9th Annual BOINC Pentathlon 2018 Strategy & Preparation

I'm going to have around 4 million points on PrimeGrid pending validation when the contest starts.

Yesterday I only got 400,000 points validated even though I had two 1080TIs on it. The slow build of validated points over this Cross Country race is going to be a nail biter!
 
There are currently around 600k PPS Sieve tasks in progress, and BOINC users are completing about 92k per day. Graphs here: https://www.primegrid.com/munin/boinc-day.html

Edit: I'm going to keep my BOINC setup pulling down a bunch of days of work, in the hopes that that will make it so all the tasks I'm processing are already assigned to someone, who also most likely returned it already.

Edit: What I think those stats mean is there is going to be a 6 day ramp up in points because the completed to in progress is 1:6 We shall see!
 
Last edited:
My custom app_config stopped working on my rig with 1080ti for some reason last night. It started doing only 1 work unit at a time on prime.. still working for both of my 1070's so i am not sure whats up with that. I re-booted, double checked the file, tried to tell it to load config, etc. performance sucks with just 1 task so i wish i knew what was up with it... wonder if it needs to start uploading again to get working correctly.
 
This is gonna be close.

I am getting the HVAC ducting / inline fans today. The tent will be here "by 9m" by today.

I am heading out in roughly an hour or so to get 2x4s for rails and a 2x2x4 project pane to serve as a gasket for the grow tent up top.

I'll have pics when this is all said and done with.


At worst, I may be a day late...but no dollar short. :D
 
Gilthanis give me a PrimeGrid config for just the sieve tasks. (CPU, GPU) and the Cullen\Woodall tasks. (CPU)
 
app_config>
<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>N</gpu_usage>
<cpu_usage>N</cpu_usage>
</gpu_versions>
</app>
<app>
<name>llrGCW</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrGCW</app_name>
<cmdline>-t N</cmdline>
<avg_ncpus>N</avg_ncpus>
</app_version>
<app>
<name>pps_sr2sieve</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>pps_sr2sieve</app_name>
<cmdline>-t N</cmdline>
<avg_ncpus>N</avg_ncpus>
</app_version>
</app_config>

Just change N to what you want accordingly. If you aren't sure... answer my question in post 328
 
please leave the CPUs on Rosetta
Screen Shot 2018-05-13 at 6.03.46.png


still some fight on rank #3; SG coming closer update by update and passed TC
 
I can't run all 64 cores on Rosetta on one of my 4Ps because I don't have enough disk space to keep the tasks stored while they crunch. Not much I can do about that right now.
 
Skillz said:
I can't run all 64 cores on Rosetta on one of my 4Ps because I don't have enough disk space to keep the tasks stored while they crunch. Not much I can do about that right now.
Click to expand...
As many core as disk allows ? And the rest on the marathon ... similar I have on limited systems ...
 
Well I have a mix of systems.

I've got a;
48 core system, 2 tasks per core.
64 core system, 2 tasks per core.
20 thread system, 18 usable on the CPU. 2 GPUs, 2 tasks each on GPUs
4 core system, 3 usable cores, 2 tasks per core, 1 GPU, 2 tasks each on GPU
 
Well there goes my bunker. hopefully that will get my 1080ti back up and running 2 tasks. still uploading but looks like a bunch of it has already validated.
State: All 2387 | In progress 622 | Pending 765 | Inconclusive | Valid 749 | Aborted 251 | Invalid | Error
 
Last edited:
EXT64, create an app_config.xml file and put this in it

<app_config>
<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

If on Windows machine put it in C:\ProgramData\BOINC\projects\www.primegrid.com\
If Linux, that depends on where your Projects folder is located.
 
Linux is most likely in /home/[user]/projects in my case it's /home/boinc/projects
 
Well i had to dump some work units but i finally got my 1080ti back on track. it was limping along for a while and hugely underperforming. GPU usage was really low but seems to be sticking around 99% now
where it was practically 50% before. makes sense though since one task 0.5 GPU..... couldn't fix it till i dropped bunker but im sure i lost about a million points for no good reason.


Capture.PNG
 
TeamChina is making a push in NumberFields. It will be great if we can win at least a medal. BTW, did we ever win a medal before in Pentathlon?

I'm expecting a big bunker drop in Rosetta in the next 22 hours by several teams. It is a good project for bunkering because of quorum 1.

We will see how we do in Asteroid in the next few days.

I think we should defend our 3rd spot in NumberFields. Would like to know what other thinks. The other option is to do the best in every event and get the team overall score up if this is preferred by most [H] members.
 
TeamChina is pushing Rosetta as well. My guess is that they had multi bunkering going with multiple projects in preparation. Let's see if they can keep up this pace on all 3 projects...but Rosetta is the one with the least amount of time left and we need to guarantee coming out ahead of SETI.Germany
 
Gilthanis can you give me an app_config for rosetta so i can set the number of threads. I think finally found out what my problem is for some reason its giving too much priority to rosetta i think and this causes me to drop a thread on prime. If i pause my rosetta i can run 2 threads of prime but even if i lower my CPU usage it seems to cause me to drop a prime thread. even when my overall CPU usage is like 75% for the system and only running 11 of 16 cpu thread for rosetta. I think if i manually configure rosetta to run to 12-14 threads and set Max CPU usage to 100% it will force it to leave the extra threads alone for my prime... hoping...
 
<app_config>
<project_max_concurrent>N</project_max_concurrent>
</app_config>

Replace N with how many cores/threads you want it to run. Then name the file app_config.xml and put it in the rosetta projects directory.
 
funny thing is on my other system its not doing this. I set overall CPU usage in boinc to 100% and it still only runs 4 of 8 threads and my GPU config seems to fully reserve them and not let any more rosetta run. This is why i think i need to manually set the number of threads on rosetta
 
Skillz said:
<app_config>
<project_max_concurrent>N</project_max_concurrent>
</app_config>

Replace N with how many cores/threads you want it to run. Then name the file app_config.xml and put it in the rosetta projects directory.
Click to expand...


This is the full config? doesn't need app names or anything?
 
Yea, full config. Doesn't require individual app names it'll do that for all of them.
 
Thanks! hope this works. If not i might try putting my cpu's on another project as my gpu is barely coasting with 1 thread..

Edit: Yes this did the trick. Simply lowering CPU usage overall was cutting back rosetta tasks but the overall amount it was allotting must have been also reduced for prime, and because it was favoring rosetta i was dropping a prime thread. funny part was it was fine for 1.5 days and just all of a sudden started doing this.
 
Last edited:
@motalden, no app_config for setting threads. Just change it in the BOINC Manager settings to not use 100% of CPU's. For example, if you needed to reserve 1 cpu for GPU work on a quad core, you would tell the client it could only use 75%
 
Yes you could do thatHowever, what is happening is that BOINC Manager thinks you are over committed and is giving CPU project priority over GPU by design. So, you will need to do that on every CPU project or just change the BM preferences to leave a few core/threads free

Well... Skillz has had a different experience with it than me. Try his way and if that doesn't suffice, try mine.
 
Gilthanis said:
@motalden, no app_config for setting threads. Just change it in the BOINC Manager settings to not use 100% of CPU's. For example, if you needed to reserve 1 cpu for GPU work on a quad core, you would tell the client it could only use 75%
Click to expand...

That does not work. You tell it to only use 75% of the CPUs on a quad core than only 3 cores will be used and the other core will NOT be used by the GPU. So you'll at best have 2 CPU projects running and 1 GPU project running. The other core wont get used.
 
Gilthanis I tried what you suggested and while it was dropping CPU tasks and freeing up overall cpu usage for the system, i was still having the problem(even at 50%). manually limiting CPU threads fixed it though. As mentioned this isn't happening on my other machine which has 2x 1070 , even when boinc CPU usage is pegged at 100%

As for Strategy: my 2 cents is it seems like it might be really hard to beat china, but we might stand a chance and dropping all possibility of overall medal just to get an individual medal would be a shame. I think 4 place with no overall medals if it came down to it wouldn't be so bad, but i think if ATP comes back online we stand a fighting chance in cross country, even with the early deficit so large. Looking at the stats i figure a bunch of people haven't dropped bunker yet for Prime?
 
I think the only team that hasn't dropped a big bunker is TAAT and P3DN.

atp1916 was posting a few minutes ago. Not sure if he's getting things fired up or not. Lets hope he is.
 
motqalden, it may have been changes in the client over time that makes it behave different. I'm thinking your client also may not have "learned" the runtimes well enough yet to gauge the amount of work it can handle. Are any of them saying High Priority in the BM? I don't recall if they took that lingo out or not as it confused some people.
 
I have seen that before with other projects but not recently. I thought that was when a task had been bounced back to the server by someone (failed / timeout etc) and had been floating around for a while so giving it high priority meant it would run first even if you downloaded it later, so that it could get validated more quickly and keep the project and or your wingman from waiting longer. (that was just my guess though) regardless the extra config seems to have done the trick and it won't be hard to add more for other projects if i need to since its only 1 system doing it. I appreciate the help and input as it makes a huge difference. I am now doing 2 tasks in 5 mins 40 seconds VS 2 separate tasks totaling 8 mins 50 seconds (while running 14 of 16 cpu threads)
 
I have 330 asteroids *Edit these are gpu tasks bunkered which is about 158k points, but other than that i have no more bunkers and my cpu's are all currently rosetta. while not a great amount of credit for 2 days of gpu work its better than nothing and much better than all of my cpu's would have done in the same timeframe i think (and the gpu's were doing nothing). I figure i should get at least 4 million points per day on prime depending on how validation goes .
 
Last edited:
CNT just dropped two big bunkers few hours ago in Rosetta. We got bumped down to 4th.
 
You must log in or register to reply here.
Back
Top