Info 12th BOINC Pentathlon 2021

[bwoochowski]

@Icecold

I have it set to generic, all the other parameters I configured got applied.

What do you typically see for GPU utilization in E@H? Mine fluctuate between 30 and 80%. Takes about 35 minutes to finish one GW WU.

Forgot to add that I had selected to only use AMD GPUs:
0.9 CPUs + 1 AMD/ATI GPU

 

[TennesseeTony]

I have had that problem before with Einstein....I just kept aborting them for a day or so till it sorted itself out and sent me the correct tasks.

 

[bwoochowski]

Could it be a new member thing? Maybe I need to hit a certain threshold in verified units before I can get different WUs? Currently have 7 waiting and 0 verified.

 

[Endgame124]

boinc on my Fedora box does not detect my GPU. Never really noticed before since I've been using it for F@H which works fine. So... what do I do to tell boinc that I actually have a GPU in the box?

 

[StefanR5R]

Did F@h run only the OpenCL application, or was it able to run its CUDA application too?

 

[Endgame124]

Did F@h run only the OpenCL application, or was it able to run its CUDA application too?

F@H runs Cuda WUs - runs about 2.4M ppd.

Boinc sees the drivers - the startup log shows something roughly like “NVIDIA GPU drivers found, but no usable GPUs”.

for now I’ve restarted F@H. I googled this last year now that I look at my search history and got nowhere

 

[Icecold]

@Icecold

I have it set to generic, all the other parameters I configured got applied.

What do you typically see for GPU utilization in E@H? Mine fluctuate between 30 and 80%. Takes about 35 minutes to finish one GW WU.

Forgot to add that I had selected to only use AMD GPUs:
0.9 CPUs + 1 AMD/ATI GPU

I'm generally seeing 98-100% utilization. That's running the Gamma-ray Pulsar Binary Search #1 WU's, but 2 of them at a time. It looks like it's pretty close to 8 minutes task time(but 2 are running in that 8 minutes). A single task runs a bit under 5 minutes.

 

[StefanR5R]

@Endgame124, post the startup part of the client log and the entire coproc_info.xml, here or somewhere befitting.

--------

The third day of Javelin Throw has been announced, post #1 is updated.

 

[cellarnoise]

Crayon eater here. Oh, correct that, Cellarnoise2. Please keep info of all kinds coming. I am learning. Details such as what to do post twice. I need to get an easy time shift task tracker here. Time is hard.

I probably with my limited hardware and human abilities will and should just focus on Sidock.... I had an old lappy running the specific WCG tasks, and turned out I uploaded my puny bunker at the wrong time and it was probably the bratworstest puter to process these tasks.....

Still learning at the expense of electricity and planet.

 

[Markfw]

I have not updated my WCG config, I will just let it run all projects I guess. I have 500 total threads on WCG, and a fair amount on MBIP

 

[TennesseeTony]

Attention all TeAm Members....message me for a link to our private communications area.

Attention all non-TeAm members....we have a public section too, some good info, message me.

Attention @VietOZ , I lubz yew and I missus yew!

 

[Skillz]

I have not updated my WCG config, I will just let it run all projects I guess. I have 500 total threads on WCG, and a fair amount on MBIP

It takes like 30 seconds to only select MIP in your preferences.

Home

World Community Grid enables anyone with a computer or Android device to donate their unused computing power to advance cutting-edge scientific research on topics related to health and sustainability.

www.worldcommunitygrid.org

Unless the stats are updating.

 

[Markfw]

It takes like 30 seconds to only select MIP in your preferences.

Home

World Community Grid enables anyone with a computer or Android device to donate their unused computing power to advance cutting-edge scientific research on topics related to health and sustainability.

www.worldcommunitygrid.org

Unless the stats are updating.

Done... But are the javalin throws like a 1 day event ?

 

[cellarnoise]

Yes, Javelin's are short time frame with like a day in between. You have to stay on top of them and then adjust for your time zone as I am learning in between slick fiber intake. Or else you eat crayons and drop your bunker or tasks after everyone has left the field.

I used to throw javelin and once "did" track events in my high school days. Before crayons.... So showing up at the field in the middle of the night and throwing into the wind for no-one is embarrassing .... Where is the wax? Hungry...

 

[TennesseeTony]

Lol, I still don't quite know what to think of you @cellarnoise!

 

[cellarnoise]

Prime grid or not: If you wish to be: or not; be somebody Sorry for the link:

I want to be some DC, Be come DC too! ?

 

[bwoochowski]

After a little tweaking one PPS Sieve WU now takes about ~ 310 seconds, not great but better. Also, after completing a few more E@H WUs several FGRP WUs got downloaded.

If anyone else is having issues with contention between cmdock and pps sieve/e@h on Linux I used these commands to set the affinity and it keeps the GPU jobs well fed:

#!/bin/bash
#Set affinity for cmdock
core=0; for pid in $(ps -u boinc -o pid,cmd r | grep 'build/cmdock' | awk '{printf "%s ", $1}');
    do taskset -apc "$core" "$pid" && core=$((core+1));
done

#Set affinity for PPS Sieve
for pid in $(ps -u boinc pid,cmd | grep tpsieveOCL | awk '{printf "%s ", $1}');
    do taskset -apc "$core" "$pid" && core=$((core+1));
done

 

[Skillz]

Done... But are the javalin throws like a 1 day event ?

Yes they are. The next one starts early tonight. Might as well switch to MIP only to clear your cache so you'll be ready for the next 4.

 

[Markfw]

Yes they are. The next one starts early tonight. Might as well switch to MIP only to clear your cache so you'll be ready for the next 4.

clear my cache ??

 

[StefanR5R]

…to get any queued MCM/ OPN etc. completed before the MIP event, is what @Skillz is referring to.


On small and medium hosts which can queue work for more than a day, you can also

• switch the web preferences to MIP exclusive,
• increase the client's buffer to something like 2 days and thus pull in >1 day worth of MIP work,
• then manually suspend each MCM/ OPN/ etc. work at the beginning of the MIP event,
• resume this other work after the event.

This is going to work only with a sufficient work buffer size, because the client will not request new work as long as some tasks are suspended. On a host with many threads, perhaps anything with 64 or more threads, may be unable to have a >1 day deep buffer of MIP work since the tasks aren't very large.

Or you could also abort + report MCN/ OPN/ work which you haven't started yet, if there are not very many of them queued in the client.

 

[Markfw]

OK, 271 MIP tasks running, and queue is full.

and 533 sidock threads going. 5 computers turned off due to heat, 2 2990wx's, 2 3900x's and a 1950x threadripper, or 144 idle.

 

[bwoochowski]

Both of my 1950s will also be idling for the afternoon as it's forecast to hit the upper 90s. Have a bunch of PPS Sieve WUs ready to drop at 0z.

 

[StefanR5R]

I am managing the cooling by means of open windows, as usual. Granted, temperatures are rather low here for now, lower than I recall from other Pentathlons. And until today, strong wind helped to carry waste heat out of the apartment too.

For the rest of the Pentathlon, my apartment will be filled with the scent from a large lilac bush which lives right outside my living room. :-)

 

[bwoochowski]

480 PPS WU stuck in pending since 0Z, this normal?

 

[Skillz]

Pending validation/verification? If yes, then yes, that's normal. You've gotta wait for your wingman to return his/her result to get it validated.