Formula Boinc 2018 Strategy

so my rig with 2 1080 TI finally finished all its GPU tasks.... I have been running the script on it since i got home from work yesterday but no new units for it :( I had downloaded a full batch but its chewed through them already . I still have CPU tasks but this is going to hurt my output a lot going forward
 
All 5k or so WUs I got when this thing was first announced is still crunching on my stuff.
 
I said we. Lol

They, as a team, are on top of these announcements as soon as they're announced giving them the advantage of crunching a 3 day sprint in 4 days instead of 3.
 
and I still don't get any assignment ... now on GPUGrid as fallback ... hope some WU will come soon'ish
 
Judging by thier forums this is a normal occurance for them to run dry on work. Generally takes a few days for things to return to normal.

IIRC, last years FB sprint for this project had the exact same issue and was dry during the whole Sprint. I think we only scored around 2M points for the entire duration.

It was one of the projects that taught me to stay on top of these announcements as soon as they are announced. If I can at least.
 
Looks like I've scored about 400,000 points so far from 856 validated tasks and have 1,484 tasks waiting to be validated. I guess when the initial batch of tasks I downloaded run out I'm done.
 
This is dumb. Doesn't help that this stuff is always started in the middle of my work day and I have no opportunity to get a jump on things.
 
Use BAM! and tell it to communicate with your hosts every hour the day before a sprint starts. Make sure you're registered with all possible sprint projects. Make sure you suspend/abort current running projects so you got a clean slate and ensure all your rigs are set for enough days to build a large enough queue.

Do that before its announced. Then when its announced log onto the BAM! web site with your phone and resume/attach/allow new work or whatever for the project announced.

Then the next hour when your hosts communicate with BAM! it'll get them started.
 
Looks like I'm half way through my queue already. If they don't generate some new tasks I'll probably be finished on Saturday.
 
Well...that is the problem with a challenge series trying to "support the smaller projects". They cannot handle the "support" being thrown at them without any notification.
 
May take Skillz approach a couple posts up.
Download it all, 10 day WU max per project, when a sprint is close.

Damn. I got all my rigs up too. Lol
 
I was fortunate enough to fill my boots with a thousand tasks per rig before going to bed on Wednesday night.

The AVX friendly intel rig ran out of work this afternoon but since then I've used the Sztaki VMs to keep it busy.

I've configured 4 Lubuntu VMs to 8 cores each with the watch command running every 30 seconds as there's no timeout on Asteroids. All 32 cores are now crunching again. I absolutely hate losing.

Looking at XS's Stoneageman's boxes, he seems to be running out of work. Maybe they all are …. glass half full and all that.
 
Oh snap you do have a bunker. Hell yeah.

I've got about 1k tasks completed already. 4k to go.
 
Skillz Naw just decreasing timout on script :p

Gonna laugh when we come back and beat them.

Edit: HERE COMES DA WORK!
 
Last edited:
atp1916 said:
Skillz Naw just decreasing timout on script :p

Gonna laugh when we come back and beat them.
Click to expand...

Provided we don't get banned for launching a DoS attack, lol. I just know XS will be doing the same thing.

Skillz said:
Oh snap you do have a bunker. Hell yeah.

I've got about 1k tasks completed already. 4k to go.
Click to expand...

No bunker this time I'm afraid. Figured that the sooner I was asking for more work the better. I didn't realise I'd only get GPU work when asking for both. The VMs are only getting a trickle but at least they're getting something.

Got over 1.8 Mil credits validated so far.
 
Last edited:
Great, one of my other systems just crashed and now I can't find any of the ~1100 tasks it downloaded. No fucking clue what happened. Looked at it in Teamviewer and seen some error about the 980Ti not being detachable or something so I rebooted the rig. Upond reboot ALL tasks for all projects are suddenly gone. Missing. Help?
 
Sorry, I've not seen this before. I've had both projects and tasks disappear before which was fixed with the boinc restart command but not just tasks.

A quick google turned up someone on Seti that had to reinstall Boinc to see his tasks again but this sounds a bit dangerous to me.

Perhaps Gilthanis has seen this before
 
Well upon restarting the computer and restart in bionic it started reattaching all of my projects so somewhere along the way something got f***** up and everything got deleted and restarted
 
If it's your E5-4620 then, yes, you're stuffed I'm afraid mate. It shows a whole bunch of tasks as abandoned.
 
1526 validated tasks uploaded earlier. Good for 732,480 points. Got 669 tasks waiting validation and 2061 tasks in the works.

Lost 1043 tasks when a GPU decided to die apparently. So I am down a GPU as well.
 
wow Skillz that blows.

It just doesn't seem like intended behaviour though unless the tasks are somehow associated to the individual compute units before actually starting them.

Wonder if it's a bug, basically.
 
I'm not sure. Everything was running fine, but when I looked in BOINCTasks I noticed one of the GPU tasks didn't seem to be running. It said "Running" but it had nothing for a completion time, but the other one did. So I looked at the rig in Teamviewer and seen a windows error about unable to detach a 980 Ti or something along those lines. So I just restarted the rig. That caused BOINC to reattach ALL the projects. All my previous settings were gone. It started grabbing work from everyone. Those Asteroid@home tasks were gone as well as the WUProp and Goofxy tasks. All gone.

The rig currently has 6 GPU tasks, but it's down 1 GPU so it's only crunching on 1 980Ti now. I think I lost around 400 completed tasks. ~150 CPU and ~250 GPU. Nearly 200k points, but judging by the current status it wouldn't have made a dent in the spread XS has on it.

Bunkering is important. ;)
 
So the fact there is no new work is send out is because all run big bunker and sucked the project empty ...
 
Actually no. I don't think that's the problem. I looked at some of the big hitters that have been producing a lot of points and they've all recently connected to the server. They have outstanding validation pending tasks and a lot of validated tasks that have been validated since the Sprint started.

The project is just unable to keep up with the load. It's happened before in the past without a large load such as this. Project admin blames it on not enough storage on the server and said the server will get upgraded this Summer.
 
now my rig with 2x 1070 is done all its tasks for gpu :( really sad that i couldn't see my full output on this project. I have been running the script for 24 hours and i couldn't get any new gpu tasks... super choked.

And i was ready and bunkered the full 10 day 22 hours before sprint drop
 
It seems the server can only handle ~300k tasks at any given time due to space restraints. Then the server's space is declining due to results coming in and thus that ~300k tasks limit is slowly decreasing as well. Seems to be around 230k limit now and steadily declining. By the end of the sprint I predict no new tasks will be sent at all and everyone with tasks wont be able to return anything because the server will be too full to accept completed work.

The admin of the project needs pay attention more frequently to the servers load.
 
Which seems to be close to on par with points. In 1 day the server's task genration has dropped by 100k tasks and the project has easily done 48 million points in that time frame.
 
Gilthanis said:
For those interested for future reference on SETI and whether the lunatics optimizations still apply
https://setiathome.berkeley.edu/forum_thread.php?id=82903&postid=1936644
Click to expand...

That is just one big mess of a thread. I see they adopted the Lunatics apps as the default for Windows. It doesn't mention if they have adopted the Linux Tbar apps. What I also understand from that thread is that credit per task has no basis in reality at Seti, and that RAC has gone down steadily this year by 30% or more.
 
I haven't finished all the posts in that thread yet. It was break time reading at work. But CreditRandom is certainly above my head if it actually does work. I've just not had much luck with it when trying to run tests for numbers.

phoenicis , I am not sure what happened with Skillz work units. That is certainly odd. However, I will say that my Intel 2P just randomly decided to detach from BAM! today. When I re-added it, about 5 projects got re-added that I did NOT manually remove. I've seen BAM! randomly detached from systems in the past but they stayed connected to all of their projects. I'm wondering if something on the back end was getting mucked with......

Oh and Asteroids was one of the projects it removed..
 
That Seti thread pointed out that the same tasks get different points depending on what system it is ran on, and what system the wingman is using. It is not consistent at all. It also doesn't take into account whether or not the host is using optimized apps (which the project is handing out by default now). I expected when I got to the end of that thread that everyone would agree that the tasks need to award a static score, but they are still discussing it.
 
You must log in or register to reply here.
Back
Top