Discussion 2019 BOINC Pentathlon thread

[StefanR5R]

Several hours ago, the Fourth Javelin Throw was announced:

                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
 WCG OpenZika   __ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄   14 days Marathon
    yoyo@home            __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄                            5 days City Run
Einstein@Home                      _ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄             5 days Cross Country
       t.b.a.                                                               3 days Sprint
     NFS@Home            __ __ __ ▄▄                                        1 day  Javelin Throw
                                  __ __ __ ▄▄                               1 day
                                              __ __ __ ▄▄                   1 day
                                                    __ __ __ ▄▄             1 day
                                                                            1 day
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa

 

[StefanR5R]

Today I thought of looking something up at the NFS@Home site. I noticed that my validated results from the past Javelin Throw days are no longer listed. That is, NFS@Home is cleaning completed WUs out of its database pretty early — similar to yoyo@home, but probably not as aggressively.

 

[TennesseeTony]

Such quick housekeeping makes it difficult to research.

 

[StefanR5R]

From Jeeper's day 8 bulletin:

I will use this quiet time to take a look at the Marathon (OpenZika). [...] TSBT (#1) continue to run as if they’ve still got a syringe stuck in their bottom. [...]
In the meantime, TAAT (#3) have shown what they are still capable of. They’ve managed to create a little breathing space with a small interim spurt to escape TC (#4) breathing hot air down their neck. The average speed of the Asians appears to be higher though which could make it very tight in the end.
[...]
Behind TSBT (#1) in the Overall Standings, the fight for Silver is in full swing. P3D (#2) and TAAT (#3) could further detach from their rivals and have both still got the potential at the Javelin Throw or Cross-Country respectively.

 

[StefanR5R]

SETI.Germany's pschoefer posted graphs of

a) PPD at World Community Grid in its entirety,



b) PPD at OpenZika in particular:



"nach Bekanntgabe" (yellow bars) = after announcement
"Durchschnitt" = average

WCG performed some maintenance on May 1.
On April 30, regular WCG contributors obviously got a somewhat lower share of OpenZika tasks (which were vacuumed up by all those pesky bunkerers) and automatically ran more of other subprojects.

 

[TennesseeTony]

@Assimilator1 Hey man. Hey. Psssst. We need you running Einstein for us right now....

 

[StefanR5R]

The last discipline has been announced just now. And it's not pretty.

                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
 WCG OpenZika   __ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄   14 days Marathon
    yoyo@home            __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄                            5 days City Run
Einstein@Home                      _ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄             5 days Cross Country
     LHC@home                                           _ __ __ ▄▄ ▄▄ ▄▄    3 days Sprint
     NFS@Home            __ __ __ ▄▄                                        1 day  Javelin Throw
                                  __ __ __ ▄▄                               1 day
                                              __ __ __ ▄▄                   1 day
                                                    __ __ __ ▄▄             1 day
                                                                            1 day
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa

 

[StefanR5R]

Only the last Javelin Throw day is still to be announced yet.

With the announcement of the last discipline, the team votes for projects was published as well.

Of the five projects which made it into this Pentathlon, LHC@home got the fewest, and merely 1 more vote than either Amicable Numbers or Cosmology@Home.

 

[StefanR5R]

From the bulletin on day 9:

Today is the third day at the Javelin Throw which means that the teams need to bare all.

Yep, that's what he says it takes!

Tomorrow, I will need to evaluate how this has been working out. There’s also half-time at the Cross-Country (Einstein@Home) which gives me a chance to check that everything is above board there. [...]
AF (#7) aren’t able to see any of the fast pacers ahead of them, the distance is too large. Even though OCN (#8) attacked and catapulted forwards, it looks like our feathered team have done enough. RKN hold on to #9 but are still able to retaliate and attack OCN. Regretfully, after a good start, TAAT are slipping back, currently Rank 10. The danger to slide even further back is limited as the advantage over BOINC@AUSTRALIA (#11) is larger than the one the Australian’s have over LTT (#12).
[...]
There’s a bit of confusion in the Overall Standings. The day at the Javelin Throw is causing unrest. Although TSBT continue to lead, they’ve forfeited most of their advantage. It is quite clear to everbody that this can’t remain so as they seem to be crunching a result which will surely be crossed out. P3D have cemented #2 and have a new first pursuer. CNT continue to remain at #3, still ahead of [H] (#4) and SG (#5). TC (#6) are ranking only just behind, barely ahead of TAAT (#7) who lost a considerable amount of feathers at the Javelin Throw. Here too, we seem to be looking at a planned strike-out result.
[...]
What’s left from today?
- Will TSBT gamble it all away?
- Will CNT be the laughing third this year?
- And will SUSA keep in the Top20?

As you can see, it remains exciting at the BOINC Pentathlon.

 

[StefanR5R]

For those who aren't avidly following the Pentathlon shoutbox:

]World Community Grid made SETI.Germany use a new stats interface this year. SG tested the import for a duration of several weeks before the Pentathlon. Nevertheless the infamous problem of negative stats during several hours after 00:00 UTC (in the worst case, for ~half a day) became visible only after the start of the Pentathlon.

And now, several hours ago, the WCG staff decided that they learned from the past week of Pentathlon running life production that they need to disallow certain simultaneous requests issued by the Pentathlon site, for performance reasons. SG's admin kleines grünes Männnel just put in a night shift and changed SG's stats import once more to compensate for WCGs sudden new restriction.

As a side effect of this latest change, the hourly stats at the Pentathlon get current at ~2 minutes after the hour (not in a matter of seconds like before). And this applies to all disciplines, not just the WCG OpenZika Marathon.

With >5 days still to go, I won't be surprised if we will see further and weirder WCG stats schenanigans yet. A reassuring aspect of all the past troubles though was that SETI.Germany always received valid data from WCG eventually, just not punctually. Hence, very final stats might not be available at the exact end of the event, but maybe later. (And by that I mean a canonical 00:00 UTC snapshot across all teams, not a random mish mash of older and newer stats.)

 

[StefanR5R]

The last event has been announced a few hours ago:
The 5th Javelin Throw happens on Friday.

                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
 WCG OpenZika   __ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄   14 days Marathon
    yoyo@home            __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄                            5 days City Run
Einstein@Home                      _ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄             5 days Cross Country
     LHC@home                                           _ __ __ ▄▄ ▄▄ ▄▄    3 days Sprint
     NFS@Home            __ __ __ ▄▄                                        1 day  Javelin Throw
                                  __ __ __ ▄▄                               1 day
                                              __ __ __ ▄▄                   1 day
                                                    __ __ __ ▄▄             1 day
                                                          __ __ __ ▄▄       1 day
                30 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
                               Su Mo Tu We Th Fr Sa Su Mo Tu We Th Fr Sa

 

[StefanR5R]

From the bulletin on day 10, looking at the outcome of Javelin Throw of Monday:

Crunching@EVGA are currently at #7. They move along with stable results but just not at the top level. It will be a challenge for them to hold on to the Top10. This will allow TAAT (#8) to easily pass them because they enjoyed some time out yesterday and saved their strength for the next throw. This could even take TAAT to the provisional top spot.
[...]
- Will tomorrow be decision day at the Javelin Throw?
- Is the Javelin Throw SG’s only chance at a medal?
- Could TAAT re-enlist in the fight for medals?

As you can see, it remains exciting at the BOINC Pentathlon.

 

[TennesseeTony]

"Re-enlist" in the fight for the medals....Hmm. Interesting use of English. Being former US NAVY, I say:
N-ever
A-gain
V-olunteer
Y-ourself

 

[StefanR5R]

This context wasn't there in Jeeper's German wording, which is more like "report back" or "show up again". Sometimes the translation loses something, sometimes it adds something. It's often a tough job to translate Jeeper's reports. If the translators ever heard of the N.A.V.Y. principle, they recklessly neglected to apply it for themselves.

 

[StefanR5R]

Day 11, marathon: "Some of the Pentathletes turn out to be lone wolves. Others prefer company and like it in groups. TSBT continue to control the field and can already see Gold shining in the distance. Likewise for P3D whose chances for Silver are good, whereby TAAT (#3) continuously accelerate, however, they are unable to come any closer. TC (#4) are the big unknown. Ordinarily, I would count on an attack by the Dragon who appears sated and sluggish this year." (This was written before Team China had enough of Jeeper's complaints and dropped some bunkers.)

 

[Assimilator1]

@Assimilator1 Hey man. Hey. Psssst. We need you running Einstein for us right now....

Hey , don't have my HD 7970 installed anymore, but I've just started crunching it on my RX 580 , how desperate are you for extra firepower? I could fire up my 2nd rig for a short burst .

 

[StefanR5R]

The Einstein race ends in 1h 51m.
But an LHC race begins in 1h 51m. :-)

 

[IEC]

Thankfully I can shut down my GPUs soon... it's insufferably warm here.

In other news, 1/3 of my completed tasks for Einstein (1172) are pending validation. Annoying.

 

[StefanR5R]

@IEC, check out the last-hour uploads according to the Cross Country stats:

[H]ard|OCP +62 M,​

SETI.Germany +54 M,​

Planet 3DNow! +16 M.​

​

Did they at least validate some of yours? :-)

 

[StefanR5R]

Einstein@Home:

I wonder, will this year's Einstein Cross Country results maintain any remote resemblance to those of 2017 Cross Country at Einstein at all? :-)

1 ‒ Overclock.net​

2 ‒ SETI.Germany​

3 ‒ SETI.USA​

4 ‒ Czech National Team​

5 ‒ Planet 3DNow!​

6 ‒ Team China​

7 ‒ Gridcoin​

8 ‒ Rechenkraft.net​

9 ‒ TeAm AnandTech​

10 ‒ L'Alliance Francophone​

11 ‒ OcUK - Overclockers UK​

12 ‒ Meisterkuehler.de Team​

13 ‒ [H]ard|OCP​

14 ‒ Crunching@EVGA​

15 ‒ USA​

Heh, at least we are consistent; by rank and by credits. (114 M then, 127 M now.)
As suspected, most of the other teams performed wildly different this year.


LHC@home:
pschoefer commented on LHC's effective disabling of the quorum 1 subprojects (translated):

Life is what happens while you make other plans. LHC@home and also yoyo@home have subprojects with quorum 1 as well as subprojects with quorum 2. We file the projects under quorum 1 if the user has the option to select WUs with quorum 1.

Those who follow the project news know that SixTrack was paused for months, and only at the beginning of May with resumed operation with new applications. Accordingly, LHC@home even had only WUs with quorum 1 when the project elections took place (April 7...28) and when we assigned the projects to the disciplines (April 29). I do not want to conceal that the LHC@home people of course know from the last time that SixTrack is usually in greater demand than the vbox subprojects. Yet the teams actually had to consider at the election that there would be only the vbox subprojects available. Now, the day before yesterday, the database issues happened too, therefore the project is focused on using additional servers to distribute the much requested SixTrack WUs as smoothly as possible, even though this makes the subprojects with quorum 1 harder to obtain.

Would it have been better to skip the sprint completely (Tetrathlon) or postpone it beyond May 19 (would be a normal challenge without the usual pentathlon situation with overlaps with other disciplines) or to switch to another project, that would have had no preparation time at all? (Of the three other possible projects, two had massive problems in the last few days.)

sources:

• LHC's announcement
• pschoefer's post in the SG forum, in German

 

[crashtech]

I guess I don't have to feel bad about bunkering Sixtrack anymore.

The only question now is what to allocate to WCG and what to allocate to LHC.

 

[Ken g6]

The only question now is what to allocate to WCG and what to allocate to LHC.

As I recall, WCG runs better on Ryzen. What about SixTrack?

 

[crashtech]

As I recall, WCG runs better on Ryzen. What about SixTrack?

Is it possible to tell? The (Sixtrack) completion times on my end are all over the place as far as I can tell, for various Intel generations as well as Ryzen.

 

[zzuupp]

Is it possible to tell? The (Sixtrack) completion times on my end are all over the place as far as I can tell, for various Intel generations as well as Ryzen.

At least you're getting tasks. For whatever reason, I've been CPU dry on LHC.
Checking preferences, otherwise back to WCG.

& yes, I have the vbox boinc installed on one box.

 

[crashtech]

At least you're getting tasks. For whatever reason, I've been CPU dry on LHC.
Checking preferences, otherwise back to WCG.

& yes, I have the vbox boinc installed on one box.

Pretty sure it's Sixtrack only this time around.