BOINC Pentathlon 2020

[Gorgon]

Had my 6-GPU rig continuously spit out results with computation errors. This was the system the new DDR4-3200 RAM went into. Disabling the XMP and dropping back down to JEDEC 2133 seems to have fixed it so far.

[homeofmew]

so is there an official sign up sheet?
I replied earlier in the thread I want in

[leadeater]

5 hours ago, Windows7ge said:

I think bunkering overall is cheating but everyone does it so it evens out. I get the feeling even if you grabbed as much work as you could with VMs it wouldn't starve the server of work to hand out.

Looking at their stats page they have something like 30k tasks available, with the number in progress I'll take that number as a per hour. With my docker method and ability to pull 6k tasks per server group and instance set I could write a script that deploys enough docker instances per hour (6 * 6k) so I might just be able to pull it off lol.

[leadeater]

8 minutes ago, homeofmew said:

so is there an official sign up sheet?
I replied earlier in the thread I want in

Just sign up to the Rosetta@home project and join LinusTechTips_Team and that is all you need to do to participate in the marathon event. It's a teams challage so the team signs up rather than each person.

 

There are 4 other events during the two weeks so if you want to help out with those keep an eye on this topic over those 2 weeks.

[leadeater]

6 hours ago, IntMD said:

The tasks are then supposed to end after that runtime has expired, and it has run it's last instance/checkpoint or something.

Do you have an idea how that effects points generation, does it also limit you to what tasks you can get? Seems odd you can just end a task like that without a proper completion point. I'd rather not change it if it limits me to a much smaller subset of tasks.

[chaozbandit]

WUs showing "waiting to report" are considered completed right? Is there any detrimental effect if you shutdown a client with several WUs in this state in regards to points?

[leadeater]

3 minutes ago, chaozbandit said:

WUs showing "waiting to report" are considered completed right? Is there any detrimental effect if you shutdown a client with several WUs in this state in regards to points?

They could expire then you don't get the points, you can go to the project tab and select the project then click update and that will send them immediately. 

[leadeater]

Next 2 weeks is going to hurt lol

[IntMD]

1 hour ago, leadeater said:

Do you have an idea how that effects points generation, does it also limit you to what tasks you can get? Seems odd you can just end a task like that without a proper completion point. I'd rather not change it if it limits me to a much smaller subset of tasks.

Not a clue, as I haven't actually changed mine from the default and haven't really been on there for very long. I think the credits are just relative to the amount of work done. I think this link kind of explains what is currently happening https://boinc.bakerlab.org/rosetta/forum_thread.php?id=669&postid=10377#10377 but is super old.

[leadeater]

1 hour countdown warning, get ready to limit your network traffic.

[Windows7ge]

43 minutes ago, leadeater said:

1 hour countdown warning, get ready to limit your network traffic.

One hour for you maybe. I still have to wait like another 24hrs+.

[leadeater]

2 minutes ago, Windows7ge said:

One hour for you maybe. I still have to wait like another 24hrs+.

You mean their event page doesn't correctly adjust for client timezone? what?

[Windows7ge]

4 minutes ago, leadeater said:

You mean their event page doesn't correctly adjust for client timezone? what?

I had to do the conversion myself and I believe it translates to Mon @ 8:00PM. I'm EST. My WU due dates haven't reached that yet.

[leadeater]

Just now, Windows7ge said:

I had to do the conversion myself and I believe it translates to Mon @ 8:00PM. I'm EST. My WU due dates haven't reached that yet.

I was just planning to abort a bunch, not all of them but enough to see if I could still grab jobs.

[Windows7ge]

3 minutes ago, leadeater said:

I was just planning to abort a bunch, not all of them but enough to see if I could still grab jobs.

I know WCG did not like it when I aborted jobs. It'd suspend me for like an hour or if I dropped a lot of jobs I'd be looking at 6 or even 12.

[cbigfoot]

yeah i was tinkering and set 4 days yesterday and they all will expire before the start so will not be bunkering as i don't want to get banned.

then again my CPUs im using are create so many points lol

[leadeater]

7 minutes ago, Windows7ge said:

I know WCG did not like it when I aborted jobs. It'd suspend me for like an hour or if I dropped a lot of jobs I'd be looking at 6 or even 12.

Fortunately the jobs have long times on them so the number you get is low, I have completed a decent amount so it won't ruin my ratio too badly, just planning on killing the ones that don't quite make it to the event start time but the very close to it, then let what is running clear and likely the next 6-8 hour set clear and hopefully that will be a full 2.5 day queue that I can drop at event start.

 

My most recent hosts I've added I can't abort jobs at all, those will get blocked most likely.

 

Will try it on one server and see what happened, then one by one more until I get kicked in the nuts for being bad.

[leadeater]

Est credits I'll have built up according to boincviewer will be 7.8 million. It'll probably be less than that, hoping to hit 7 million.

 

Anyone know how Rosetta compares to WCG RAC wise? Specifically OpenZIKA, not that I expect anyone to remember that detail.

 

Winning Marathon points last year was 330 million for reference, I expect that to be much higher due to all the big CPU jumps and buying.

[leadeater]

Javelin has been announced, first round starts same time.

 

I know nothing about this project, CPU or GPU? Anyone?

[Windows7ge]

1 hour ago, leadeater said:

Fortunately the jobs have long times on them so the number you get is low, I have completed a decent amount so it won't ruin my ratio too badly, just planning on killing the ones that don't quite make it to the event start time but the very close to it, then let what is running clear and likely the next 6-8 hour set clear and hopefully that will be a full 2.5 day queue that I can drop at event start.

 

My most recent hosts I've added I can't abort jobs at all, those will get blocked most likely.

 

Will try it on one server and see what happened, then one by one more until I get kicked in the nuts for being bad.

I am just now starting to get jobs dated 5/4/20 9:05:17PM which should suffice but this means I will only have 1 hour of error to get everything submitted before it expires. I should be home at this time Monday so so long as I'm paying attention it should work out.

 

I'll sacrifice one VM to see how bad it is if I ditch all the jobs dated earlier.

[leadeater]

@Windows7ge

Do you know how to change the project resource share for just a single computer, not the whole project and everything running it.

[Windows7ge]

1 minute ago, leadeater said:

@Windows7ge

Do you know how to change the project resource share for just a single computer, not the whole project and everything running it.

Are you referring to Options -> Computing preferences... ?

 

It'll change your web preferences to local on just that machine allowing independent control.

[leadeater]

5 minutes ago, Windows7ge said:

Are you referring to Options -> Computing preferences... ?

 

It'll change your web preferences to local on just that machine allowing independent control.

Nah I just want to set the share so different projects and run at the same time on the same system nicely but only need this set one a single computer.

 

I found out how to do it though. Just create a file in the project directory called app_config.xml and place this in to it

<app_config>
<project_max_concurrent>4</project_max_concurrent>
</app_config>

 

Now that project will only run that many at once and then any other project that has tasks in the queue will run with the remaining slots available.

[Windows7ge]

1 hour ago, leadeater said:

Will try it on one server and see what happened, then one by one more until I get kicked in the nuts for being bad.

Alright, first offense, dropped 117 tasks:

Communication deferred 00:06:00

Not bad. Will try again on the next VM.

[leadeater]

@Windows7ge

NumberFields@home deadlines are much wider, maybe bunker those instead of Rosetta? Would probably be good to build up multiple VMs with competed work that we can drop at one of the throws we wish, might be more valuable than bunkering Marathon?