BOINC Pentathlon 2020

[marknd59]

2 minutes ago, Gorgon said:

So if the NumberFields server is unreachable at the start of the Javelin throw and you release a bunker at the start it might not count.

I've set mine to release at 02:00 UTC.

[ConceptOW]

Getting there...

[Gorgon]

15 minutes ago, marknd59 said:

I've set mine to release at 02:00 UTC.

Are you using the time-of-day feature in BOINC Manager or BOINCview?

[marknd59]

2 minutes ago, Gorgon said:

Are you using the time-of-day feature in BOINC Manager or BOINCview?

The feature in BOINC Manager, Daily schedules, Network. I've had network activity supended I set the release time a couple of hours ago and then turned network activity back to based on preferences. They should get released 02:00UTC, I have to do it that way because I'm in the UK so it's the middle of the night for me.

[Gorgon]

4 minutes ago, marknd59 said:

The feature in BOINC Manager, Daily schedules, Network. I've had network activity supended I set the release time a couple of hours ago and then turned network activity back to based on preferences. They should get released 02:00UTC, I have to do it that way because I'm in the UK so it's the middle of the night for me.

Yeah, that makes sense. I have 15 bunkers across 4 systems to release so I suspend the BOINC instance then restart it when I want to release. I shut them down rather than leaving them running and suspended as a) I didn't know how much resources an idle boinc-client might use and, b) I'm lazy and re-used the same port (31422 - which has been playing hell with my dsylexia) for every instance

[cbigfoot]

ok so i found 32 GB of DDR-4 2400 bender is getting that so he can run 2xGPU and 1xCPU task of Amicable mumbers...

[cbigfoot]

ok what did i miss how do you get boinc to see both vcards? it will only see my 2060 not my 2060 ko

[Gorgon]

I'm not seeing much of a surge from the WAN Show shout-out. Likely we needed some prizes to entice more casual people into running with us.

 

I'll keep my CPUs on Rosetta so we can hopefully maintain our ranking.

 

Unfortunately I can move any GPUs onto Amicable Numbers as I just don't have the RAM and already upgraded the dual GPU systems to 16GB so they could run both Rosetta and Number Fields without getting bogged down so getting them on 32GB each isn't likely.

[Gorgon]

5 minutes ago, cbigfoot said:

ok what did i miss how do you get boinc to see both vcards? it will only see my 2060 not my 2060 ko

in your BOINC Data Directory (C:\ProgramData\BOINC on windows) edit cc_config.xml: - then "Read Config Files ..." from Options menu in BOINC Manager

root@fold9:/var/lib/boinc_skel# cat cc_config.xml
<cc_config>
 <options>
  <use_all_gpus>1</use_all_gpus>
  <allow_multiple_clients>1</allow_multiple_clients>
 </options>
</cc_config>

[cbigfoot]

Just now, Gorgon said:

cc_config.xml:

root@fold9:/var/lib/boinc_skel# cat cc_config.xml
<cc_config>
 <options>
  <use_all_gpus>1</use_all_gpus>
  <allow_multiple_clients>1</allow_multiple_clients>
 </options>
</cc_config>

 

well maybee the KO isnt supported as i put that in and restarted and same only the non KO running...thats ok though for now as it got me 40 GPU units now...will keep F@H on the KO

once amicable gets fully bunkered i will reenabled Rosetta. and during city Run will keep my GPUs on Amicable and CPUS on Rosetta but till bunkering complete its full speed ahead amicable.

[Gorgon]

2 minutes ago, cbigfoot said:

well maybee the KO isnt supported as i put that in and restarted and same only the non KO running...thats ok though for now as it got me 40 GPU units now...will keep F@H on the KO

It should be - BOINC is less smart than F@H. As long as the Nvidia driver sees the card it should work

[Gorgon]

@cbigfoot did your text editor perhaps create cc_config.xml.txt?

[cbigfoot]

Just now, Gorgon said:

@cbigfoot did your text editor perhaps create cc_config.xml.txt?

nope i had it already so i could remotely monitor: Here is the contents

<cc_config>
   <options>
       <allow_remote_gui_rpc>1</allow_remote_gui_rpc>
       <use_all_gpus>1</use_all_gpus>
       <allow_multiple_clients>1</allow_multiple_clients>
   </options>
</cc_config>

[Gorgon]

53 minutes ago, cbigfoot said:

nope i had it already so i could remotely monitor: Here is the contents

<cc_config>
   <options>
       <allow_remote_gui_rpc>1</allow_remote_gui_rpc>
       <use_all_gpus>1</use_all_gpus>
       <allow_multiple_clients>1</allow_multiple_clients>
   </options>
</cc_config>

weird. I don't need the remote_gui option there. I just have a remote_hosts.cfg and that works for me

[ConceptOW]

Is there a noticeable difference with a higher clock speed?

[Gimli]

Who should I get in touch with if I want to donate a couple of prizes for this?

[Ithanul]

Yeah, these Amicable WUs are no joke.  8GB of RAM usage and higher temps on my 1080Ti which is averaging around 71C (note, air cooled with custom fan curve).

 

Crap, just had a driver crash on the 1080Ti, yeah, seems this project is going to be a no go for me on that card.  May have to look into if there is some app tuning for this one.

[Windows7ge]

3 minutes ago, ConceptOW said:

Is there a noticeable difference with a higher clock speed?

I can't confirm this project to project but at least comparing World Community Grid on a 5960X to a Xeon of the same architecture clock-rates do make a difference in terms of WU completion time.

[Ithanul]

18 minutes ago, Windows7ge said:

I can't confirm this project to project but at least comparing World Community Grid on a 5960X to a Xeon of the same architecture clock-rates do make a difference in terms of WU completion time.

Same for Universe.  When I had my previous 5960X running 4.6GHz under W7, it spat those out even faster than my 1900X under a Linux distro.  Could be different now as that was a year or more ago.   

 

Heads up Amicable is one that has some bunker resistance.  It only allows 20 GPU tasks per GPU per client.  So, having multiple clients is the method to get around that.

 

Also, by the looks, we should be able to claim another spot in the City Run, Universe.

 

 

I aimed my 1950X full out on Universe for the moment.  Xeon 12c/24t still on Rosetta and my 1900X is a mix of Rosetta and Universe.

[leadeater]

4 hours ago, Gorgon said:

I think I'll just start another instance of NumberFields for the start of the event and wait for the first hour to complete then if I'm seeing points being updated release starting then.

Numbers didn't have any problems on the first one but waiting is best, also if you check the Pent chat there is usually a post in there saying the first stats gather was done and it;s safe to start.

[leadeater]

34 minutes ago, Windows7ge said:

I can't confirm this project to project but at least comparing World Community Grid on a 5960X to a Xeon of the same architecture clock-rates do make a difference in terms of WU completion time.

@ConceptOW It's a trade off between number of cores (Xeon) and clock speed (HEDT), Xeon usually wins as long as it's a high core count one otherwise HEDT does better. It can also depend on memory bandwidth required, 3990X could be faster than 7742 but if memory bandwidth is required then the 7742 has nearly double.

 

Can't wait for Zen2 TR and EPYC to hit ebay at good prices 

[Gorgon]

47 minutes ago, Gimli said:

Who should I get in touch with if I want to donate a couple of prizes for this?

@Ithanul

[cbigfoot]

2 hours ago, Gorgon said:

@cbigfoot did your text editor perhaps create cc_config.xml.txt?

ok so had to reoot computer...good news Unlike F@H the 2060KO actually acts more like a 2080 than a 2060 it crushes amicable in 5-6 mins...vs 10-11

[Gorgon]

OK - Last system updated before the start - 5700 tasks completed and 1700 tasks in the queue for the Number Fields Javelin throw

[leadeater]

28 minutes ago, cbigfoot said:

ok so had to reoot computer...good news Unlike F@H the 2060KO actually acts more like a 2080 than a 2060 it crushes amicable in 5-6 mins...vs 10-11

Nice! I was hoping one of these KO situations would finally show itself.