SETI.Germany's 14th Annual Pentathlon (2023)

FLECOM said:
added a couple GPUs, lets see if they can do anything besides generate heat hehe
Click to expand...

Thanks man! Need all the help we can get. Rechenkraft.net team just dropped a big 20M points in the last hour. LTT team just stepped up their game a few hours ago. Hourly stats.

For PrimeGrid AP27 (City run), each task requires a quorum of two for validation and then points awarded if there are no errors/invalids. Watch your pending credit. Mine is hovering around 3.6M.

To increase the validation success rate, you will need to fill up your work cache as soon as you can for it to last till the duration of the race which ends in about 3 days.





1683674884846.png
 
pututu said:
To increase the validation success rate, you will need to fill up your work cache as soon as you can for it to last till the duration of the race which ends in about 3 days.
Click to expand...

ok how do I do that?
 
is it normal for yoyo@home to just eat a ton of ram and not cpu? it was using 100% cpu for a while but now it's not

Screenshot_2023-05-09_20-08-03.png
Screenshot_2023-05-09_20-04-37.png
 

Attachments

  • Screenshot_2023-05-09_20-04-37.png
    Screenshot_2023-05-09_20-04-37.png
    249.6 KB · Views: 0
  • Like
Reactions: EXT64
like this
Yes, many of the ECM work units can periodically use a lot of ram. The _ru in particular periodically use 4GB+ each!
 
FLECOM said:
ok how do I do that?
Click to expand...
Open up your Computing preferences inside the BOINC client

1683678148172.png

Go down to Other and change the Sore at least X days of work to however many days of work you want.
1683678182930.png
 
If anyone is bunkering for SRbase / Javelin you should probably drop that now. We only need enough points to tie our first throw and it will be good to find out if we need to continue to crunch through this throw or not. Any tasks saved past this throw are likely not going to make it to throw 5 (just a guess). would suck to have completed work that is past deadline and doesn't get counted. We will see how things stand after this next update and if we need anymore points for Jav.
 
FLECOM said:
added 8x RX480s dunno if these numbers are reasonable?

View attachment 569332
Click to expand...
My RX570 will crunch through these AP units on PrimeGrid in about 45 minutes each. You might need to wait until it's finished or almost finished crunching the work unit to see how long it will actually take. For whatever reason the client or PrimeGrid thinks it would take over 2 days to crunch one of those work units when it's actually 45 minutes. Even after running work units the whole time of the race it still thinks its going to take more than a day so I can't actually bunker as I'm not going to make a new instance to run for 6-10 work units. This may be the "issue" you're seeing with regards to estimated time for completion.

It did something similar with my 6750xt but after running that for a few days it finally has the correct estimate of 12ish minutes per work unit so I can actually bunker that.

And yes, the YoYo@home work units are somewhat screwy. I had several work units earlier sitting there "waiting for memory" leaving cores unused. Looks like 16 gig of RAM is nowhere near enough for this project. And from the look of things it was the _ru work units which were sitting around waiting for RAM. They also tend to need a much longer time for crunching. Either that or the massive RAM requirements means they're sitting around "active" but not actually crunching much which leads to a much longer run time.

motqalden said:
If anyone is bunkering for SRbase / Javelin you should probably drop that now. We only need enough points to tie our first throw and it will be good to find out if we need to continue to crunch through this throw or not. Any tasks saved past this throw are likely not going to make it to throw 5 (just a guess). would suck to have completed work that is past deadline and doesn't get counted. We will see how things stand after this next update and if we need anymore points for Jav.
Click to expand...
I had a small bunker which I just released.
 
There are tricks to building a bunker when expected completion times are an issue. Slack or pm me ;)
 
SmokeRngs said:
Congrats. I don't know how rare these are but I've "found" 2 20 length and 2 21 length so far.
Click to expand...
Low ones aren't rare or at least not rare enough to even make it to the list. Basically you want either length 26 or 27 to be really meaningful.
 
SmokeRngs said:
My RX570 will crunch through these AP units on PrimeGrid in about 45 minutes each. You might need to wait until it's finished or almost finished crunching the work unit to see how long it will actually take. For whatever reason the client or PrimeGrid thinks it would take over 2 days to crunch one of those work units when it's actually 45 minutes.
Click to expand...

hrmm ok, I'll keep an eye on it, I am wondering if the mining risers are killing their performance
 
AP1.PNG


Think I started a bit late. For some reason only 1 compute section is getting used on my GPU, Probably driver related. Going to 2 threads usually leaves me with 1 thats pretty slow.
 
FLECOM said:
hrmm ok, I'll keep an eye on it, I am wondering if the mining risers are killing their performance
Click to expand...

I dont think so. I have two 3070 on x1 risers and they are just as fast as my x8 and x16 ones. Some boinc projects have an issue on risers but not this one.
 
Jeeper's Day 6 BOINC Pentathlon commentary

Notable commentaries:
  • Obstacle (yoyo) - The battle for #3 also remains exciting, with three teams involved, fittingly. [H] (#3) currently hold the best cards, followed by SG (#4) and LTT (#5).
  • City Run (AP27) - The Rechenkrafties are doing very well and just overran [H] (#7)
Can we possibly be on the podium for yoyo?
Can we spring back in gpu arena?


Day 6 – 10 May 2023 – All Good Things Come in Threes?​


Today the teams are competing on two tracks. In addition, there is the third javelin throw. Here the teams have to put their cards on the table today at the latest. After the third throw we will know who is capable of what. I will evaluate this tomorrow.

And so we can focus on the track disciplines again.

Obstacle Run (yoyo@home: only subprojects ECM, ECM P2) P3D managed to take the lead and outperform TAAT (#2) at least for now. However, the bonus days could still shake up the picture considerably. The battle for #3 also remains exciting, with three teams involved, fittingly. [H] (#3) currently hold the best cards, followed by SG (#4) and LTT (#5).

icon-obstaclerun-small.png


GPUUG and CNT are fighting for #6 at the moment. They might still join the battle for bronze as well, but their pace is not sufficient right now. SUSA's (#8) gap to the duo is already substantial, but we know from the past that the Americans always like to gamble. Whether they are able to do so this year is another matter.

In #9 and quite close behind are RKN, followed by AF, TC, and MK. The Chinese currently hold the best cards and are closing in on AF. Behind them is a quartet. AMD Users (#13), Ukraine (#14), B@P (#15), and BOINC.Italy (#16) form an illustrious travel party and we cannot yet estimate who will be in front in the end.

OCN continue to wander in no man's land and are calmly securing #17. Behind them, EVGA (#18) have just overtaken UBT (#19), and Chinese Dream (#20) might become a threat for the British as well.

LITOMYSL (#21) allowed for a small gap to open. TSBT (#22) and behind them BC and TFFE could well close the gap in front of them if they step on the gas.

Russia and B^S are furiously fighting for #25 and could witness another duel if they looked around. BOINCstats have passed PiGyS and reached #27. 2ch (#29) and Crystal Dream (#30) are not too far behind either, and with luck and skill, could also still make up places.

City Run (PrimeGrid: only subproject AP 27)

icon-cityrun-small.png
The City Run is halfway through for the teams. Unfortunately, it is not quite as exciting at the top as it is in the Obstacle Run or the Javelin Throw. TAAT have invested an insane amount of energy here and continue to put the pedal to the metal.

P3D continue to hold #2, but GPUUG (#3) are closing in. A bit behind them LTT (#4) have passed SG (#5). For both of them winning a medal is going to be tough, and they even have to watch out for RKN (#6). The Rechenkrafties are doing very well and just overran [H] (#7).

CNT (#8) are also quite fast at the moment and have the opportunity to catch up with the duo in front of them. The Czechs also have the advantage that there is currently no threat from behind. TC (#9) are kept at a distance, and AF (#10), who are fighting with AMD Users (#11), is even further behind.

Meanwhile, OCN (#12) have an eye on the AMD guys and is slowly stalking them. But they should watch out not to be surprised by SUSA (#13) while stalking, who are approaching just as steadily.

MK (#14) are running out of coolant and so they lose energy and contact. However, there is still time to get back up to speed, as Crystal Dream (#15) are a bit away. The dreamers also have the luxury of having opened a gap, so LITOMYSL (#16) and EVGA (#17) are already quite far behind.

EVGA's strong intermediate spurt has cost TSBT (#18) one place, and now the Scots have to be careful not to be chased away by TFFE (#19) as well. They not only managed to pass 2ch (#20), but also increased their lead over BOINC.Italy (#21), Ukraine (#22), and Russia (#23) in the process.

In contrast, B@P (#24) had to let go. It is a small consolation to have left Chinese Dream (#25) further behind. BC (#26), B^S (#27), UBT (#28), BOINCstats (#29), and PiGyS (#30) follow.

  • Can GPUUG silver their City Run?
  • Can P3D hold on to the lead in the Obstacle Run?
  • Can SG win the Javelin Throw?

You see it remains exciting at the BOINC Pentathlon.

Yours, Jeeper

[Based on stats at 11:00 UTC]
 
Please do not drop anymore Javelin / SRbase tasks today or tomorrow. We should be doing throw 5 and that means either bunkering for that (risky) or putting your CPU's on Yoyo (bunker for possible bonus day if you want or run without bunker is fine). Yoyo is the safe bet so so you are not sure then do that.
 
I don't know whats going on with the yoyo projects but the machine was sitting at like 3% cpu usage and none of the WUs had progressed (they were all stuck at like 80% for 8 hours) I aborted them and now it started new ones and is using 100% cpu again
 
Gilthanis said:
Open up your Computing preferences inside the BOINC client


Go down to Other and change the Sore at least X days of work to however many days of work you want.
Click to expand...

ok done, did it for 5 days that should be good?
 
ok i apologize for all the noob questions but is it normal for the GPU tasks to not be using 100% of the GPU? all of mine seem to use 70% or so of compute 1 and nothing else
 

Attachments

  • primegrid2.png
    primegrid2.png
    314.6 KB · Views: 1
FLECOM said:
ok i apologize for all the noob questions but is it normal for the GPU tasks to not be using 100% of the GPU? all of mine seem to use 70% or so of compute 1 and nothing else
Click to expand...
Depending on the GPU it is very possible. Some people run more than one instance on their GPU to squeeze a bit more production out but is only recommended on higher end GPU's.
If you choose to do so, you can create an app_config.xml file using Notepad and inside put

<app_config>
<app>
<max_concurrent>2</max_concurrent>
<name>ap26</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.1</cpu_usage>
</gpu_versions>
</app>
</app_config>

You would then put this file into your BOINC data directory. In Windows it is typically c:\\ProgramData\BOINC\projects\www.primegrid.com\
You then will want to restart your BOINC client.
 
Numberfields has been announced for the sprint. You can start bunkering that if you like. The good news is that you can use any Intel iGPU's (newer than Sandy Bridge IIRC) on Numberfields. So, if you want to keep your CPU on YoYo you can dedicate your iGPU to Numberfields. We don't know if the last project to be announced for the Cross County event will a GPU capable project yet but most likely will be. So, it will be up to you if you want to bunker AMD/nVidia work for Numberfields as well or not after PG ends.
 
For the moment I'll probably finish off what I have for PrimeGrid and see what things are like at that point with regards to Numberfields. It would be good to get a jump on Numberfields but if it also means losing another spot in PrimeGrid I'm not sure it's worth it. At the same time, there is the requirement of PrimeGrid needing everything checked before it counts so the last day or so of crunching may not be worth trying if my bunker runs out before then.

Decisions, decisions.
 
Looks like Cross Country is Einstein@Home. I'm pretty sure Einstein was a complete shitshow last year so let's hope their servers can handle the load this time.
 
Einstein@home is announced for the Cross County event. It supports CPU and GPU work units which also include Intel iGPU. Thanks SmokeRngs for posting above. Yeah, it struggled a ton last time. It would be best for anyone running the project to load up a large enough cache of work to last the remainder of the event so that it increases their chances or validating/crediting.
 
Gilthanis said:
Einstein@home is announced for the Cross County event. It supports CPU and GPU work units which also include Intel iGPU. Thanks SmokeRngs for posting above. Yeah, it struggled a ton last time. It would be best for anyone running the project to load up a large enough cache of work to last the remainder of the event so that it increases their chances or validating/crediting.
Click to expand...
So 2 days from now.
Is this the [H] Team I presume Hard OCP
https://einsteinathome.org/community/teams/179351

1683907114912.png
 
Last edited:
Jeeper's Day 7 BOINC Pentathlon commentary (one more grueling week to the finishing line)

Day 7 – 11 May 2023 – Days of Decision​


In the next two days the City Run and the Javelin Throw will be decided. Since today we also know when all the other disciplines will run. A challenge for the teams who now have to manage their forces. Today I will look at the situation in the Javelin Throw and in the overall standings, which I have neglected severely so far. Although there is still a lot of variability in it, it is worth taking a look. And after that, you will get today's winners and losers in the Obstacle Run and the City Run.

Javelin Throw (SRBase: all subprojects except TF)

icon-javelinthrow-small.png
Three throws are done and the fourth is today. A rare sight presents itself right now. Ukraine are leading ahead of RKN and LTT. But I probably do not have to tell you that this will not stay this way. Ukraine have a good chance to reach the top 6 and RKN still have a good chance to win a medal. However, for this the final two throws have to be spot on. Still in #4, MK have shown their potential, but it is questionable whether it will be enough for the top 10.

The hottest bet for gold is TAAT (#5). Yesterday's throw was absolutely top notch and today's follows on from that. They still have to confirm it with the final throw, but it would be a real surprise if they failed to do so. BOINC.Italy (#6) and AF (#7) will likely lose a lot of ground. The Italians made three solid throws, whereas the Francophones saved their energy yesterday. Nevertheless, they will probably finish behind the cooling experts, unless they get too hot.

CNT (#8) made two strong and stable throws and have to confirm those once again. If they do so, they will probably end up just below the podium. It will be interesting to see where P3D (#9) are going to end up. Silver, bronze, or iron, that is the question here. The first throw was not good enough to bag precious metal. RKN and SG (#11) are their competitors.

SG became a bit nervous yesterday and once again put the medal at risk. As in previous years they might be forced to make a fourth strong throw. However, this is a challenge because of the narrow time window. Currently it looks like the Green Planet is skipping today, but SG are doing so as well, which might cost them the medal.

AMD Users (#10), TSBT (#12), and BC (#13) can already get in the elevator down. For the Scots, the javelin was already their weak point last year and this is unlikely to change this year.

[H] (#14) will probably go for the final throw and end up behind CNT. TC (#15) will not be involved in the decision for the top places either, their throws are just too short. TFFE (#16), Crystal Dream (#18), Russia (#19), and OCN (#20) will decide the ranking among themselves. However, the quartet will not be able to keep EVGA (#21) behind. They are saving their energy today, but this will probably change on the final day.

B@P (#17) are puzzling us. Their first throw was excellent, but from then on, there was nothing coming anymore. Will there come anything later today? BOINCstats (#22) have probably finished their job. They will almost certainly have to let LITOMYSL (#23) and SUSA (#24) pass. The Americans are banking on the final days, but they are far from their form in the past.

Chinese Dream (#25) will have to let B^S (#26) pass, and maybe GPUUG (#28) as well. PiGyS (#27) are a few centimeters short. While 2ch will probably not touch the javelin at all, UBT are also puzzling us. Only one good throw so far and nothing else countable.

Overall standings

ringe.png
So far, I have somewhat neglected the overall standings because the javelin has introduced some strong distortions. This has not changed yet, but it is time to have a closer look.

TAAT are at the top and their likely result in the Javelin Throw is not even fully included yet. LTT are on the silver rank, but will probably have to let it go. P3D (#3) and SG (#5) will surely pass them. RKN (#4) still benefit from their strong result in the Javelin Throw and will surely come under pressure from [H] (#7) and CNT (#8).

Ukraine are still holding on at #6, and GPUUG (#9) are also doing well. For both of them, it will be interesting how they will do in the next disciplines. AF (#10) are again very consistent. You can just count on strikes in France and by AF in the Pentathlon.

MK in #11 still benefit from the javelin, but they have some cushion on TC (#12) and AMD Users (#13). SUSA is currently only in #14, in part because the completely fail in the Javelin Throw. But this also means that they can improve in the next disciplines. However, reaching the top 10 will be a challenge.

BOINC.Italy in #15 might still be able to follow. OCN (#16), B@P (#17), EVGA (#18), and TSBT (#19) are basically on the same level. They all have potential waiting to be exploited.

The next close battle is just behind them. Crystal Dream (#20), TFFE (#21), and BC (#22) are fiercely battling each other. And LITOMYSL (#23) and Russia (#24) are also trying everything not to lose contact for good. Except for LITOMYSL, all five teams are losing a lot of ground in one project.

Chinese Dream (#26) are doing very well in the obstacle run, as are UBT (#27). But their advantage is fading slowly as BOINCstats (#28) and 2ch (#29) are catching up. The PiGyS round off the field. Obstacle Run (yoyo@home: only subprojects ECM, ECM P2)

icon-obstaclerun-small.png

#teamΔ
3SETI.Germany+1
10Team China+1
13Ukraine+1
19Chinese dream+1
#teamΔ
4[H]ard|OCP-1
11L'Alliance Francophone-1
14AMD Users-1
20UK BOINC team-1


icon-cityrun-small.png City Run (PrimeGrid: only subproject AP 27)

#teamΔ
16Ukraine+6
19Russia+4
11SETI.USA+2
2GPU Users Group+1
3LinusTechTips_Team+1
6[H]ard|OCP+1
#teamΔ
7Rechenkraft.net-1
20The Final Front Ear-1
4Planet 3DNow!-2
13AMD Users-2
18LITOMYSL-2
22team 2ch-2
23The Scottish Boinc Team-5


Lots going on, lots of thoughts, lots more questions:

  • Will P3D stop TAAT's winning streak?
  • Will GPUUG win silver in the City Run?
  • Which one of the three German teams will come away empty-handed from the Javelin Throw?

You see it remains exciting at the BOINC Pentathlon.

Yours, Jeeper

[Based on stats at 11:00 UTC]
 
rgMekanic said:
Wow, no idea that was a thing. Thank you for letting me know before I dropped the bunker I'm working on of those lol
Click to expand...
Yeah...it was because of the GDPR that has no bearing on most things BOINC. Especially for those of us outside the USA. So, there are a hand full of projects that require it to have your stats exported.
 
You must log in or register to reply here.
Back
Top