TPU's WCG/BOINC Team

[windwhirl]

Well, I don't know about that. My client just runs by itself for at least 8 to 10 hours a day of the 12 or so that is likely to be active. Basically I leave for work and leave my computer doing work. The only thing that could be out of the norm is that it's set up to ask for work units that can last for around 48 hours.

 

[phill]

Interesting. On the same network, one rig has a full queue, the other has absolutely nothing.


If I may make a suggestion... I read somewhere before that doing things like hammering the Update button, or setting WCG up to report completed work units immediately (thus forcing a connection every time a work unit is completed) are actually quite bad for the servers. Apparently, it's a lot easier for the servers to maintain a connection once in a while, whenever WCG is configured to communicate with the servers naturally, than it is to be constantly bludgeoned with forced updates, whether it's the user hammering the button or the advanced setting that forces WCG to communicate as soon as a work unit is completed. Maybe if we all let WCG run naturally it would work out better?

I think its worth a try but I have noticed that sometimes re-tries are hours not just a few minutes and when you have no work and a load to download but can't its frustrating so I know I'm guilty of retrying perhaps a little more than I should.. That said, I do forget and sometimes that just works too... Its a hit and miss for sure....

 

[hat]

Yesterday my main machine had no work. Today I've got loads.

@phill I think that's because, as the client fails to communicate to the server, it sets a time to try again. As it continues to fail, the time increases. I think the way to go is to increase "Store up to an additional x days of work" option to a reasonable amount, so you still have work even if the servers are on the fritz. Personally, I use 3 days for now.

 

[Nordic]

I am probably going to begin crunching again next month when it gets cold.

I also am considering swapping my 5950x for a 5800x3d.

 

[hat]

I am probably going to begin crunching again next month when it gets cold.

I also am considering swapping my 5950x for a 5800x3d.

Is the 5800x3d any better for WCG specifically? I would think the 5950x would be the better choice.

 

[Nordic]

Is the 5800x3d any better for WCG specifically? I would think the 5950x would be the better choice.

The 5800x3d would be a downgrade for crunching. It would be for gaming. It is hard to ignore a 50% fps boost in certain cpu bound games.

 

[phill]

I've just noticed today that my TR + 3090 seems to be only doing one GPU unit at a time... Checking the APP_CONFIG.XML file and it all seems to be good to go but WCG seems to be having none of it.. That said, on my other machines, it seems to be working fine... Go figure I know it was working before today, so I'm curious as to what has gone on with it.. Anyone else been seeing this??

Yesterday my main machine had no work. Today I've got loads.

@phill I think that's because, as the client fails to communicate to the server, it sets a time to try again. As it continues to fail, the time increases. I think the way to go is to increase "Store up to an additional x days of work" option to a reasonable amount, so you still have work even if the servers are on the fritz. Personally, I use 3 days for now.

I'd like to consider doing the longer time for work but sadly sometimes when the sun doesn't come out, I don't have the machines on.. Some of these beasts can take up to 800w for each rig so I'm trying to keep the power consumption down as much as I can.. I tend to give about a days work spare but sometimes the PCs just munch their way through it all without any problems at all!!

That said, comms today has been great, files have been downloading/uploading without a hitch Long may it continue!!

 

[mstenholm]

I've just noticed today that my TR + 3090 seems to be only doing one GPU unit at a time... Checking the APP_CONFIG.XML file and it all seems to be good to go but WCG seems to be having none of it.. That said, on my other machines, it seems to be working fine... Go figure I know it was working before today, so I'm curious as to what has gone on with it.. Anyone else been seeing this??

I had my 3900/2070 on for about an hour this morning. I got 8 OPNG in one go and my APP_CONFIG is set to 8 and they all ran concurrent. I did notice that my 5950X got nice points, 494.7 to be precise per OPN1 job. The work size for the jobs sent last night was 231 gflops later reduced to 168. Normally they are around 30, so the points are following the estimated work load but the run-time was the normal 1.60 hour. See Odd estimates

 

[phill]

I had my 3900/2070 on for about an hour this morning. I got 8 OPNG in one go and my APP_CONFIG is set to 8 and they all ran concurrent. I did notice that my 5950X got nice points, 494.7 to be precise per OPN1 job. The work size for the jobs sent last night was 231 gflops later reduced to 168. Normally they are around 30, so the points are following the estimated work load but the run-time was the normal 1.60 hour. See Odd estimates

I did manage to have a look about on the forum, noticing that page but it never really said much about why one machine was doing only one GPU job at a time and the rest where managing 4 to 8 or whatever they where set to!! Damn WCG scores haven't worked either so when I've gone to check on my device stats, nothing been updated since their last work units from February... Gutted

 

[debs3759]

Just received an email saying WCG is officially up and running.

 

[mstenholm]

Just received an email saying WCG is officially up and running.

Got it too and saw the word functioning infrastructure in the very top of it. Hmm, right now the old problem of downloading jobs is back. Well

 

[debs3759]

So much for being back up and running. I have plenty of work, but have to keep retrying downloads, and most of the website is broken!

 

[Zachary85]

I noticed the same things. They sent out the email that said they were officially up and running, I wonder if that email caused a lot of inactive users to come back online and overload the servers. Who knows at this point in time.

 

[stinger608]

So much for being back up and running. I have plenty of work, but have to keep retrying downloads, and most of the website is broken!

Yeah, I've seen that problem for the last couple of weeks. I just don't pay attention to it and they end up downloading okay.

 

[windwhirl]

Can't complain since I've a lot of work

 

[80-watt Hamster]

Can't complain since I've a lot of work
View attachment 264070

So you're hoarding it all! *shakes fist dramatically*

My issue actually seems to be comms ATM.

 

[windwhirl]

So you're hoarding it all! *shakes fist dramatically*

My issue actually seems to be comms ATM.

An observation to be made, though, these Mapping Cancer Markers WUs are pretty simple ones. Take like an hour and half to be done each, and that's for my Ryzen 3600 system capped to 45 W (because Eco mode). So, there might be some truth to that joke you made there.

 

[Zachary85]

The website and forums are back up, but boy are they slow.
Still having download issues the past 2 days, but I have enough work to keep my crunchers busy.

 

[mstenholm]

Testing, testing

 

[phill]

Had nothing but problems with downloads and such last few days, crunchers going dry because of lack of work downloading... Most frustrating but have been trying to leave things alone as @hat suggested, not really making any difference sadly

Apologies about the lack of updates of late, been getting over the flu and busy with the utmost boring job of scanning documents for home... Thank god that's done!! How's everyone been??

 

[HammerON]

I haven't had any work for several days now. Just sitting idle
Downloading issues continue.

 

[windwhirl]

Apologies about the lack of updates of late, been getting over the flu and busy with the utmost boring job of scanning documents for home... Thank god that's done!!

No worries, phill. Hope that flu leaves already (if it didn't yet).

How's everyone been??

Fine, nothing much to complain really.

About WCG, it seems I'm finishing soon-ish with my queue and will be out of work units too

 

[Nordic]

I turned on boinc for the winter this morning. I downloaded WCG tasks just fine. I have WCG at 65% resource share, Rosetta at 30% resource share, and Milkyway@home at 5% resource share. This resource allocation means I will never run out of work but WCG will be prioritized if work is available.

I still can't make up my mind about a 5800x3d. It trades 50% less performance in boinc for 50% more performance in certain bottlenecked games.

 

[windwhirl]

Back to doing nothing I suppose

 

[Zachary85]

I have stopped trying to download any WCG work for now. It's just not worth all the effort it takes to try and get work.

If you scroll down to the bottom of any page on the WCG website it shows a blue box with "Krembil ..Relentless" I'm starting to think they should change that to read "Krembil ... CLUELESS !!!!"