BOINC Pentathlon 2025

[Markfw]

MW had 1k tasks available.
WCG is barely sending out any.

Just run PG, E@H, NF, NFS and another that can keep up with work for everyone.

well, I guess I have 2 days to decide. I have plenty of WCG and a buffer for most of my hosts (trusted). The 2 I have on milkyway are running great ! If they run out, I will change.

 

[TennesseeTony]

No pics of the dual EPYC !!!

Top shelf, right hand side, 3rd picture, I see 2 Noctua coolers sticking up.

But here is one without the tray installed yet. Hiding behind the monitor.

 

[Skillz]

except 2 host refuse to be trusted. 10 or less jobs ? instead of 128 ???? what a waste

They will eventually get trusted. It hasn't started yet. Just keep everything you got on World Community Grid: Mapping Cancer Markers.

MW had 1k tasks available.
WCG is barely sending out any.

Just run PG, E@H, NF, NFS and another that can keep up with work for everyone.

It seems they auto generate tasks on the fly and keep a ~1000 task bumper. So no signs of running out.

 

[Markfw]

How to you get out of beta projects ???? I can't find an option to opt out of it, and I have 151 tasks for it, and it has the word arthritis in it !!! not the word cancer.

editL found it. Duhhh its right on the menu, I opted out of it.

edit 2> takes forever to exit out. They have been running fir 7 days and have one more day left !

 

[Skillz]

How to you get out of beta projects ???? I can't find an option to opt out of it, and I have 151 tasks for it, and it has the word arthritis in it !!! not the word cancer.

 

[Markfw]

So, here it is over an hour later, and the tasks that I said (and confirmed) to kill are still running !

 

[emoga]

After reading through this thread and viewing all the rig pics, I’ve come to a grim realization: my Cheez-It levels are critically low 😏

 

[TennesseeTony]

https://stats.free-dc.org/userbycpid/33c55b16c9b0fcc5446bc4e21803fec4

Whoa. Gonna have to worry about LTT on WCG. His host list is over 3 pages long on FreeDC

 

[Markfw]

OK, I just found out something. If you need to abort those beta tasks *that have a minimum EYA of 7 days !) you have to go directly to the machine and use boinc manager to do it. BOINCTASKS will not do more than one at a tome. For `151 tasks I was not going to do them one at a time !

 

[Markfw]

https://stats.free-dc.org/userbycpid/33c55b16c9b0fcc5446bc4e21803fec4

Whoa. Gonna have to worry about LTT on WCG. His host list is over 3 pages long on FreeDC

That page won't load for me. Do they look to cloud hosts ??

Going by points, he has had 17 mill in the last 7 days, and I have run it pretty exclusively for that time (only half my boxes) and got 2 mil, same as he did today. Looking at his sat stats on sunday will really tell the tale. He is 27 overall WCG and I am 38.

So Sunday mid day will tell us if we need to worry. I am all in now, except the 2 boxes that can't get trusted, 2 of 128 and 7 of 128 !

 

[cellarnoise]

That page won't load for me. Do they look to cloud hosts ??

Going by points, he has had 17 mill in the last 7 days, and I have run it pretty exclusively for that time (only half my boxes) and got 2 mil, same as he did today. Looking at his sat stats on sunday will really tell the tale. He is 27 overall WCG and I am 38.

So Sunday mid day will tell us if we need to worry. I am all in now, except the 2 boxes that can't get trusted, 2 of 128 and 7 of 128 !

MARK! It does not look good! Stay on WCG if you can! This guy is serious!!!!

I'm firing up my old junk, but even as a team, this one dude is putting up a challenge!!!!!!!

 

[Markfw]

MARK! It does not look good! Stay on WCG if you can! This guy is serious!!!!

I'm firing up my old junk, but even as a team, this one dude is putting up a challenge!!!!!!!

So here is the farm. The 4 in blue highlight will NOT get trusted ! They have been on WCG for 2 days ! The milkyway@home tasks are just taking up some cores on a couple of those boxes, but does NOT affect getting trusted.

 

[TennesseeTony]

Do they look to cloud hosts ??

Going by points, he has had 17 mill in the last 7 days,.....

No. He's running them all from home, a bunch of 8000-series Intel i5's. A BUNCH of them. Sheesh!

 

[Icecold]

No. He's running them all from home, a bunch of 8000-series Intel i5's. A BUNCH of them. Sheesh!

It'll honestly be pretty impressive if we get beat on WCG by a bunch of 6-7 years old off lease Intel prebuilts. He might be the top dog on WCG during the pent but it'll depend on how much output the rest of his team can support with.

 

[Markfw]

No. He's running them all from home, a bunch of 8000-series Intel i5's. A BUNCH of them. Sheesh!

if its 6-7 year old I5's I am sure the farm I have can do way better than 500k a day. We will see Sundays stats (looking at Saturday points.). Now that I have aborted 151 beta thats were getting no points, and won't in this competition. and maybe some of those 9554's can get trusted overnight ? I mean they have been running strait for 2 days. How long do they have to go ?

 

[emoga]

We might actually be in some trouble with the marathon.
Honestly, that’s kind of nice. Finally some real competition for a change.

 

[cellarnoise]

We might actually be in some trouble with the marathon.
Honestly, that’s kind of nice. Finally some real competition for a change.

Don't worry! If my 2014 4 core /8 thread laptop does not crash again.... we may have a chance?

 

[Markfw]

if I can't get full utilization of my CPU's because they are not "trusted" and thats why we lose, I will be more than pissed at WCG, So a little math. If he had 6 core I5 8400's, he would need 21 of them to equal one of my 9554's. So 6 of these would be 126 of those boxes for my 6 (assuming I got all the cores filled. Then my 7950x's 30 more, then the 9654 would be 30 more. Thats 186 of those boxes he would need. And thats assuming they have the same IPC, which I seriously doubt. 186 of those boxes, at 100 watts each is 18600 watts or 155 amp breaker box, not including furnace, stove, washer, dryer, etc...

Now all of that is wild estimation, but I should be able to beat him IF... Thats the big thing.... IF I could get full utilization of all my boxes.

And the pentathlon as well.

edit: so right now those 6 64c/128t boxes should have 768 tasks, But they have 174. THIS is the problem.

 

[cellarnoise]

if I can't get full utilization of my CPU's because they are not "trusted" and thats why we lose, I will be more than pissed at WCG, So a little math. If he had 6 core I5 8400's, he would need 21 of them to equal one of my 9554's. So 6 of these would be 126 of those boxes for my 6 (assuming I got all the cores filled. Then my 7950x's 30 more, then the 9654 would be 30 more. Thats 186 of those boxes he would need. And thats assuming they have the same IPC, which I seriously doubt. 186 of those boxes, at 100 watts each is 18600 watts or 155 amp breaker box, not including furnace, stove, washer, dryer, etc...

Now all of that is wild estimation, but I should be able to beat him IF... Thats the big thing.... IF I could get full utilization of all my boxes.

And the pentathlon as well.

edit: so right now those 6 64c/128t boxes should have 768 tasks, But they have 174. THIS is the problem.

Keep working it! I am still trying to get wcg tasks also!

 

[Markfw]

Keep working it! I am still trying to get wcg tasks also!

The problem is, I have THOUSANDS of tasks, But only for "trusted hosts". and I don't know how long it takes to be trusted. already been like 3 days.

 

[Ken g6]

The problem is, I have THOUSANDS of tasks, But only for "trusted hosts". and I don't know how long it takes to be trusted. already been like 3 days.

I think you get trusted when one of your WUs is double-checked. I got the one machine I have that wasn't trusted to become a trusted host pretty quickly. Edit: In the past it's taken longer.

 

[TennesseeTony]

He has about 1950-2000 CORES, whole lee smoke 'em if you got 'em!!! 325-ish machines. Wow. We're in trouble. Unless even trusted hosts start to have problems getting tasks, not everyone bothers to bunker and WCG may not yet have felt the full force of the Pentathlon. We know he isn't willing to reconfigure his machines to run a different project, because, well, 325 machines. So there is that going for us on other projects.

 

[StefanR5R]

I am catching up after returning from travel.
I updated the calendar in #17 and will continue to do so, but certainly often with random delay because of other stuff happening IRL.

I am lending my computers to parsnip soup, like I did in recent PrimeGrid challenges. Therefore my contribution to Pentathlon will be a few late validations at MilkyWay at best.

Apropos MilkyWay and late validation: Like many other projects, MilkyWay's validator needs 2 matching results per workunit before it gives credit. But unlike any other of such projects that I know of, MilkyWay does not generate 2 tasks per workunit at once and sends out both ASAP. Instead, it sends 1 task, waits for its result to return (success or error), then sends a next task, and so on until there are two matching returns. That is, a workunit is worked on serially, not in parallel, which obviously increases the time until credit is given for the first good result of a workunit.

Reading this thread, it is obvious that the recent cancellations of SETI.Germany's WCG Birthday Challenge (used to happen on November 16–23 every year) caused some WCG bunkering knowledge to get forgotten. I forgot some specifics of that myself, but at least I know that I put some notes somewhere and simply would have to dig those up if I needed them. There were some of us who built multi-day bunkers in WCG Birthday Challenges, and there were others who decided not to yet knew what it took to do so. (With other WCG subprojects, 10 days deep bunkers were doable, which let you put 17 days computer time into the 7 days long contest. Additionally, I for one maintained several of such bunkers in a staggered fashion, such that I had peak output on the last day, not on the first day.)

I don't want to make a point that many days deep bunkers should be done again. After all, Krembil's servers << IBM's servers. And we had more time to prepare for WCG Birthday Challenges than we have for Pentathlon's subcontests. All I want to say is that my first impulse, when I was reading up on this thread, was surprise how project details can get forgotten although they seemed to be well known once.

 

[Markfw]

He has about 1950-2000 CORES, whole lee smoke 'em if you got 'em!!! 325-ish machines. Wow. We're in trouble. Unless even trusted hosts start to have problems getting tasks, not everyone bothers to bunker and WCG may not yet have felt the full force of the Pentathlon. We know he isn't willing to reconfigure his machines to run a different project, because, well, 325 machines. So there is that going for us on other projects.

well, I have 1132, but I think its equivalent to about 1400-1600 of his cores, but yes, I can't beat him. Your 256 core box alone will help a lot !

@StefanR5R did you see my 12 hour estimation of my bunkers ? I can't help there. That's a system problem I can't get around.

edit: I looked. He has 8500 6 cores, so my estimation is sort of right one, except 325 machines ? where could find room for that many in any house ?

and the dell box I looked up uses 260 watts. 325 of those is 84, 500 watts ! or 704 amp ! How could you get that large a service ???? something smells about this user. assuming its true they are all in his house like said above. they also make a 200 watt box and that 65,000 watts or 542 amps.

 

[StefanR5R]

@StefanR5R did you see my 12 hour estimation of my bunkers ?

This is the old problem with its old and well-known solution: A small and slow host can take a long time to work through its work buffer, but a big and fast host tends to take only a short time to complete its work buffer. In other words, work buffer sizes do not scale arbitrarily in proportion to host performance. The obvious workaround to this is to have several work buffers on the physical host, i.e. to have several BOINC client instances on a computer.¹

Not everybody wants to occupy² oneself with this, which is fine, as there may be more important things in life.³ :-)

Spoiler: footnotes

¹) The buffering limitations into which big and fast hosts run are partially caused by the BOINC client, partially imposed by the BOINC server. One could use a BOINC client which is recompiled from edited source code, which changes the conditions under which the client requests more work, and maybe conditions under which it reports results. I never went there myself though. Because in the end, the limits which are set to the host by the BOINC server often cannot be addressed by changes in the client, but can be by running multiple clients per computer. At that point, modified clients can still be a help but are not a necessity.

²) I.e. to learn it, to practice it until it works without surprises, and maybe to scale it up to multi-day bunkering on several computers. To get there is ultimately more a matter of motivation than a matter of prior knowledge and ability.

³) Besides, performance of BOINC servers lags more and more behind performance of contributors' hosts. This makes massive⁴ bunkering activity less and less advisable.
But maintaining certain work buffers on the hosts is still beneficial: At the start of a contest — or e.g. at Steeplechase's infamous bonus days — when lots of people rush to fetch new work and many try to unload some mini bunkers too, it's good if users have work buffers and avoid to contact the server during these rush hours more than necessary. It's good for these users but also good for the server.

⁴) It doesn't matter if few big participants bunker massively individually, or if many regular participants bunker small time but in large numbers. If anything, there is a chance that the former do it with some care and coordination, whereas the latter always do it in plain rush hour style.