PrimeGrid Challenges 2020

[StefanR5R]

@Ken g6, thanks for the stats.

@VirtualLarry, I think what Ken g6 was referring to is that not a lot will change during the "cleanup" phase anymore, because

1. invalid results are now more likely to be detected already locally by the application,
2. the validation tasks are very small and therefore quick to complete.

Calculation errors are therefore less likely to remain undetected until later when the wingman turns in the result to compare with, and in addition, the wingman now needs very little CPU time to to complete the validation task. There is not a lot left to be done in the cleanup stage anymore.

Right now, host 973560 is showing

• 47 results which the application itself flagged as unrecoverable error,
• 44 results which the application flagged as error which it attempted to correct and submitted for validation, but then they failed validation when the wingman completed the little validation task,
• 82 results which the application flagged as error which it attempted to correct and submitted for validation, and were validated successfully after all,
• 1 result which the application flagged as error which it attempted to correct and submitted for validation; the corresponding validation task has not been completed yet,
• 146 results in which the application did not detect an error and which were validated successfully,
• 12 tasks in progress.

That is, you have just one task still pending validation, therefore nothing much can happen anymore to your score during the "cleanup" phase of this challenge.

 

[VirtualLarry]

82 results which the application flagged as error which it attempted to correct and submitted for validation, and were validated successfully after all,

Thank you for the precise summary.

The above that I quoted is a little bit worry-some to me in regards to the project. If the application flagged it as an error, but also, it validated? Meaning, the wingman had the same error, and thus, passed validation? Is that what that means? It's a little frightening, from the science perspective.

 

[StefanR5R]

No, the application detected an error, re-ran the portions which it identified as faulty, figured that the re-run was good, and then submitted this corrected result for validation.

The PrimeGrid scientists/ admins definitely are not making any compromise in this regard.

 

[Ken g6]

44 results which the application flagged as error which it attempted to correct and submitted for validation, but then they failed validation when the wingman completed the little validation task,

Yet they say only 13 were that way. I think they're still mis-counting invalid WUs.

 

[StefanR5R]

Michael Goetz is all enigmatic about it.

My guess: "Of the tasks that have been returned successfully: x-hundred thousand have been successfully validated" does not count tasks, but workunits, and it includes all workunits which had an invalid result first but a valid result after replication and before the stats posting.

 

[Ken g6]

He clarified.

I really should be clearer because the LLR2 mechanics are so new.

Those statistics, to be meaningful, do not count the fast DC tasks. Only the main tasks are counted.

The third line of the statistics says:

Fast DC tasks are NOT included.

All of Ken's teammate's errored tasks (that computer has some serious problems) were fast DC tasks, and thus are not counted in the statistics.

And by the way, if those tasks hadn't errored, they would have been significantly faster than they were, which might be why I didn't notice the difference.

 

[TennesseeTony]

My remaining 3800x had 13 or so "WARNING!" tasks, but validated (uh, sorry, ran for [H] in the final hours, only 186k points).

Ken, we really appreciate the stats, and also appreciate the others who present stats while you are busy. I only have to worry about stats once a week, and I know it must be far more difficult to compile DAILY stats during these events. I tip my hat to you and the others!

 

[VirtualLarry]

My remaining 3800x had 13 or so "WARNING!" tasks, but validated (uh, sorry, ran for [H] in the final hours, only 186k points).

So it's not JUST me, that had those "warning!" in red on their tasks.

 

[VirtualLarry]

Ken, we really appreciate the stats,

This! 200%!

 

[Ken g6]

Bump for the next challenge!

There are two projects: Cullen and Woodall LLR. You can run either or both. Choose wisely!

 

[Ken g6]

And the challenge has started!

I think I'm going to be slow getting going. Too many WCG WUs downloaded.

 

[VirtualLarry]

Is there a points advantage or CPU/memory usage advantage to one or the other? I checked off both in my PrimeGrid project prefs on their site, and then set my two Ryzen R5 3600 PCs to "Update", and "Allow New Tasks", and "Run based on Preferences".

WUs showed 1d and 2d (days) when downloading, but now that they're running, they're showing 5d and 6d. Is that normal? I've shut down my CPU-mining on those rigs.

 

[emoga]

WUs showed 1d and 2d (days) when downloading, but now that they're running, they're showing 5d and 6d. Is that normal?

How many threads are you using to Multithread those tasks? 1 or 12?
Sounds like you're running those large tasks with a single thread by the sound of that deadline.
The single Cullen task that I have running on a 3600 with all 12 threads working on it will finish in about 8ish hours.

 

[Ken g6]

Day 1 stats:

Rank___Credits____Username
70_____82288______emoga
111____41173______Ken_g6
128____40021______SlangNRox
161____20499______waffleironhead

Rank__Credits____Team
23____209430_____Ukraine
24____206563_____BOINCstats
25____187785_____Noobs Of Kryta [NOOB]
26____183982_____TeAm AnandTech
27____143348_____Team-Goobee.org
28____102641_____BOINC World
29____86266______Special: Off-Topic

I wanted to wait to post stats until my first work finished.

 

[VirtualLarry]

Hmm, I must be approaching this challenge with the wrong settings, I won't have any results to report for 5.5 days.

 

[waffleironhead]

Hmm, I must be approaching this challenge with the wrong settings, I won't have any results to report for 5.5 days.

Im not sure of your hardware, or if you have enabled multithreading in your primegrid preferences.
For reference, my 3700x was getting 2.5% done per hour when I was running 8 threads and 2 WU's at a time. Then I tried running one WU at a time using 8 threads(not using the extra thread per core) and was getting 7.5% per hour progress. Then I switched to one unit running 16 threads and am getting ~9.7% per hour(those hyperthreads are helping!).
Also, my lowly 2400g systems are taking 3.24% per hour with all 8 threads.(1 day 8 hours per wu)

 

[Ken g6]

Day 2 stats:

Rank___Credits____Username
112____123690_____emoga
124____102334_____SlangNRox
126____100434_____Ken_g6
141____82104______waffleironhead

Rank__Credits____Team
24____486263_____Ukraine
25____475316_____USA
26____409547_____Team-Goobee.org
27____408563_____TeAm AnandTech
28____320635_____Alien Prime Cult
29____309270_____The Scottish Boinc Team
30____268610_____Canada

We actually lost a place. I hope things get better when the WCG challenge ends.

 

[Ken g6]

Day 3 stats:

Rank___Credits____Username
114____222297_____Ken_g6
124____186131_____emoga
131____164649_____SlangNRox
139____144892_____waffleironhead
242____43950______zzuupp

Rank__Credits____Team
23____1027781____Ukraine
24____845846_____Noobs Of Kryta [NOOB]
25____769411_____USA
26____761921_____TeAm AnandTech
27____700669_____Team-Goobee.org
28____591164_____Alien Prime Cult
29____580898_____Canada

You know we're in trouble if I'm the leader of the pack.

 

[Ken g6]

Day 4 stats:

Rank___Credits____Username
118____285123_____Ken_g6
130____232214_____waffleironhead
139____207636_____SlangNRox
140____207082_____emoga
244____68411______zzuupp

Rank__Credits____Team
23____1336664____Ukraine
24____1247080____Noobs Of Kryta [NOOB]
25____1063280____USA
26____1000468____TeAm AnandTech
27____973140_____Team-Goobee.org
28____873028_____Russia
29____813408_____Canada

I really hope we pick up speed after the WCG challenge ends tonight. It's embarrassing to be beaten by Noobs.

 

[lane42]

Ill be in. A couple more hours

 

[Orange Kid]

Is there a better choice of the two or are all things equal in the long run?

 

[Ken g6]

Is there a better choice of the two or are all things equal in the long run?

Good question. Apparently, right now, Woodalls are slightly bigger than Cullens. Otherwise, no, there's not much difference.

 

[Ken g6]

Day 5 stats:

Rank___Credits____Username
115____415428_____Ken_g6
149____270878_____SlangNRox
150____270771_____emoga
158____252754_____waffleironhead
250____92876______zzuupp
255____86514______Orange Kid
276____65404______Tejas
281____64521______Lane42
315____43515______Icecold
316____43492______biodoc
382____19538______VirtualLarry

Rank__Credits____Team
22____2144574____BOINCstats
23____1882364____Ukraine
24____1667589____Noobs Of Kryta [NOOB]
25____1625696____TeAm AnandTech
26____1409805____USA
27____1246890____Team-Goobee.org
28____1175318____Canada

I was hoping to wait until we passed the Noobs to post this, but like the WUs themselves, it's taking longer than I expected.

 

[lane42]

Those noobs are Toast

 

[Ken g6]

Day 6 stats:

Rank___Credits____Username
114____505062_____Ken_g6
116____483462_____VirtualLarry
141____358560_____emoga
148____336319_____SlangNRox
154____316987_____waffleironhead
171____279332_____Orange Kid
228____173559_____Tejas
238____151947_____Lane42
255____130726_____biodoc
268____117401_____zzuupp
284____92815______Icecold

Rank__Credits____Team
19____3357861____BOINC@Poland
20____3089745____Duke University
21____3030418____UK BOINC Team
22____2946174____TeAm AnandTech
23____2513056____BOINCstats
24____2207501____Ukraine
25____2050027____Noobs Of Kryta [NOOB]

The "I can't believe I'm still leading!" edition. I hope that's because somebody's building up bunkers. On second thought, I don't! (See sig.)