8th Annual BOINC Pentathlon by SETI.Germany

For whatever reason, my quad E7-8867 box won't max out all cores if HT is turned on. I've seen this happen in the past and not sure WTF is going on. Sometimes it would run at 100% and then drop down to some other number and stay there. It's not thermal, power, RAM or HDD space; all checked out fine. Sometimes I wonder if it's Server 2012 R2 related. Am tempted to upgrade it to 2016 or switch it to Ubuntu; but that would mean losing a lot of work units. Since it wants to sit at 70%, I've gone the conservative route and set BOINC to 89 CPUs.

Added a 4P E5-4640 v4 and, to the best of my knowledge, have all 96 cores humming along at 100%. :)

There is a NUMA assignment problem with Windows and BOINC with machines over 64 threads. Sometimes the problem occurs immediately, sometimes after the initial work units complete. The work units don't get assigned evenly over multiple procs. I have a 2p machine and I've seen every single BOINC work unit get assigned to proc 1 and nothing to proc 2, I have corrected this by manually setting assignment in windows' affinity manager, but then after the work units complete it reverts back. I haven't found a third party affinity manager that supports machines over 64 cores. I have discussed this with Bryan from SUSA and he has tried all sorts of Win server versions without success.
 
Looks like were a few hours from jumping to 9th on cosmology, after LHC had it's issues last night I got a bunch of WU's that I completed really fast, then nothing, so I un-suspended cosmology so it was working on something
 
Took 10th in Cosmology and 13th in LHC, should take 9th in Cosmology in a few hours if outputs stay the same, but going further in LHC is gonna be a bitch
 
I'm only about an hour away, may take the wife for a ride sat. and see if the prices are in line with the lint in my wallet heh



You're a beast and we love you.

Pick me up something good if you find it.

Group hug........
 
Its a little late for that...lol. Not enough of our Folders made that switch. :mooning:
 
I have to be honest, folding seems so much easier. I really wish BOINC was quick and simple to setup.

What are the Galaxy S8+ phones capable of? I used to BOINC on a LG G2 lol.
 
Not sure. I don't crunch on any that expensive. I also don't recommend running any CPU intensive apps on a phone that expensive simply due to wear/tear battery bloat. I stick to the $10 quad core prepaid phones. Much better bang for your buck. And if you run money maker apps, they can actually pay for themselves while they crunch.
 
And I agree about the client. I personally feel the problem isn't really the complexity of the client as the simple view really suffices to an extent. The problem as I see it is that anyone used to FAH have it in their head that they must maximize. Then they get a taste of all the BOINC options and realize how little control over their hardware they really had. They also don't have as much hardware supported by their client. So, when they make the switch they feel overwhelmed due to the "need to maximize" when in reality, they could leave stock settings and just do some good with little headache. Yes they would certainly be underperforming but in my opinion they are doing that running FAH anyways. WCG is a much better project IMO. The rest of BOINC related DC projects are just ways to keep things fresh, unique, new, and challenging. Again, just my opinion.
 
This was weird. My computer just crunched a lot of SixTrack v451.07 (pni) tasks. Most of them ran less than 10 seconds.
 
Just looked an have had some as well. 13 seconds run time 5.74 seconds cpu time earned like .13 points.
 
I don't think BOINC is all that complicated... but the VirtualBox stuff is REALLY starting to piss me off. :mad::banghead: I mean, FFS, I use VBox all the time, but this shit is supposed to "just work" and it most certainly doesn't. The regular non-vbox BOINC stuff is easy peasey.
 
I don't think BOINC is all that complicated... but the VirtualBox stuff is REALLY starting to piss me off. :mad::banghead: I mean, FFS, I use VBox all the time, but this shit is supposed to "just work" and it most certainly doesn't. The regular non-vbox BOINC stuff is easy peasey.

Did you manage to get it to work tho?
 
I haven't messed with the Linux boxes today; just don't have the time to fight with a possible huge headache today. Did try to get the E7 system moved over, since I thought it would be pretty simple... it's not being nice. Am waiting for the last of the Cosmo units to clear out, then will nuke it.
 
I have insurance lol.

That's a quick turn around. Does a CPU do fairly well on BOINC these days vs a GPU? I have a Ryzen 1700x.
 
By the way, did you folks see Jeeper's update for today?

Rank 4 has never before been so coveted as in this year’s Marathon. OCN would love to keep this position and would have earned it, without a doubt. Team China (#5) and SG (#6) are two teams who are fighting for every point. Once those two get into their stride, the hard won advantage will begin to shrink. Whilst AF continue to secure #7, it’s gone eerily quiet at CNT (#8). The Czechs love to bunker and lead their opponents into a false sense of security. RKN (#9) and OcUK (#10) are still having to throw the dice to decide the order of rankings. [H]ard|OCP (#11) who are coming closer are also bringing their over-sized dice to the table.

I have no idea what he's talking about! :whistle: :ROFLMAO: Never mind the fact that we're now in 9th and making the Czech's nervous. ;)
 
I think I call it a day with LHC; no more work since a day or so and still some pending validations. Cleaned up my BOINC boxes and removed all those yoyo, cosmo and for sure LHC for the time being; back to WCG and GPUgrid. Lots of work there.

Thanks for you guy helping me to (re)understand some boinc stuff; still lots to learn.
 
Great to have you and all the other [H]ordes in our team. Remember the next pentathlon (penta = five) for 2018 is May 5th. The date is easy to remember.

As for the project selection in the pentathlon, I would like to see the breakdown of the project votes by team. According to SG website, they are transparent in publishing the result. Does anyone know here to find this?
To keep it as transparent as possible, we will publish the suggestions of every team when announcing the final project.
 
I couldn't pull LHC on my phone. I wondered if it was just because I'm on Android. Kept saying no work.
 
Still 307 pending on LHC with 1 hour left, we could have passed USA for 12th had that moved at all :(
 
yea, i think someone just sitting on all them WU for LHC :(
 
LHC does not have Android work units so, phones/tablets would not be helpful.

pututu, they posted it on their site last year, I will keep an eye out.

Sorry guys, I was on the road all day today. Great job to everyone.
 
My Atlas (LHC) vbox point keeps going down over time but I'm happy with my planck vbox tasks (cosmology). Managed to finish 73rd in the Planck app 2017 Pentathlon contest.

Sure vbox sometimes can be a PITA to setup and requires huge amount of RAM which I didn't have on my rigs.

Great work by Phoenicis who finished 61st in our team.
 
Cheers! I didn't even realise there was a parallel planck competition! Probably a good thing or else I might not have moved on to the anticlimactic agony of LHC.

Great work by all involved ,,,, it was a lot of fun especially after our very own thermonuclear weapon arrived.

I'm off to do a bit of folding until the FB Sprint kicks off.
 
Well..when I saw the users that were on top of the list... I knew there was no way I could even make top 10 so didn't try on Planck. Mumps essentially has a data center at his disposal...
 
Final ceremony and top 15 teams performance review by pschoefer.

We did better than last year event. 2015 was our best year at 9th with fastgeek & grandpa in.
12. [H]ard|OCP (12-13-15-14-9) – previous year #15
The further they got into the Pentathlon, the harder it became. They invested a lot at the Marathon and reaped the rewards. What a contrast to last year. There, they claimed their best ranking in the Sprint. At the very least, they are showing their rivals an ability for tactical changes which isn’t something all teams can call their own.

Well done team and see you all again in next year 9th BOINC Pentathlon on May 5, 2018. Looking forward for better turn-out next year and maybe a medal in one of the five events?
 
While this is still fresh in our mind, here is my own personal post-mortem review of each of the five skills needed to be successful in the next Pentathlon. Hopefully when we revisit this forum thread next year, it may help us to remind and prepare better for it. Again, my own personal view and not reflecting everyone's view here and not to criticize any specific person or group.

Quote directly from the BOINC Pentathlon website:
The BOINC Pentathlon is an innovative cross-project competition for teams in the BOINC world and runs for two weeks. Various skills of the teams are tested.

  • Mobilization: who recruits the most active crunchers in the own team for the Pentathlon?
  • Communication: which team arranges the best strategy and tactics?
  • Strategy: clever resource sharing in terms of the overall stats or concentration on few disciplines?
  • Flexibility: who copes best with different requirements of the projects (e.g., CPU vs GPU) and reacts fastest to the announcement of the projects?
  • Tactics: which team manages fast project switching best?

Mobilization: aka recruitment. Having one or two big crunchers certainly help but the successful teams always have large number of participants. I think this is the most important team skill needed without saying. Just look at the top three teams participating in this year Pentathlon. See the BOINCstats. Use this to gauge where we stand with respect to the rest of the team.

Our team is at #41 and if you click on that little chart icon
upload_2017-5-21_9-28-51.png
, you will see that we have 674 members participating in all BOINC projects and 69 active crunchers. Top teams have hundreds to a few thousands.

upload_2017-5-21_9-27-10.png


What can we do to recruit more crunchers or make folks aware of this? Constructive suggestions are welcome here.
- Awareness. Create a sticky post in the [H]ard|Forum post about this event or perhaps in the [H]ard|OCP main page? Kyle needs to approve this but not sure if he or this organization truly supports this once a year event, so perhaps other senior members can comment. More like an ad but with very limited time frame as I know ad spaces are not cheap.
- PM other [H]ordes or even other [H]ard|Forum or F@H members that you know so as not to appear like doing door-to-door sales person.

One thing that I often hear during this event is that "we are stretched very thin". Hence I think this section of the challenge skill set is the most important.

Communication: Thanks Kyle for creating and providing password protected thread ([H]DC Strategy) where this is by invitation only as Pentathlon battle tactics are being discussed. We also have quite a number of PM sent out to members, so I think we did very well here.

Strategy & Tactics: Initially we thought of focusing on a few events but with fastgeek involvement in marathon, other members were able to deploy their assets to some other events and helping the team to move up to overall #12 position. Though we have 50+ active WCG crunchers, quite a number did not switch to Zika project which will be a bonus. It shouldn't be that difficult to switch WCG projects, perhaps others have personal goal, not an active member of this forum or maybe not interested, etc. Perhaps they are now aware of this Penthathlon (communication and mobilization skills?). Same goes for E@H where we could get some help from folks in F@H.

Bunkering is something that should be included especially for small team like ours. Not sure how many take advantage of the bunkering. Whether bunkering should be allowed in Pentathlon is a separate discussion.

Projects with quorum 2. Need a good planning to minimize "pending for validation" which is not counted to the overall points. I think we did not do well here in general i.e. one should download as many tasks as possible spread out over a short period early to improve chances of tasks being "completed and validated".

Also, some of us are not familiar with other BOINC projects, hence I think by participating in the Formula BOINC throughout the year will give exposure to other projects as well and is a good long term investment if one wishes to participate in the next Pentathlon.

Flexibility: At least for me, I switched projects and sub-projects a few times but I'll try to keep the team informed of what I'm doing in case there is any miscommunication or wrong focus.

Other comments:
- project selection plays an important role here. Some are familiar with projects/sub-projects yielding good points (eg in WCG).
- Not sure why these 4 projects other than cosmology were selected by the participating teams. Would like to see how this breaks down later by each team project nomination.
-Server projects are being tested heavily (once a year for this event). Frustration is pretty much everywhere but this is where I think a well thought off strategy, communication and patient are essential.
 
Back
Top