8th Annual BOINC Pentathlon

[StefanR5R]

@Ken g6, they do have tasks, but hand them out by the spoonful.

In other news, here is a PC which looks adequate to run Yoyo/ECM:
https://www.servethehome.com/hpe-the-machine-reaches-160tb-uses-cavium-thunderx2/

 

[Ken g6]

@Ken g6, they do have tasks, but hand them out by the spoonful.

Are you sure? My laptop is still getting that, and now has one core free.

 

[TennesseeTony]

Might have enough memory, lol, but....ARM processors? Uhm....

*******************************************************************

We're almost famous guys, Jeepers used our full name! Snippets:

(re: YoYo) TeAm AnandTech seem to have taken a liking to #9. With an end spurt, they secured #9 for the 3rd time in the third discipline.

(re: LHC) The battle behind is completely open. TAAT (#12) are currently pushing ahead with the aim to claim #9 of course. [I love that part. ] That’s where we can find RKN at the moment. To get there they have to push past OcUK (#10) and BOINC Confederation (#11) [which happened not long after this bulletin].

********************************************************************

For those running out of LHC, try....

Uhm, maybe this should be classified info....posting in our conversation instead. Sorry ocukguest, Jondi, Coleslaw, 4thKor, you guys are on your own.

 

[petrusbroder]

I am also crunching LHC. But I do not get sufficient WUs ...

 

[TennesseeTony]

Hey, I think I reached my maximum invites for the private conversation, somebody get petrusbroder in there, please. Off to the right on the conversation window, there is a box showing the participants in the conversation. At the bottom you should see "invite" so let's make sure we haven't forgotten to invite anyone else, except the competition.

Sending you a PM for the secret recipe petrus....

 

[StefanR5R]

@Ken g6, they do have tasks, but hand them out by the spoonful.

I suppose the "unsent tasks" shown at the server status page are in fact a pool of buggy tasks which remain in circulation until their max number of client errors is reached.

 

[TennesseeTony]

Over half my machines were in full blown buggy mode this morning with no time to attend to them before leaving for work. :'(

 

[[H]Coleslaw]

For those not able to pull work, are you using the old six Track address or did you connect using the new one? I've heard some claim that the old one doesn't always deliver....

 

[StefanR5R]

[Swimming, LHC@home]

Mentioned by @[H]Coleslaw in the Pentathlon shoutbox:

Nils Høimyr, project admin - Message 30374, May 17, 11:52 UTC -
There is a slight lag with scheduling due to the large number of short tasks. We're looking into it. We had a temporary issue with the feeder, now fixed.

Please try to reset your BOINC client and do "update" if you do not get any tasks soon.

(There are more problems caused by these short tasks of theirs, but they are at least aware and may address that eventually too.)

 

[crashtech]

My 24-thread Westmere machine is neck-and-neck with the 16 thread Ryzen, both run very similar clockspeeds, giving Ryzen about a 50% IPC advantage in LHC.

 

[IEC]

Looks like I missed this. Too late to join or would a few days output from some Ryzen CPUs help out? I'd also need an idiot's guide to getting started. Not because I'm an idiot, but because a quick start would help make up for lost time.

 

[crashtech]

Looks like I missed this. Too late to join or would a few days output from some Ryzen CPUs help out? I'd also need an idiot's guide to getting started. Not because I'm an idiot, but because a quick start would help make up for lost time.

We'd love to have you! There is 24hrs left in both LHC@Home and Cosmology @Home! Are you familiar with BOINC?

 

[IEC]

We'd love to have you! There is 24hrs left in both LHC@Home and Cosmology @Home! Are you familiar with BOINC?

Not familiar with specifics, unfortunately. I used to run Folding@Home a long time ago but haven't really been in the DC scene for a while now.

 

[crashtech]

Not familiar with specifics, unfortunately. I used to run Folding@Home a long time ago but haven't really been in the DC scene for a while now.

Start here:

https://boinc.berkeley.edu/download.php

If you're not familiar with VBox you may wish to avoid it, but some projects require it. I'll PM you on how to start.

 

[TennesseeTony]

I'm still having difficulty with several machines, where the short LHC tasks has elevated my supposed GFLOPS. What is the solution? Run out of tasks, re-run the CPU benchmark from the tools menu, then resume downloading?

 

[IEC]

Looks like I'm up and running on 1 rig:

 

[iwajabitw]

I'm still having difficulty with several machines, where the short LHC tasks has elevated my supposed GFLOPS. What is the solution? Run out of tasks, re-run the CPU benchmark from the tools menu, then resume downloading?

Possibly the benchmark, or reset the project.

 

[TennesseeTony]

Wow. We got mentioned at the end of the bulletin.

Will SUSA still attack Gridcoin at Swimming?
What are CNT planning?
Can TAAT continue their series of 9th rankings?

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

******************************************************************



Welcome to the party IEC, and thanks for joining!



****************************************************************

 

[IEC]

Rig #2 is up and running:

 

[IEC]

Rig #3 is up and running. The trusty box o' parts had an extra wireless adapter and I rebuilt the 1700's WC loop.

 

[crashtech]

Rig #3 is up and running. The trusty box o' parts had an extra wireless adapter and I rebuilt the 1700's WC loop.

That's the spirit!

 

[StefanR5R]

I'm still having difficulty with several machines, where the short LHC tasks has elevated my supposed GFLOPS. What is the solution? Run out of tasks, re-run the CPU benchmark from the tools menu, then resume downloading?

No, the benchmarks are not helping. @Kiska pointed me towards a thread with a workaround:
https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4273

Recipe:

1. set project to "Won't get new tasks"
   
2. shut down boinc-client
3. open client_state.xml with a capable text editor
4. use a "search and replace all" function to replace all occurrences of
   0.000000</rsc_fpops_bound>
   by
   0000.000000</rsc_fpops_bound>
   i.e. add three more zeros to all tags like
   <rsc_fpops_bound>180000000000000000.000000</rsc_fpops_bound>
   which are located within <workunit> tags
5. save the modified client_state.xml
6. restart boinc-client
7. It should now be able to finish tasks even if their estimated remaining time is a lot lower than their real time for completion
8. after you completed a good number of longer running tasks that way, set project to "Allow new tasks" again

If anybody has a better solution, I would be interested.

 

[StefanR5R]

PS,
if you don't have time to babysit the machines, you can perform step 8 sooner than later. But it involves the risk that you receive another number of tasks in which rsc_fpops_bound = the self-destruct timer is set too low.

 

[StefanR5R]

Rereading the thread, a better workaround may be to modify

<app_version>
<app_name>sixtrack</app_name>
[...]
<flops>whatever</flops> <-- decrease this number?
</app_version>​

in all versions of sixtrack app blocks in client_state.xml (while boinc-client is completely shut down). This should allow to set "Allow new tasks" immediately after boinc-client is restarted.

I haven't tried this type of the workaround myself, only the workaround from post #472.

 

[TennesseeTony]

Thanks Stefan. I didn't know a "replace all" type of edit was possible. I started doing that then realized there was too many to effectively complete.

I stopped new work, aborted what had not started yet, let the others finish, ran the benchmark, restarted BOINC, re-allowed downloads and things were 50/50, some good, some only a few seconds. This lasted an hour or so before they are 99% good now.