TAS-member thread for September: µFluids & TANPAKU

[networkman]

I like the beer idea - I was already planning to do that anyway this evening.. having some friends over for a small LAN party to play SWAT4 Co-op. Check out the video review - it's pretty good.

 

[caferace]

In my sometimes annoying, yet oft amusing way I was seeing what this damned bplugz thing was all about. You math whizz'es may be intrigued:

http://www.ufluids.net/download/0/bplugz1_13_4.8_1_0_0_0_0.fe (a simple text-type file)

-jim

 

[petrusbroder]

Originally posted by: caferace
In my sometimes annoying, yet oft amusing way I was seeing what this damned bplugz thing was all about. You math whizz'es may be intrigued:

http://www.ufluids.net/download/0/bplugz1_13_4.8_1_0_0_0_0.fe (a simple text-type file)

-jim

I have not laughed this hard for a very long time ...
From the text-file:

Computes quarter of a plug in a squashed tube for Aspect Ratios > 1 and volume = 0

...plugz1 - a plug in one (1) end (end = z)
Where do you have a "squashed tube"?

The math is elegant, though ....

The fun part is that they calculate this for a gravity of "0", i.e. in free fall in space. Probably they also have WUs with different amounts of gravity ...

I have the impression that there are very few solutions to this math and that is why the crunching time is so short - i.e. the model hits one of the constraints. I may be wrong though - I have looked at the math only for a short time and I do not understand all of their thinking.

 

[petrusbroder]

OK Team- and TAS mates: here are the stats for the first 2½ days of crunching.

Stats for µFluids as of September 3,2006 09:15 UTC:

New members: Very welcome, CupCak3!
TAS-members who have joined since Sept. 1: Strates, CupCak3, Lizardth (Liz), Jened, Fullmetal Chocobo


TAS' stats
___ Credits _____ RAC ______ Credits ________ RAC _____ Production
___ Sept 1 _____ Sept 1 _____ Sept 3 ________ Sept 3 _____ in Sept__
160 885.63 ___ 2 676.04 ___ 169 112.46 ____ 2 788.08 _____ 8 226.83

Comments:
12 TAS'ers are crunching as of now.
After two full days of crunching TAS has produced 8 226 cobblestones. Considering the problems with µFluids (the very small WUs) this is not too bad, as quite a few crunchers have hit the WU-limit (200WUs/day) and do not crunch full time... No prognosis can be made today.


The team-stats:

Sorted according to credits:
Rank ___ Members ______ RAC _______ Credits _____ credits added in July ____ Name
1 _______ 44 ______ 5 392.38 ____ 1 390 380.96 ________ 21 502.09 ___________ BOINC.BE
2 _______ 76 ______ 2 712.55 ____ 1 139 821.67 ________ 10 749.61 ___________ Dutch Power Cows
3 ______ 180 ______ 3 880.28 _____ 822 306.39 ________ 17 967.81 ___________ BOINC Synergy
4 _______ 32 ______ 4 446.72 _____ 456 692.64 ________ 17 009.48 ___________ Team Anandtech
5 ______ 123 ______ 1 593.02 _____ 407 669.08 _________ 4 206.18 ___________ Czech National team


Sorted according to RAC:
Rank __ Members _____ RAC ________ Credits ______ Name
1 ______ 44 ______ 5 586.35 ___ 1 388 773.48 ____ BOINC.BE
2 ______ 32 ______ 4 271.78 _____ 452 150.32 ____ Team Anandtech
3 ______ 38 ______ 2 922.64 _____ 148 613.06 ____ AMD Users


TAS-member's stats: Climbers:
Change in
Team Rank ___ Name[/color]
+7 _______ petrusbroder
+6 _______ cooper76
+1 _______ BlackMountainCow
+1 _______ Strates
+1 _______ Liz


Edit: The numbers in this post were off by "one cruncher" due to a error in my spreadsheet.
I appologize to TAS and TA for having published errornous numbers. This was of course not my intent.
I also appologise to my fellow cruncher caferace whos numbers were excluded by the error mentioned above. I will strive to proof my posts much better in the future.

 

[CupCak3]

thanks for the stats peter!!!

ufluids is back up!!! they seemed to have fixed the small WU issue. I'm on my first WU and it looks like it will take about half an hour with one core of a 3600+ @2.72 to complete

 

[Freewolf]

"These Tiny WUs should be gone within a day. We will have 1 hr / 2hr size WU up and coming."
Posted by the man in charge on their forums

 

[petrusbroder]

Yeah, I got some upbend...- and nonlinear...-WUs, they are good for at least an hour. I have not got the bplugz1...-WUs for the passed 2 hours.

 

[Jened]

thanks - all set up

 

[strates]

I had a P3 -900 I was trying to do ufluids with, but its only got 128megs of ram, and the client said I need 512 or something. The weird thing is that my p3-600 only has 256. so I don't know whats going on. Craziness.

 

[BlackMountainCow]

IMPORTANT

If you run TANPAKU as a backup project of µfluids, make sure you realize that is has NO CHECKPOINTS!!! So, if TANPAKU switches to µfluids at 90% completion of a WU, you'll lose those 90% and the WU will start from scratch later on. At this monet, TANPAKU is a "run only" project, until checkpoints are implemented!

I'm doing this because I just saw 7,5 hours of crunching going down the drain because I didn't know about that

 

[CupCak3]

Originally posted by: BlackMountainCow
IMPORTANT

If you run TANPAKU as a backup project of µfluids, make sure you realize that is has NO CHECKPOINTS!!! So, if TANPAKU switches to µfluids at 90% completion of a WU, you'll lose those 90% and the WU will start from scratch later on. At this monet, TANPAKU is a "run only" project, until checkpoints are implemented!

I'm doing this because I just saw 7,5 hours of crunching going down the drain because I didn't know about that

mine have been restarting just fine.

9/4/2006 8:36:19 AM|Project TANPAKU|Pausing task pep4_147_27_8_4_1 (removed from memory)
9/4/2006 8:36:19 AM|uFluids|Starting task upbend2_jon_120_9_5.45999999999993_0_0_0_0_0 using evolver version 410
.
.
.
9/4/2006 9:10:08 AM|Project TANPAKU|Restarting task pep4_147_27_8_4_1 using BD version 504
9/4/2006 9:12:10 AM|Project TANPAKU|Computation for task pep4_147_27_8_4_1 finished
9/4/2006 9:12:10 AM|Project TANPAKU|Starting task pep4_129_26_20_3_2 using BD version 504

pep4_147_27_8_4_1 was at 98% or so when it was origionally paused

 

[BlackMountainCow]

You sure? I intentionally tested it with a couple of WUs and it never worked for me

 

[CupCak3]

i'm using BOINC manager 5.4.11 maybe that makes a difference?

i'm pretty new to BOINC so i'm probably not going to be able to be of much help.

i recieved full credit also:

2725785 5751 4 Sep 2006 9:36:09 UTC 4 Sep 2006 13:41:08 UTC Over Success Done 3,887.55 15.65 49.25

though i still dont' fully understand the claimed credit and the granted credit.

 

[petrusbroder]

Hmmm, I have had no such problems, I have delivered some 10 TANPAKU WUs all crunched as a beckup projects and the crunching times fit ... no loss of work AFAIKS. OTOH: I have had a load of short WUs - some 400 - 900 seconds of crunching - so that may be a reason. Longer WUs have completed without loss of work too ... :Q

 

[Ken g6]

Originally posted by: CupCak3
though i still dont' fully understand the claimed credit and the granted credit.

It's a challenging system, but the idea is fairly simple. Every time you download a TANPAKU WU, you're in a race to return it. The first person to return that WU completed gets granted credit equal to their claimed credit. The rest of the people also get granted credit equal to the race winner's claimed credit.

Why they did this I don't know, but it poses an interesting challenge. Those who get high credit, like me with my Athlon XP, Windows XP, and optimized client, want to win the race. For running TANPAKU, I've set my computer to request 0.02 days (about 1/2 hour) of work. That way I get started on my WU shortly after downloading it. So far I've won a lot of my races. But my computer isn't the fastest, so it's harder to win the longer the WUs are.

If you're running Linux, on a fast machine (especially a P4), without the optimized client, the rest of us would appreciate it if you increase your buffer to 1 day or so. You're also more likely to get more credit this way.

 

[Philippart]

i set boinc to leave the application in memory, so no work for TANPAKU is lost.

 

[Rattledagger]

Originally posted by: BlackMountainCow
If you run TANPAKU as a backup project of µfluids, make sure you realize that is has NO CHECKPOINTS!!! So, if TANPAKU switches to µfluids at 90% completion of a WU, you'll lose those 90% and the WU will start from scratch later on. At this monet, TANPAKU is a "run only" project, until checkpoints are implemented!

I'm doing this because I just saw 7,5 hours of crunching going down the drain because I didn't know about that

As long as you don't shut-down the BOINC-client, this shouldn't be a problem... But of course, if you're only running for some hours before stopping/rebooting or whatever, you'll never make any progress...

My recommendation is to always select "leave in memory", this way you'll not lose the progress since last checkpoint even if switches to something else, and leaving in memory shouldn't normally be a problem, since using a little more of the pagefile shouldn't normally be a problem...

But, even if you've not selected "leave in memory", the BOINC-client will keep application in memory until it's checkpoined once, for projects that doesn't checkpoint, this means until wu finished... It should work this way in BOINC-client v5.4.9 and later, if it's not there's a bug somewhere. For older BOINC-clients it's possible it works, but for some it works until next benchmark, and in some it don't work at all... So, together with various bug-fixes, if you're running older BOINC-clients my recommendation is to upgrade.

 

[petrusbroder]

Hey TA-mates! This is good:

RAC from µFluids stats:
1. BOINC.BE: 5 170.54
2. TeAm Anandtech: 5 124.94 <--- up from 4 495 yesterday!

Soon we will - again - be the bigest producer in the world! :thumbsup:

 

[BlackMountainCow]

:evil: ... one TeAm, to rule them all, one TeAm to find, one TeAm to RAC them all, an in the stats bind them... :evil:

 

[caferace]

My comps are in super-ramp mode right now. The RAC is climbing at close to 20 per hr. Gotta love it. If I have to displace someone for the #1 spot in RAC for uFluids, I'm glad it's at least another TA cruncher. :thumbsup:

-jim

 

[petrusbroder]

Update:

RAC from µFluids stats:
1. TeAm Anandtech: 5 307.11 <--- up from 4 495 (Sept. 5)!
2. BOINC.BE: 5 256.56

Originally posted by: BlackMountainCow
... one TeAm to RAC them all,...

 

[Stalin PoG]

Concerning TAS. What is Boinc month?

 

[BlackMountainCow]

Originally posted by: Stalin PoG
Concerning TAS. What is Boinc month?

Chose any two BOINC projects you like and crunch them for 30 days.

 

[caferace]

uFluids is out of work again. :|
Hopefully, they'll get some going soon...

edit: Hmmm. One of my boxes just got work, while the fastest box is reporting this:

9/7/2006 5:31:40 PM|uFluids|Message from server: No work sent
9/7/2006 5:31:40 PM|uFluids|Message from server: (there was work but it was committed to other platforms)

That's annoying.

-jim

 

[CupCak3]

Originally posted by: caferace
uFluids is out of work again. :|
Hopefully, they'll get some going soon...

edit: Hmmm. One of my boxes just got work, while the fastest box is reporting this:

9/7/2006 5:31:40 PM|uFluids|Message from server: No work sent
9/7/2006 5:31:40 PM|uFluids|Message from server: (there was work but it was committed to other platforms)

That's annoying.

-jim

still none for my box