projects on all 8 CPU cores. DNETC tasks run on dual core GPUs freeze every few hours or so but the watchdog timer ends frozen tasks after 20 minutes so DNETC still has greater output than the other 2 ATI projects.

It's strange with Nehalem class CPUs though, on quite a few CPU projects due to contention issues the throughput is very similar for 6, 7 or 8 cores. When MilkyWay ATI is sucking 3 CPU cores I sometimes run a 2 core VM on AlmereGrid. Running CPU projects on less than the full number of cores reduces the CPU temp quite a bit, I will definitely be making use of this during the heat of summer.

[dyeman]

Thanks guys

@veebee the real killer bandwidth-wise is almeregrid.  Downloads are nearly 10 MB and WU runs around 35 minutes.  By my calculations that's about 12 GB download per core per month.  I couldn't work out why my broadband usage suddenly doubled - eventually worked out that almeregrid was the culprit.  Still, all the effort has been worthwhile - 1.3 cobblestones per WU soon adds up and BOINC@AUSTRALIA has already hit 5,000  biggrin

[clownius]

Hmm well so far so good.  Each Core is churning out a WU in around 1hr 15mins and claiming around 24 credits.  Getting around 20credits.  Looks like some of the apps claim really low creits and drag my claim down a bit with them.  Wouldn't call this a fast credit project but ill do what i can for a while.  Haven't crunched it in the past so i will give it some clown love.

[kashi]

Good one Clownius. 

I'm running simap v5.12 tasks on 8 cores at default 2.67GHz claiming about 25 and being granted about 21. Tasks are taking about 67 minutes.

It doesn't seem to make any difference what my wingman claims, I think SIMAP must not use the BOINC benchmark based credit system but have fixed credit calculated somehow according to the estimated amount of work in each task.

So yes credit is not generous by any means but it's better credit and more stable than some other "in the red" CPU projects I've been crunching recently, so even 19.4 cr/hr with stability makes a pleasant change. I had a nice batch of hmmer v5.09 tasks a few days ago, they pay a little better, but unfortunately there is always relatively few of them, usually I don't get any at all.

[veebee]

that's strange, I seem to be getting slightly MORE than what I claim.

two i7-920's (1 @ 3.36, 1 @ 3.15) On the faster machine, on a cpl WU's on the first page, it claimed around 21 and got over 29 per WU = but average claim is around 21 and getting around 23.2. WU's are taking about 53 minutes (= 26 / hr/ core).

the second machine however, is not quite as good, but the difference in core speed is obviously the "problem" - taking around 58 minutes, claiming about 19 and getting about 21.3. Thats about 22 cr/ hr/ core.