9th Annual BOINC Pentathlon 2018 Strategy & Preparation

Skillz said:
Same way I recruit people to the Christmas Race and the Formula BOINC Sprints.

I'll make a post in General Mayhem and see where it goes.
Click to expand...

Should post in some of the other forums as well. Lots of people not in Genmay. I think i managed to get a few people from a post in the AMD forum while you were cruising and boozing
 
Yeah I did that for the Christmas race. I posted in the Intel and AMD forums, multiprocessing forum, genmay and I think Montu made us a front page news article that he was able to slip in a shout out to the WCG team
 
In about 3 hours from now, the first marathon project will be announced on Apr 30, 00:00 UTC, if I read this correctly. Its 5pm/8pm (PT/ET) here in US.

I'm almost finishing up other WUs in progress and making room for this.

upload_2018-4-29_13-49-21.png
 
There is a limit of 6 tasks per thread. Remember to disable your network or block the host to prevent upload until the event begins.

There are two ways to overcome the above limitation.
1) Create and add cc_config.xml file to C:\ProgramData\BOINC (for Win). Sorry I don't know much about linux. In doing this, it seems that there is a hard limit of 384 tasks per rig. Anyone else seeing this hard limit? After you have downloaded the tasks, remember to switch the number of cpus back to your original setting.

In below example, I set 64 threads = 64 threads x 6 tasks per thread = 384 task limits.
<cc_config>
<options>
<start_delay>60</start_delay>
<ncpus>64</ncpus>
</options>
</cc_config>

2) Install virtualbox. Run as many virtual machines you want but limited to your RAM and diskspace. When you download BOINC, you are given a choice to download with or without virtualbox. Hmm, I don't have any good tutorial for setting up virtualbox or maybe see this youtube video for Win7?
 
Last edited:
option 3) I go via the multiple client installation bases on source code from boinc ... each client gets it own folder and use a different port to communicate. Worked well last year with the WOW! and GPUs (and CPUs). Right now setting up a refreshed Ryzen-Ubuntu 17.10 this way and try to document the steps/scripts needed
 
pututu, it would be better if you put a startup delay in that cc_config. Otherwise you may make the host unusable before you get the chance to edit it back to normal.
 
See green font. Normally I start as normal and then increasing the CPU count in one or two steps depending on the target CPU count that I want to achieve.
 
On a side note, would it be better to create a new thread consolidating all the "tricks" (not cheating) for each project? No strategy discussion, just for newbie to get off the ground? Maybe the thread can be called "Marathon CPU project: NumberFields@home. Useful tricks & setup". I'll start this one first and make the first post (post #1) a living document as we get more useful tips from all of us here?

I reckon the marathon project will be useful for new recruits since it will last for the next 18 days or so
 
Feel free to make one. It cannot hurt. Just post a quip with a link back to the main Pentathlon thread so people know there is more to it. I think that will help for any new recruits whom want an easy set and forget participation.
 
Oh...and I'm running SETI on the GPU's so that they have long due dates. I'm expecting Rosetta to be the 3 day sprint. I would be shocked if it weren't. So, here is where we really need to know everyones thoughts on strategy. Do we focus mostly on the Marathon or do we want to focus everything on one of the other events to try and get top 3 in one of the categories this year? fastgeek, I haven't seen you chime in yet since announcement. What are your thoughts since you tend to bring a lot of heat to the CPU table? Obviously we will roll with the punches..but would like to know what everyone has in mind.
 
first Ryzen-1700x bunker is filled,
deadlines around May 7;

three clients on the same hardware; one active crunching; the other two disabled

State: All (267) · In progress (267) · Validation pending (0) · Validation inconclusive (0) · Valid (0) · Invalid (0) · Error (0)
Application: All (267) · Get Decic Fields (25) · Minimum Discriminant Septics (242)

Those "Minimum" have a forecase of 9 hours runtime; "Get Deric" around 7 hours.

I will let few first WU through as its the first time for me on this project; and want to test of the finish well.
 
So, seeing as numberfields reduced the number of WU per thread from 24 to 6 last year due to a server load, when all the teams dump their bunker, aren't we just going to crash the server? lol
 
Possibly... that is why I would recommend bunkering 6 or 7 days worth of work just in case. I personally will only have a minor dump as I have not had the proper time at home to do all the things I wanted to get done.
 
EXT64 said:
So, seeing as numberfields reduced the number of WU per thread from 24 to 6 last year due to a server load, when all the teams dump their bunker, aren't we just going to crash the server? lol
Click to expand...

I was reading the site and wondering that myself, but doesn't it say that was due to bad hardware / HDD's? surely he has fixed those by now right???:ROFLMAO:
Ok i guess the second part of the issue was "10,000 work units that all timed out at the same time" don't know about you guys but the 900 WU's i grabbed are all deadline May 6th so should be fine to bunker till the project start without time out.
 
Heavy hitter? That's cute. ;) (that's a clicky clicky BTW)

That's what I managed to wrangle up to max WU counts today. Bit of a PITFA; but done. Will see about a few more tomorrow. Some of those boxes will exhaust their WU store in a day or two... so might try the VM thing if I have time.
 
It's nice to have a CPU project I can do some damage on vs. the GPU big boys here. :D Unfortunately I'm down four 2P 5118 boxes vs. last time around... so wouldn't be able to give the CE/GC pools quite the run for their proverbial money. :LOL: (Just in terms of my RAC vs. the RAC of their top "single user" account; obviously their combined might far surpasses me)

For giggles. One box was being stubborn about loading up; but it's one of the weakest and was sick of fighting with it, so not too worried about it.

State: All (6483) · In progress (6349) · Validation pending (0) · Validation inconclusive (0) · Valid (134) · Invalid (0) · Error (0)
 
I keep getting the message that the project is out of work...everyone hammering away at bunkering sapping it dry too quick...lol
 
Seems I forgot to set the host file on the 4P-E7 box yesterday; so it has been reporting WUs for ~24 hours. Verified it still had 384 WU's on deck and fixed that. :p Guess I should make sure nothing else slipped through the cracks! :oops:
 
fastgeek said:
Seems I forgot to set the host file on the 4P-E7 box yesterday; so it has been reporting WUs for ~24 hours. Verified it still had 384 WU's on deck and fixed that. :p Guess I should make sure nothing else slipped through the cracks! :oops:
Click to expand...
nice strategy; hiding in plain sight ... give "the others" the feeling that this is what we have with a bit of a leak; then knock them down ... :punch:

my bunker is only 10% of yours; not much of a knock ... :dummy1:
 
Well..I have a small bunker brewing but nothing impressive. I also have a few VM's and a few borgs reporting some work too. Some things just aren't worth the extra effort...lol. I did add another quad core laptop tonight. Was sent home a few Core 2 Quads that used to be DVR's, but I think I've tapped my limit for adding stuff. So, those 2 boxes will probably sit cold.
 
We should be able to perform quite well in the marathon given the current numbers

1 SETI.USA 233,812,464 109
2 L'Alliance Francophone 141,038,432 211
3 Czech National Team 91,653,112 105
4 SETI.Germany 85,845,184 139
5 [H]ard|OCP 82,349,960 35
6 BOINC@Poland 69,133,400 129
7 LITOMYSL Boinc Team Czech Republic 56,041,452 8
8 UK BOINC Team 40,747,616 40
9 BOINC.Italy 31,740,392 86
10 USA 31,214,262 89
11 TeAm AnandTech 30,336,230 23
12 Overclock.net 25,417,000 41
13 Crunching@EVGA 24,689,554 36
 
Well crap! Turns out it wasn't just one box. Rather it was six of the damn things; including a Linux box from hell. 674 WUs snuck out in the last ~24 hours. They're all stuffed and blocked now. Oh, and the count is up to 19 boxes now. Have more systems that might come online in the next day or two if I can find space and power. (Couple 4P Ivies, 2P highend-ish Haswell and a passel of 2P Skylakes; the last would be pretty temporary though.)

State of the Buffer...

State: All (7889) · In progress (7212) · Validation pending (0) · Validation inconclusive (0) · Valid (674) · Invalid (0) · Error (3)
 
Based on terrible, unscientific, SWAG math... think I might be sitting on ~3.5M points as is.

And, Skillz, that might work for a moment; but once the reported WUs dry up I'd rather suspect folks would get wise. ;)
 
Universe@home was chosen for the Swimming event for 5th-12th. I will be focusing more on Numberfields. fastgeek, both of these projects have badges....lol.
 
Some of the competition is reporting this if anyone can confirm.

Swimming project is Universe@Home. If you can get them, BHDB give much better credit than BHSpin. Also have to watch out for BHSpin tasks that run for hours and hours longer than normal for no credit. The tasks are all around the same length, so abort anything that runs abnormally long. It's just a bad task.


For reference on my 1950x, the BHDB tasks ran at less than half the time of BHSpin for the same 666.67 credit. So I would select BHDB app and the extra checkbox about sending task from other apps if there is no work.

Edit: There is a task limit here as well. I think it was 256 on my 32 thread machines. Not sure if that was 8 per thread or 256 per host.
Click to expand...
 
Hard decision to make. Split half on Numberfields or Universe? If we want a shot at the medal podium on a project and disregard overall team position, need to find which event to run 100% or just have fun first and see where this leads us to?
 
I will take a chance on this one and say the best bet if focusing would be Numberfields.
Here are my reasons:
The applications are more refined and less buggy.
The project is the whole event so after one event is done we aren't essentially playing keep up.
Anyone on our team wanting to set and forget will most likely be on this project anyways.

If the team would just like an overall better standing rather than taking the chance of seeing some metal in a particular category, then sure divy the resources. That is kinda how we went last few years.
 
You must log in or register to reply here.
Back
Top