TPU's WCG/BOINC Team

[Deleted member 202104]

Day Six of the Time 2022.02 challenge complete.

Looks like we might have this one in the bag. Steady at #1.

The Top 10 for today:

-----------------------------

Day One of the Compute for the Cure Jan 2022 complete.

Starting to see a pattern with these... We're #1.

The Top 10 for today:

Quick link to the Challenge Stats:

Team Challenge Detail

What if you could support causes you care about while reading this post? Your device’s unused computing power can help scientists tackle some of the world’s biggest problems: from finding better #HIV and #cancer treatments to more efficient sources of renewable energy? I just joined World...

www.worldcommunitygrid.org

 

[80-watt Hamster]

Day Six of the Time 2022.02 challenge complete.

Looks like we might have this one in the bag. Steady at #1.

The Top 10 for today:

View attachment 232554
-----------------------------

Day One of the Compute for the Cure Jan 2022 complete.

Starting to see a pattern with these... We're #1.

The Top 10 for today:
View attachment 232555

Quick link to the Challenge Stats:

Team Challenge Detail

What if you could support causes you care about while reading this post? Your device’s unused computing power can help scientists tackle some of the world’s biggest problems: from finding better #HIV and #cancer treatments to more efficient sources of renewable energy? I just joined World...

www.worldcommunitygrid.org

Yeah, it seems like the big hitters don't bother to join the challenges for some reason.

 

[phill]

Yeah, it seems like the big hitters don't bother to join the challenges for some reason.

I do wonder if they don't because of how big a contribution they can make... It's not exactly I guess, 'fighting fair'

I hope everyone is doing well??

 

[Deleted member 202104]

I do wonder if they don't because of how big a contribution they can make... It's not exactly I guess, 'fighting fair'

I hope everyone is doing well??

That's a really good point. I guess it really hadn't set in with me how many points/results and runtime the team produces in comparison.

I'm always up for a good challenge but I suppose, as they say, no one likes a showoff. :shrug:

 

[Caring1]

I noticed I actually have one HSTB task
Estimated run time for it is just over 9 hours for one task

 

[phill]

I noticed I actually have one HSTB task
Estimated run time for it is just over 9 hours for one task

Some of the Africa tasks I get take about 12 to 14 hours..... Wish I could actually get some of those tasks, I'd like to get myself up to 2 years on the project before they close it down, whenever that might be....

 

[80-watt Hamster]

Some of the Africa tasks I get take about 12 to 14 hours..... Wish I could actually get some of those tasks, I'd like to get myself up to 2 years on the project before they close it down, whenever that might be....

My pool seems to get an ARP every day or three on each machine, but never more than one assigned at any given time to a specific client.

 

[Deleted member 202104]

My pool seems to get an ARP every day or three on each machine, but never more than one assigned at any given time to a specific client.

Do you have them limited in your device profile? I have mine set to 8 maximum at a time and usually have 6-8 on each machine that use that profile.

 

[mstenholm]

Do you have them limited in your device profile? I have mine set to 8 maximum at a time and usually have 6-8 on each machine that use that profile.

View attachment 232841

You have the hardware to support that setting now that you abandoned Rosetta. I remember that my old i7 plus 8 GB struggled with more than 2-3 if I was using the PC. I settled for the five year badge with a slight 4 day margin.

 

[Deleted member 202104]

You have the hardware to support that setting now that you abandoned Rosetta. I remember that my old i7 plus 8 GB struggled with more than 2-3 if I was using the PC. I settled for the five year badge with a slight 4 day margin.

Very true - I think I remember reading that the maximum recommended number of ARP is no more than your physical cores. Memory requirements are higher as well.

I did suspend Rosetta for the WCG challenges we entered, but after seeing the results I resumed Rosetta yesterday. Nothing but the 'VirtualBox is not installed error' so far. I think the team at U of W just doesn't want to send me any jobs.

 

[mstenholm]

Very true - I think I remember reading that the maximum recommended number of ARP is no more than your physical cores. Memory requirements are higher as well.

I did suspend Rosetta for the WCG challenges we entered, but after seeing the results I resumed Rosetta yesterday. Nothing but the 'VirtualBox is not installed error' so far. I think the team at U of W just doesn't want to send me any jobs.

You have to strike at the right time and that was yesterday evening, my time. I got more than I can handle since I changed the run-time from 8 to 12 hours. Ups.

 

[phill]

Guys I'm having a senior moment I think but I just wished to check in with you...

I've got some OPNG work units in and they are ready to go. Sadly for some reason today, they only seem to be completing one at a time. Yesterday they where doing 12 at a time as I had it set in the settings when we did the app_config and cc_config setup back along.

Would anyone know why or what would be stopping it carry on with the GPU work at all??

Here's my config files as they where yesterday -
app_config

<app_config>
<app>
<name>opng</name>
<max_concurrent>12</max_concurrent>
<gpu_versions>
<gpu_usage>0.0833</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

cc_config
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Am I missing something or am I having a senior moment here??

 

[Jacky_BEL]

app_config.xml looks OK.

I don't really know what cc_config does , but syntax also looks OK.

 

[mstenholm]

Guys I'm having a senior moment I think but I just wished to check in with you...

I've got some OPNG work units in and they are ready to go. Sadly for some reason today, they only seem to be completing one at a time. Yesterday they where doing 12 at a time as I had it set in the settings when we did the app_config and cc_config setup back along.

Would anyone know why or what would be stopping it carry on with the GPU work at all??

View attachment 232849

View attachment 232850

Here's my config files as they where yesterday -
app_config

<app_config>
<app>
<name>opng</name>
<max_concurrent>12</max_concurrent>
<gpu_versions>
<gpu_usage>0.0833</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

cc_config
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

Am I missing something or am I having a senior moment here??

The app_config looks ok. The last six lines isn’t a part of it, right? Ninja…
Run the file and ask for work or re-start if that doesn’t work.

 

[phill]

Tried shutting down Boinc and restarting it but that came back with the same issue... Tried a restart (cause in IT, that fixes everything more often than not!!) but sadly that's not worked either... I'm getting this now -

I think it started last night/this morning after a reboot as I noticed I had a few jobs to do but they were just sat there... I've checked over the preferences...

Couldn't see anything in there that would have caused anything that I'm aware of.... Rarely look in there as it is anyways...

I can't see what is going on there either....

Also noticed that the GPU is not constantly under load via MSI...

Am I being stupid or is there an issue there? Is anyone else getting this sort of thing at all?

 

[mstenholm]

The fluctuating load is normal when you run one or few units. The rest of the settings look ok. All that I can suggest is to try 10 and 0.1. The PC is running the correct program, it just read the numbers wrong. This is the best I can do.

 

[Jacky_BEL]

Aren't the maximum number of tasks already running, and is it showing the tasks that are ready to run next as soon as another task is completed?

You can switch between views with a click on the button on the left side to :

"show active tasks" vs "show all tasks"

 

[phill]

The fluctuating load is normal when you run one or few units. The rest of the settings look ok. All that I can suggest is to try 10 and 0.1. The PC is running the correct program, it just read the numbers wrong. This is the best I can do.

Sadly I've just ran out of the OPNG tasks for the moment, but I've changed it to 10 just for a trial and to see if that makes any difference...

Aren't the maximum number of tasks already running, and is it showing the tasks that are ready to run next as soon as another task is completed?

You can switch between views with a click on the button on the left side to :

"show active tasks" vs "show all tasks"

Well that's the thing, before hand, when the tasks would come, they'd put other tasks on hold, so I'm guessing like a priority type thing? At the moment, this doesn't seem to have been happening and has just done the one at a time... I'll see what happens I'll keep my eye on it... If anyone else gets the same sort of thing happening, please do let us know back here

 

[mstenholm]

My two 2xOPNG GPUs are still running like that and my 8x has dried out and has been from before your first post. I will of course report if I see a similar behavior

 

[80-watt Hamster]

Do you have them limited in your device profile? I have mine set to 8 maximum at a time and usually have 6-8 on each machine that use that profile.

View attachment 232841

Well, it was limited to whatever Maximum Output sets. I'd never dug into the Custom profile until now. Although I think all machines are set to use local prefs (although maybe that only applies to schedule?), so would editing web profile have an effect?

 

[Deleted member 202104]

Well, it was limited to whatever Maximum Output sets. I'd never dug into the Custom profile until now. Although I think all machines are set to use local prefs (although maybe that only applies to schedule?), so would editing web profile have an effect?

I think I remember that when you opt in to ARP, it defaults to one concurrent job because of the greater system requirements. I can't find anything on the WCG site about it, so I may not know what I'm talking about (wouldn't be the first time).

I think it's either/or for web or local profile. If you choose web it ignores anything you've changed in the client settings.

I moved to using web profiles so I could pick and choose without touching the computer.

 

[phill]

My two 2xOPNG GPUs are still running like that and my 8x has dried out and has been from before your first post. I will of course report if I see a similar behavior

Well, it was limited to whatever Maximum Output sets. I'd never dug into the Custom profile until now. Although I think all machines are set to use local prefs (although maybe that only applies to schedule?), so would editing web profile have an effect?

I couldn't see anything to maximum output set or anything like that, I know there's something there for the how many types of tasks do you wish to have but nothing more that I could see...

I think I remember that when you opt in to ARP, it defaults to one concurrent job because of the greater system requirements. I can't find anything on the WCG site about it, so I may not know what I'm talking about (wouldn't be the first time).

I think it's either/or for web or local profile. If you choose web it ignores anything you've changed in the client settings.

I moved to using web profiles so I could pick and choose without touching the computer.

I normally just leave it to whatever if I'm honest, I don't tend to worry about the tweaking and such, I just crunch whatever comes down Here's hoping it's sorts itself out a little quickly... I'm on the hunt for points!!

 

[Jacky_BEL]

I am still puzzled why 3 of my machines won't run OPNG's.
Same program configurations, same profiles and it just won't run them.

 

[80-watt Hamster]

I am still puzzled why 3 of my machines won't run OPNG's.
Same program configurations, same profiles and it just won't run them.

Connect them to F@H instead? My impression from some of @mstenholm's comments (though I can, as always, be wrong) is that there's generally less OPNG work than there are clients to run them. May as well put them to good use somewhere.

(I realize that's not actually a solution to the stated problem.)

 

[mstenholm]

Connect them to F@H instead? My impression from some of @mstenholm's comments (though I can, as always, be wrong) is that there's generally less OPNG work than there are clients to run them. May as well put them to good use somewhere.

(I realize that's not actually a solution to the stated problem.)

A small number of crunchers from two teams uses cron and Windows Scheduler to query for work every 3-5 minute. They have a great number of decent hardware but they have decreased their involvement in WCG when the nature of jobs increased the run-time without a similar increase in points. When the jobs return to normal they might return and then we will be back to the few jobs again. My low-end GPUs have had jobs more or less constant the past week but my 2070S spend 90-95% of its time in 2D clocks.

F@H is a better bet if you want to stay warm during the winter.