9th Annual BOINC Pentathlon 2018 Strategy & Preparation

ChelseaOilman, are your GPUs doing anything?

atp1916, are you still working on an asteroid bunker? If yes, keep them steady. If they're outta work switch to PG.

We'll have to see where everyone goes. Most likely, use your GPUs on PG. It'll be better there, but let the asteroids bunker go until PG starts.
 
My older slower GPUs are doing PrimeGrid alone as they don't burn through them before their next chance to connect to the internet for 3 hours during the night. My 1080s and 1080 Ti have multiple projects going and still occasionally run out of work to do before the next chance to connect to the internet.
 
Asteroids bunker not going well. Pretty sure I messed something up on my way out from work earlier this morning. They downloaded virtually no tasks all day long. :/

Anyhow, I am just going to set these guys to PrimeGrid some time tomorrow and we will go from there.

My CPUs are still on Numberfields...
 
ChelseaOilman said:
My older slower GPUs are doing PrimeGrid alone as they don't burn through them before their next chance to connect to the internet for 3 hours during the night. My 1080s and 1080 Ti have multiple projects going and still occasionally run out of work to do before the next chance to connect to the internet.
Click to expand...

PrimeGrid upped the allowed tasks per GPU to 1000 on PPS Sieve. Maybe you won't run out of work anymore.
 
MN Scout said:
PrimeGrid upped the allowed tasks per GPU to 1000 on PPS Sieve. Maybe you won't run out of work anymore.
Click to expand...
If that's the case I should probably stick to just PrimeGrid then. I have the BOINC client set up to get 2 days worth of work on my 1080 Ti and have Einstein, GPUGrid and PrimeGrid selected. Clearly I'm not getting 2 days worth of work for each project. Seems like the BOINC client sees the other projects and limits me.
 
BOINC can get weird with multiple Projects for sure, and trying to balance scheduling the tasks. For me I lowered the number of days of tasks stored down to 0 for PrimeGrid. I've seen that messes up Seti, not sure about other projects. That doesn't work for your situation with limited network connection.

Their old limit for PPS sieve was 100 tasks per GPU, they upped it to 1000.
 
Last edited:
My last second Rosetta bunker was good for ~225K I think. Better than nothing! Forgot all of the recent 2P-5118 boxes aren't on BOINC Tasks, so they're still banging away on Numberfields.
 
ChelseaOilman said:
If that's the case I should probably stick to just PrimeGrid then. I have the BOINC client set up to get 2 days worth of work on my 1080 Ti and have Einstein, GPUGrid and PrimeGrid selected. Clearly I'm not getting 2 days worth of work for each project. Seems like the BOINC client sees the other projects and limits me.
Click to expand...

I had completely filled my bunker with asteroids and the event log says my Nvidia GPU cache is full when i try to download more tasks for primegrid. This means i can't really bunker primegrid unless i dual boot or something. I am about 2 hours from finishing all my asteroids so i guess i am just going to let primgrid run and upload as she goes. hopefully alot of them will be pending validation for a bit...
 
Moved all the 5118's and the 8867 box to Rosetta; or at least they will be once BAM! updates. Wit the exception of the 4110's, which are staying on Universe, everything should be on Rosetta now.
 
I still have some but can’t open the pipe from work ... hope the server are not cancel them as I have seen some in the past. Some are 24hours and still crunching
 
Last edited:
ChristianVirtual said:
I still have some but can’t open the pipe from work ... hope the server are not cancel them as I have seen some in the past. Some are 24hours and still crunching
Click to expand...
not happy with the outcome ... a number of WU got cancelled by server or aborted.
I dumped quite a number of tasks and reloaded Rosetta with reduced duration.
 
Also, for anyone planning on running PrimeGrid, make sure to load up a lot of work now. The PPS sieve has zero chance of finding an actual prime so no glory for turning in first. However, they do need wingmen. So, increase your validation chances by having your work pulled early.
 
OK my issues with getting more primegrid are over after i let it run and upload overnight. I was able to download 1500 or so WU this morning and then i grabbed a few days of rosetta and i am going to bunker it all for now. also got the primegrid app config running and this is making a big difference!
 
yeah..the server will start you out with a smaller limit of work units until your host has returned a few to indicate it is reliable enough to run the work. They have highly tweaked the server to really know what your system can do and will tailor the work units to it all automatically if you don't interfere. That is probably why you can download more work now.
 
Something for us to test is another bunker tactic for future events. Apparently if you set your upload rate to something horribly slow, some projects will time out trying to upload work rather than move at a snails pace. Food for thought.
 
why not just UNC path each one on your network? Copy and paste the same hosts file to each one...done. Easy peasy.
 
Gilthanis said:
why not just UNC path each one on your network? Copy and paste the same hosts file to each one...done. Easy peasy.
Click to expand...

Or that. BoincTasks has spoiled me. :p

If there is one thing i dislike, however - it is "per host" changes. If i have to go to each host and flip levers on or off to 'make stuff work', i get irritated with the act in very quick order.

Gonna try blocking IPs at the router level. Hopefully that works.
 
I suspended some projects to get my full ~1000 task limit on Primegrid. They're almost full and I'll be closing off the pipe.

atp1916 what router do you have?

edit
Just seen your post. I block all my projects on my router. I have a EdgeRouter Lite.
 
Decided the bunker from one of my rigs wouldn't really be all that impressive so decided to try it out... I can confirm that setting my upload speed to 0.01 KB/s in the BOINC preferences allowed me to pull work from another project that I had not loaded up yet on it and not upload. Numberfields and PrimeGrid are on the box and is currently downloading some rosetta.
 
Definitely reduce the length of time. I put a link in the first post of Pentathlon thread for easy reference. Default setting is 8 hour tasks. Points scale. I don't know if exact but they do scale. This way you don't have a lot of long running tasks at the end and it also allows for quick project switching.
 
I was mostly just wondering how fast your army was updating in the stats.

You should be able to download a lot of tasks unless limited on disk space.
 
Take a look at my two rigs performance over the past few hours. I'm getting much higher credits with 2-hr run task without bunkering and limited tasks being downloaded. Not sure how the credit works as I tried to figure this out a few months back but was not successful. Not sure how many are seeing this. Maybe extra random credit for quick return of result over the weekend only?

https://boinc.bakerlab.org/rosetta/results.php?hostid=3227492&offset=0&show_names=0&state=4&appid=
https://boinc.bakerlab.org/rosetta/results.php?hostid=3213122&offset=0&show_names=0&state=4&appid=
 
when we keep crunching we might reach #1 in FB, too (at least for a while)

Screen Shot 2018-05-12 at 20.12.10.png
 
You must log in or register to reply here.
Back
Top