microFluids project - uFluids

[Lagu]

Gamer007

Thank you for joining this project. We need as many we can get and of cource a AMD x2 outputting 5 GHz ;)

Lagu :D

[AQUAJOE]

I joined the team and have an account but have not recieved an email with my code.

[vaughan]

That's the spirit guys. The more the merrier, hey :!:

Has anyone had any long running molecules to crunch in uFluids since Robert upgraded to evolver 4.02 :?:

Maybe its just my luck but I have a few bubble4a tasks that are estimated to take 12 hours on my xp3200+

On the P4 2.66GHz notebook these tasks are estimated at 48 hours :!: :shock:

[NeoGen]

all the latest ones I've had take around 14/15 hours on my XP 2000+

[Lagu]

Hej

I have 2 WUs whish have a limit set to 11.47.30 on my AMD 3200+ and 2 limited to 11.59. But I will see if it takes this time or if the task will be longer or shorter. Sometimes a limit is set but the task is done earlier if there not is any fault on my eyes.
I ran a benchmark yesterday because I don´t receive a work. The server got new information.

Measured Floating Point Speed: 2.032.53 million ops/ sec.
Measured Integer Speed: 3.867.97 million ops/sec.

I think this Benchmark should be a little better yet if I should close ZoneAlarm and Panda antivirus program.

I´m surpriced that Vaughan´s Intel need 48 hour. Thank God I have an AMD. There we can se the differense between Intel and AMDs arcitecture and philosophy.

Lagu :D

[Lagu]

Hej

Observations: The task who should be running in 11.47 hour was done after 7.04 hour. My eyes have not lurk me :D

I think the Intels 48 hour will be lesser than the limit as have been set.

Lagu ;)

[mitro]

I've put a machine on μFluids atleast for the time being.

[vaughan]

Hej Lagu,

I don't know if its an Intel versus AMD thing or just the fact that uFluids is an ALPHA project :!: but ...

bubble4a_45_1_90_0.764_15000_0.0001_0.382_4 took 4:15:41 (so far) not 48 hours and its Progress is currently 360% ...

I'll just let it crunch as long as it wants.

[Lagu]

Hej

I should watch uFluids ”Top participants” and check after WU´s as I have done and these who was pending. I found 8 tasks worth 460 points who was pending. I have computer-id 24. I have run 40 tasks so far.
Clicked on computer-id to see whish other was running the same task. There was other, one with computer-ID 11. Clicked on this a new page opened and the owner is Jeffrey Robinson. He has done 428 work units. How the hell has he done that I think. Clicking on that number a page opened who show all his 428 works had erroded out as far he begins to crunch them. No points there. He have run the same work unit as me whish now is pending. In the stats page that guy have 3.697.15 in “Total credits” and 70.71 in “Recent average credits”. It is a little smaller than me and Vaughan together! I wonder where/how he has got all credits. All 428 WU´s, use zero CPU-usage.
I also noticed computer-ID 27 had got more points for the same WU as me:
Computer-ID 24: Claimed credits=94.91, Granted credits=94.91 CPU-time=27.480.91
Computer-ID 27: Claimed credits=142.59, Granted credits=94.91, CPU-time=31.436.53.

Have he got better claimed credits because he have crunch this WU longer than me?
I think he has over clocked his CPU because of higher Integer speed. But this seems not to be to any advantage because he use longer time to crunch this WU.

I don’t know how they calculate the points but it seems not to be fair. It should be much better if we all got the same points and not need to wait for other to done the same work units. If any stop runs this project any can lose CPU-hours to no use. If I run a task and got say 100 points immediately and the other 100 if they should run the task to the end.

Have you any comments to above?

Lagu

[NeoGen]

The computer id 11 problem was already reported in the uFluids forums, and a few others like it were also found. It's not a problem for us if they turn back errors on the workunits because the boinc scheduler resends that same workunit to someone else.
And the differences in claimed credit for more or less equal machines sometimes have to do with different boinc versions. I heard that boinc v5 claims more credits per workunit than v4. In this project it could be good for us to use boinc v5, as the lowest claimed credit seems to always be the chosen value for granting. If we claim more, we got a better chance to get more.