BOINC Pentathlon 2024

[leadeater]

Guess I will go back home,  managed to find two 2x 5220 servers and get them plugged in but if RNMA is going to run out of jobs to send there isn't much point me sticking around trying to find something else in the junk pile and give it CPR.

[leadeater]

44 minutes ago, GOTSpectrum said:

Anyone else having issues getting WUs?

Looking at logs on one of my systems I'm also not getting any new tasks. Up shot is I have a lot of them and should last me out until the end.

 

Edit:

Ah I see, actually 0 to send. The other one I don't think is actively giving out and not what we have been running

[GOTSpectrum]

2 minutes ago, leadeater said:

Guess I will go back home,  managed to find two 2x 5220 servers and get them plugged in but if RNMA is going to run out of jobs to send there isn't much point me sticking around trying to find something else in the junk pile and give it CPR.

I've just added PG too so my system is being used one way or another 

[leadeater]

Just now, GOTSpectrum said:

I've just added PG too so my system is being used one way or another 

Have your setup your PG preferences correct for the number of threads and L3 cache in your CPU? If not you'll have run times in the days 

[GOTSpectrum]

10 minutes ago, leadeater said:

Have your setup your PG preferences correct for the number of threads and L3 cache in your CPU? If not you'll have run times in the days 

Uhh... Sure... 

 

*Googles*

[SkillzTA]

8 minutes ago, GOTSpectrum said:

Uhh... Sure... 

 

*Googles*

You need to allocate 25MB of L3 cache to a single Primegrid CUL task. You should assign them the number of cores (8 is about the max of not being any faster) as that CPU has L3 cache space. For AMD CPUs you need to see where the L3 cache falls within' the CCD or CCX. Zen2 and lower each CCD has 32MB split between 2 CCX's with 16MB each. These CPUs (and older then Zen2) are not that great at PrimeGrid CUL, but they're not bad either. So you'll want to assign 1 task per CCD (8 cores, 16MB L3 x2) and for Zen3 and newer its 1 CCD and 1 CCX with 32MB, that are 8 cores total. You'll want to assign the tasks to run on 8 cores and within' a CCD domain.

For Intel it's a little easier as the L3 is shared across all cores. So just assign it enough threads to fill the L3 of the processor. Disable the E-cores.

 

Windows you'll want to use process lasso to assign tasks to specific cores/ccd/ccx and Linux you'll want to use a utility called taskset. Luckily these tasks take a long time so you could manually just make sure the tasks are assigned to a specific set of cores with a common shared L3 cache domain.

[GOTSpectrum]

7 minutes ago, SkillzTA said:

You need to allocate 25MB of L3 cache to a single Primegrid CUL task. You should assign them the number of cores (8 is about the max of not being any faster) as that CPU has L3 cache space. For AMD CPUs you need to see where the L3 cache falls within' the CCD or CCX. Zen2 and lower each CCD has 32MB split between 2 CCX's with 16MB each. These CPUs (and older then Zen2) are not that great at PrimeGrid CUL, but they're not bad either. So you'll want to assign 1 task per CCD (8 cores, 16MB L3 x2) and for Zen3 and newer its 1 CCD and 1 CCX with 32MB, that are 8 cores total. You'll want to assign the tasks to run on 8 cores and within' a CCD domain.

For Intel it's a little easier as the L3 is shared across all cores. So just assign it enough threads to fill the L3 of the processor. Disable the E-cores.

 

Windows you'll want to use process lasso to assign tasks to specific cores/ccd/ccx and Linux you'll want to use a utility called taskset. Luckily these tasks take a long time so you could manually just make sure the tasks are assigned to a specific set of cores with a common shared L3 cache domain.

So on my 5950x I want to run two 8 threaded tasks?

[leadeater]

1 minute ago, GOTSpectrum said:

So on my 5950x I want to run two 8 threaded tasks?

Yep, the other thing you could do is bunker NFS since we are going to need to do the next one

[GOTSpectrum]

Just now, leadeater said:

Yep, the other thing you could do is bunker NFS since we are going to need to do the next one

Will see how this goes first

[leadeater]

@SkillzTA

Hah! We can throw a Javelin further

 

[GOTSpectrum]

Here's the update for the last hour, We outproduced SG by 14k 

 

[leadeater]

Well that's not so good....

 

[GOTSpectrum]

2 minutes ago, leadeater said:

Well that's not so good....

 

oof 

[GOTSpectrum]

Checking in for the hourly update and we outproduced SG by 20k this update, keep up the good work guys 

[Kilrah]

6 hours ago, will0hlep said:

Crazy (and probably worthless) plan that’s just come to mind at 1AM: Could I fold and/or Boinc on my NAS (runs TrueNAS Core) by setting up a Windows (or Linux) VM?

Way simpler, there's a Docker container, here's all it needs from the template I have on Unraid:

 

 

docker run --name='Boinc' -e 'BOINC_GUI_RPC_PASSWORD'='blah' -e 'BOINC_CMD_LINE_OPTIONS'='--allow_remote_gui_rpc' -v '/mnt/user/appdata/boinc':'/var/lib/boinc':'rw' 'boinc/client'

[Kilrah]

Waking up and it seems I'm running out just now on the desktop... looking at the power graph it didn't fall to idle during the night

 

 

 

Laptop for some reason had downloaded like 1000 at a time the other day so it's got plenty of work still

[GOTSpectrum]

Update on the city run, we outproduced SG by 26k points, we gotta give it everything we've got and we could be taking home a medal. 

[TheLANguy]

3 hours ago, leadeater said:

Have your setup your PG preferences correct for the number of threads and L3 cache in your CPU? If not you'll have run times in the days 

Where in PrimeGrid's settings do you change that?

[leadeater]

1 minute ago, TheLANguy said:

Where in PrimeGrid's settings do you change that?

 

 

 

 

 

 

[kiska3]

YIPPEEEEEEEEE!

[Kilrah]

Bunkering all the NFS units the server wanted to give me on the desktop now, will switch to PG when that's done since my CPU's not a... prime choice for that project

[GOTSpectrum]

1 hour ago, GOTSpectrum said:

Update on the city run, we outproduced SG by 26k points, we gotta give it everything we've got and we could be taking home a medal. 

the latest update on the city run, we outproduced SG by a further 25k points this hour 

[Kilrah]

I just dropped my laptop's production during the night (was offline) of 200 units, not much but meh... got another 350 left on it which will last till tonight, the ~40 left on the other machines will be done within an hour or 2

 

Oh NFS actually needs 900MB of RAM per task, gotta watch usage if you have a high thread count CPU, suddenly saw RAM usage way higher than normal

[GOTSpectrum]

29 minutes ago, Kilrah said:

I just dropped my laptop's production during the night (was offline) of 200 units, not much but meh... got another 350 left on it which will last till tonight, the ~40 left on the other machines will be done within an hour or 2

 

Oh NFS actually needs 900MB of RAM per task, gotta watch usage if you have a high thread count CPU, suddenly saw RAM usage way higher than normal

Glad I have 128gb of ram haha (over 22 cores)

 

45 minutes ago, GOTSpectrum said:

the latest update on the city run, we outproduced SG by a further 25k points this hour 

We once again outproduced SG this update, by 20k points 

[kiska3]

10 minutes ago, GOTSpectrum said:

56 minutes ago, GOTSpectrum said:

the latest update on the city run, we outproduced SG by a further 25k points this hour 

We once again outproduced SG this update, by 20k points 

Very good!