Pentathlon 2023

[UBT - Timbo]

Hi all

I just received this invitation by email:

hi UK BOINC Team,

once again May is approaching rapidly, once again SETI.Germany invites to the BOINC Pentathlon https://www.seti-germany.de/boinc_pentathlon/start.php.

Starting on 5 May, fourteen days with five disciplines at five projects are on the schedule for the fourteenth time. There are no major rule changes this year, only a slight downward adjustment of the bonus points during the Obstacle Run. Nevertheless, we hope to be a bit more lucky with the stability of the projects and to have one or the other surprise in stock again this year.

Of course, even with the best preparation, sometimes things can go wrong, even though we do our best to make the event run as smoothly as
possible. It should also be noted that some projects see the Pentathlon as a welcome stress test and may even learn more from any problems than
from smooth sailing. The Pentathlon will not be perfect because it cannot be perfect.

But don't let that discourage you from taking up the challenge, dusting off your computers, and signing your teams up on {removed link} (possible until 2 May), because:

It will be exciting again at the BOINC Pentathlon!

Best regards
pschoefer
SETI.Germany

Historically we've not done well with this Challenge, but I've signed us up and I hope anyone interested will crunch those projects that are chosen.

regards
Tim

[stephenhunter]

I will try my best

[wbiz]

Sounds way too complicated for me, my iphone runs my life, anything more complicated than setting an alarm and I've got no chance.

[UBT - Timbo]

Sounds way too complicated for me, my iphone runs my life, anything more complicated than setting an alarm and I've got no chance.

Hiya

Yes, it does require a certain degree of dexterity to manually manage which projects should be running on which days. Plus they all start and end at 00:00 UTC, so you have to be able to switch projects with ease at the appointed time

And some of the Challenges last only 1 or 2 days, so one could argue that any credits earned would be relatively low, so it might not be worth crunching them and just concentrate on the Challenges taking place over a longer period (typically 3-5 days)?

At least we might have a chance, depending on how many UBT members are actually active on the chosen projects (as this is a Team Challenge and is not dependant on individuals signing up).

regards
Tim

[UBT - Timbo]

Hi all

The First Pentathlon project is yoyo@home (CPU subprojects ECM and ECM P2 only) and will take place as follows:

Obstacle Run - https://www.seti-germany.de/boinc_penta ... llengeid=1
Start: 05 May 2023 00:00 UTC
End: 19 May 2023 00:00 UTC

The Second Pentathlon project is PrimeGrid (subproject AP27 only) and will take place as follows:

City Run - https://www.seti-germany.de/boinc_penta ... llengeid=2
Start: 08 May 2023 00:00 UTC
End: 13 May 2023 00:00 UTC

The Third Pentathlon project is SRBase (all the CPU subprojects and *NOT* the GPU subproject TF) and will take place as follows:

Javelin Throw (Day 1) - https://www.seti-germany.de/boinc_penta ... llengeid=3
Start: 07 May 2023 00:00 UTC
End: 08 May 2023 00:00 UTC

Javelin Throw (Day 2)
Start: 08 May 2023 00:00 UTC
End: 09 May 2023 00:00 UTC

Javelin Throw (Day 3)
Start: 10 May 2023 00:00 UTC
End: 11 May 2023 00:00 UTC

Javelin Throw (Day 4)
Start: 11 May 2023 00:00 UTC
End: 12 May 2023 00:00 UTC

The FOURTH Pentathlon project is NumberFields https://www.seti-germany.de/boinc_penta ... llengeid=4

Sprint run
Start: 13 May 2023 00:00 UTC
End: 16 May 2023 00:00 UTC

The FiFTH Pentathlon project is Einstein@home https://www.seti-germany.de/boinc_penta ... llengeid=5

Crosscountry run
Start: 14 May 2023 00:00 UTC
End: 19 May 2023 00:00 UTC

UPDATE:

The overall standings for all teams taking part is here:
https://www.seti-germany.de/boinc_penta ... athlon.php

regards
Tim

[wbiz]

There goes my team lead for yoyo on FormulaBoinc Marathon

[UBT - Timbo]

There goes my team lead for yoyo on FormulaBoinc Marathon

Hiya

Well, the stats say you have 140,478 BOINC credits on yoyo so far, during this FB season.

I'm in 2nd place with 32,433 and Stephen is 3rd with 15,992

(see here: https://formula-boinc.org/equipes_v3.py ... &projet=59 )

So, you are way ahead as it stands, and earning 100,000+ BOINC credits might take me more than 2 weeks, even assuming I am crunching yoyo 24/7 for all this time !

However, i have returned 3 tasks already (in advance of the Pentathlon) and earned 559 credits already, so maybe I can overtake you?

On the flip side though, they do take CPUs some time to crunch: one task took 5,800 seconds (1.5 hours) and the other two took 13,300 seconds (3hr 40mins) each.

14 days = 1,209,600 seconds so if each task takes 13,300 seconds per CPU, then I can complete 91 tasks in that time, and if each CPU earns maybe 231 credits each = 21,021 credits. PER CPU. Just as well I have my 16-core Xeon CPU on this

So, the gauntlet has been laid down...give it your best shot !!!!!

regards
Tim

[stephenhunter]

So this clashes with the FB sprint, damn!
Though I do have a GPU now (1080 from Ebay) So hopefully the FB sprint will be a GPU project, leaving the CPU for Yoyo.
I have started with a few Yoyos as well, trying to get a few 'pending' banked before the start

[wbiz]

So, the gauntlet has been laid down...give it your best shot !!!!!

regards
Tim

Lol, I intend to try and defend my position.

I opened up yoyo on four machines last night, there were still a few other tasks I wanted it to complete first (long Sidock tasks especially), credits should start rolling in.

Yoyo is quite long on the Raspberry Pi's (up to 23 hours) and I have a big gripe that for most of yoyo tasks they checkpoint in 20% increments which means I can lose up to 8 hours work if they task switch or computer becomes busy.

It would be funny if FB sprint turns up yoyo as well but that would put you in a very awkward position.

[UBT - Timbo]

So this clashes with the FB sprint, damn!

Hi Stephen

Not quite.

I've been in touch with Patrick at SETI.Germany and we have worked together to ensure that the chosen Sprint project and the Pentathlon Disciplines will not be in contention with each other for tasks (which could mean projects run out of tasks and everyone gets shirty, or project websites simply become overwhelmed with uploads and downloads, causing it to crash !!).

Though I do have a GPU now (1080 from Ebay) So hopefully the FB sprint will be a GPU project, leaving the CPU for Yoyo.
I have started with a few Yoyos as well, trying to get a few 'pending' banked before the start

And yes, you've got the right idea about being able to run CPU tasks for one Challenge and GPU for the other

Isn't it nice when people work together to ensure things just "work out?"

regards
Tim

[UBT - Timbo]

Lol, I intend to try and defend my position.

I opened up yoyo on four machines last night, there were still a few other tasks I wanted it to complete first (long Sidock tasks especially), credits should start rolling in.

Yoyo is quite long on the Raspberry Pi's (up to 23 hours) and I have a big gripe that for most of yoyo tasks they checkpoint in 20% increments which means I can lose up to 8 hours work if they task switch or computer becomes busy.

It would be funny if FB sprint turns up yoyo as well but that would put you in a very awkward position.

Hiya

Go for it !! And lets see who takes the yoyo leadership (on FB) by the 19th May !!

The 6th Sprint (starting tomorrow) will NOT be yoyo, I can guarantee that, as Patrick (from SETI.Germany) and I have chatted about the scenario where projects run out of tasks or servers go offline due to being overwhelmed !! And neither of us want that !

regards
Tim

[stephenhunter]

Just noticed that it's just ECM and ECM P2 tasks for Yoyo. 10 hour tasks

AP27 for the PrimeGrid city run

Just tried some AP27 PrimeGid tasks and they all failed. So, just Yoyo for me

[wbiz]

Just noticed that it's just ECM and ECM P2 tasks for Yoyo. 10 hour tasks

AP27 for the PrimeGrid city run

Just tried some AP27 PrimeGid tasks and they all failed. So, just Yoyo for me

AP27 averages 24 hours on CPU and 12 minutes on GPU so I'm out on that as well.

[UBT - Timbo]

Just noticed that it's just ECM and ECM P2 tasks for Yoyo. 10 hour tasks

Hiya

On my Xeon CPU, the tasks have taken from just under 2 hours to over 4 hours (and in about 5 cases a lot longer - see below). But they do vary quite a bit, in terms of time to complete, so I guess you just have to be lucky and hope you get some short ones.

My shortest tasks took: 5,802.38 secs (1.5 hrs) - earning just 95.11 credits
And the longest one took: 41,805.20 secs (11.6 hours) - and earned 679.35 credits

At least the credits given are: a) proportional to the time taken, and the granted credit, tends to be much higher than the claimed credit.

AP27 for the PrimeGrid city run

Just tried some AP27 PrimeGid tasks and they all failed. So, just Yoyo for me

Yup, GPUs are normally the best way to crunch PrimeGrid tasks !

regards
Tim

[ChelseaOilman]

Just tried some AP27 PrimeGid tasks and they all failed. So, just Yoyo for me

Is this on your GTX 1080? Unless things have changed since I ran AP27 GPU tasks, they use OpenCL, not CUDA. Make sure your using the GPU driver from the Nvidia site and use something like GPU-Z to check if OpenCL got installed with the driver. If your running the card on Linux you may need to install OpenCL separately.

[stephenhunter]

Checking the Nvidia X Server settings, it looks like OpenCL is installed, but clinfo shows nothing!
Will look into it.

On a related note, I got a GPUGrid task completed, ATMBeta. But the credit does not show on FB or BG. Any ideas?

[wbiz]

On a related note, I got a GPUGrid task completed, ATMBeta. But the credit does not show on FB or BG. Any ideas?

New projects are slow to appear on FB, I finished a Ramanujam two days ago to put me on the scoreboard, it hasn't arrived yet, they usually arrive by the third day. Happens on the stats sites sometimes as well, its not just a FB thing.

[stephenhunter]

Ok, I will keep my eyes open

[UBT - Timbo]

Just noticed that it's just ECM and ECM P2 tasks for Yoyo. 10 hour tasks

Hiya

On my Xeon CPU, the tasks have taken from just under 2 hours to over 4 hours (and in about 5 cases a lot longer - see below). But they do vary quite a bit, in terms of time to complete, so I guess you just have to be lucky and hope you get some short ones.

My shortest tasks took: 5,802.38 secs (1.5 hrs) - earning just 95.11 credits
And the longest one took: 41,805.20 secs (11.6 hours) - and earned 679.35 credits

OMG - I've now had 3 tasks that took over 108,000 seconds - that's just over 30 hours.

And what's worse is that my PC slowed down to a crawl, as some of these ECM tasks were taking between 1Gb and 1.8GB of RAM, EACH.

This is well documented here:

https://www.rechenkraft.net/forum/viewt ... 57&t=17709

So, I've had to limit the number of active CPUs on yoyo, as this host only has 8Gb of RAM.

regards
Tim

[wbiz]

Yep "yoyo" is a very apt name for time taken, credits awarded and memory used.

[stephenhunter]

Well I'm having to sideline Yoyo, it persistently causes the PC to freeze

Tried dropping the cpu usage%, tried running just a few tasks, still freezes. Same thing happened the last time I ran Yoyo, it just doesn't like me i guess

Still got Einstein going though, and SR base will be ok

[UBT - Timbo]

Well I'm having to sideline Yoyo, it persistently causes the PC to freeze

Hiya

Yup, same happened with me, and one is forced to reboot - but at least yoyo does checkpoint each task, so only a few seconds might be lost (depending on your settings).

The issue (in my case) is simply down to using older (but still perfectly usable/acceptable) hardware, and more recent applications are being created that require more CPU and memory resources.

This has been mentioned on the yoyo messageboard, but no apps have been created that are more suitable for legacy type hardware.

Tried dropping the cpu usage%, tried running just a few tasks, still freezes. Same thing happened the last time I ran Yoyo, it just doesn't like me i guess

I've been running yoyo on my work host (16 vcore Xeon) and had to throttle back to only using 50% of the cores, just so it didn't lock up. But with older SIMMS being cheaper these days, I just need to buy some more and then it can run these memory intensive tasks a bit better. But that's all the benefit I'd get from the extra cost, as this host is used at work, for simple printing and text editing processes, so 16Gb (or more) is simply wasted for these mundane work functions.

Still got Einstein going though, and SR base will be ok

Nice

Note that the SRBase "Run" excludes GPU (ie TF) tasks. So only CPU tasks will count for this "Run", which will of course conflict with yoyo CPU tasks.

There will be ONE conflict though as the City Run Challenge (on PrimeGrid AP27 subproject only which can use CPUs and GPUs) starts at 00:00 on 8th May, so it'll be concurrent with the last 5 hours of the Einstein Sprint.

regards
Tim

[UBT - Timbo]

There goes my team lead for yoyo on FormulaBoinc Marathon

Hiya

Keep an eye on this stats page:

https://formula-boinc.org/equipes_v3.py ... &projet=59

Currently (as of time of writing this):

Code: Select all

Posn	Member			FB Pts	Credits (this year)	Credits (3rd May stats)
1	UBT - wbiz		25	169,599			140,478
2	Stephen Hunter		18	58,991			15,992
3	UBT - Timbo		15	55,741			32,433
4	CaffeinatedSloth	-	0

So, Stephen has slipped into 2nd place, but wbiz is still over 110,000 credits ahead !

regards
Tim

[stephenhunter]

It's all gone TU for me, Einstein has stopped as well. Think it may be something to do with getting OpenCL working yesterday. That was in an effort to get Prime AP27 tasks to run.

Think I will have to roll it back and try again

[wbiz]

Not all yoyo tasks have regular checkpoints I can lose 7 hours work sometimes with infrequent checkpoints.

I swapped everything to Einstein for a few hours but I can't get worthwhile credits out Einstein so I'm back on yoyo, it has got me on the sprint score board and that's about it.

I had to switch off ECM 2 tasks from yoyo, some were getting up to 8GB ram usage which is crazy.

[wbiz]

Obstacles 00:00 7th - 00:00 8th (UTC), 50% bonus on yoyo points (ECM/ECM 2) for the Obstacle Run.

.... but of course the Javelin Throw is running at the same time using SRBase (not TF)

[UBT - Timbo]

.... but of course the Javelin Throw is running at the same time using SRBase (not TF)

Hiya
But don't forget that the Javelin Throw will take place over multiple days and (IIRC) only the 2nd best credit score (per team) will actually earn points on the Pentathlon.

So, concentrating on yoyo makes more sense (compared to SRBase) when there's a "clash" for CPU processing time.

regards
Tim

[UBT - Timbo]

Hi all

We're currently in 24th place (out of 30) in this years Pentathlon as per their stats page here:

https://www.seti-germany.de/boinc_penta ... athlon.php

So, if you fancy joining in and helping us "leap up the league table" , please keep an eye on the following website to see which projects should be crunched and on which days. (Each days event starts from midnight 00:00 and lasts for 24 hours).

https://www.seti-germany.de/boinc_pentathlon/start.php

regards
Tim

[UBT - Timbo]

Hi all

Sadly, it looks like my work PC has lost it's internet connection, so I do not think it will contribute much, if anything, until I am back at work on Tuesday (it's at my office about 20+ miles away).

But it might at least have crunched its cache of tasks so there might be a glimmer of hope, credits wise, but I'll still have lost maybe 2-3 days potential crunching.

Cést La Vie, I guess.

regards
Tim

[wbiz]

Bad luck Tim that you've lost the tasks on bonus day and Stephen on his tribulations as well. This is the first challenge where everything is going smoothly for me with only a few tiny operator errors so far.

We and others were overtaken by Crystal Dream who did a lot of bunkering prior to bonus day. I bunkered a small amount but some tasks (ecm_cn) have no checkpoints at all which prevented me bunkering more.

EDIT: Then everything went wrong and we've dropped loads however we are pulling back again.

[stephenhunter]

Glad someone is holding the fort. I think your Yoyo lead is safe

[wbiz]

Glad someone is holding the fort. I think your Yoyo lead is safe

Others in the team are putting in a lot more muscle than me, I have about 20% of the UBT Pentathlon yoyo points but only 0.34% on FB.

Tried to bunker more for the next bonus day but yoyo issued me the really long tasks with no checkpoints I guess I could abort them but I have a feeling that could backfire in some unforeseen manner, I don't know what the etiquette is regarding aborting loads of tasks.

[ChelseaOilman]

Tried to bunker more for the next bonus day but yoyo issued me the really long tasks with no checkpoints I guess I could abort them but I have a feeling that could backfire in some unforeseen manner, I don't know what the etiquette is regarding aborting loads of tasks.

Are those the ECM P2 tasks? After changing my preferences to only accept ECM tasks I aborted the few ECM P2 tasks I received when I first started up after noticing they were taking up so much memory. How many points do the ECM P2 tasks earn and how does the crunching time compare to the ECM tasks?

[wbiz]

Tried to bunker more for the next bonus day but yoyo issued me the really long tasks with no checkpoints I guess I could abort them but I have a feeling that could backfire in some unforeseen manner, I don't know what the etiquette is regarding aborting loads of tasks.

Are those the ECM P2 tasks? After changing my preferences to only accept ECM tasks I aborted the few ECM P2 tasks I received when I first started up after noticing they were taking up so much memory. How many points do the ECM P2 tasks earn and how does the crunching time compare to the ECM tasks?

Its the plain ECM application, I had to kill off ECM 2 because of ludicrous memory usage, some went up to 8GB per task on Intel.

Its the ecm_cn tasks are often very long and no checkpoints. around 50 hours on my Pi's and 12 hours on Intel at the moment.

From memory the ECM 2 applications were very roughly the same credit per second as ECM but they may have been getting bottle-necked because of memory constraints.

[ChelseaOilman]

Its the plain ECM application, I had to kill off ECM 2 because of ludicrous memory usage, some went up to 8GB per task on Intel.

Its the ecm_cn tasks are often very long and no checkpoints. around 50 hours on my Pi's and 12 hours on Intel at the moment.

I haven't noticed any ECM CN tasks on any of my computers. I'll keep an eye out for them.

[wbiz]

Well you learn something every day, something I never realised despite a couple of years of playing with settings......

If I set my boinc-client to buffer one day of tasks, it tries to give me tasks that take less than one day.

If I set it to download three days worth of tasks, it tries to download longer tasks of less than three days.

etc

When I say "tries", it seems to prioritise those rules but if it hasn't got tasks of relevant length then anything goes, it will underload or overload.

It might just be a yoyo thing?

[ChelseaOilman]

I set my client to Store at least 0.1 days of work and Store up to an additional 0 days of work. Probably why I didn't see any of the tasks you were getting.

[UBT - Timbo]

Well you learn something every day, something I never realised despite a couple of years of playing with settings......
.
snip

Hiya

Yes, if the cache is set quite low, then long deadline tasks seem to be ignored, which is understandable, even if a bit cock-eyed.

There are loads of algorithmy things "going on" with BOINC Manager, when it comes to how and why it downloads specific tasks and with what frequency or which projects are "prioritised".

And to be honest, I gave up trying to figure it out, so I created my own algorithm, which is to only allow tasks from one project for CPU tasks and another project for GPU tasks. These projects are then set to 100% resource share and maybe a 1-2 day buffer.

And then I also allow work from some "backup" projects, but which are set to 0% resource share - this way, if tasks from either of the main projects (set at 100% RS) are not available, the backup projects will kick in and only one task is then downloaded (per processing unit). Once that task is uploaded, BM with then try to download a main project task and if not possible, another backup task will be downloaded.

This keeps things under my control and I can just leave it alone, until a Sprint or some other worthwhile Challenge comes along

regards
Tim

[wbiz]

I'd never noticed the long tasks for long caches before, it does make some sense but there are other mechanisms in boinc specifically for this eg rosetta has a target length.

The boinc-client scheduler is madness but its not until I came across multi-threaded tasks that I realised how complex the problem is.

I've also been tempted write my own controller as the API is reasonably simple, its on my endless list.

I tend to control from boinctui but also edit the project settings, global overrides and project overrides quite often which is easy enough when I running my usual two computers, when I'm running four or five it becomes a bit of a chore. Boinc manager does some weird things, like removing the override links from /var/lib/boinc-client to /etc/boinc-client which irritates my ocd.

When the pentathlon is over I'm going to again try and get my Intel hd gpu going, although most projects have either dumped Intel support or use Intel ARC, some projects still appear to support Intel hd, I have 2GB of GPU memory which should be usable. Its a shame projects don't also have a generic OpenCL offering instead of GPU specific. It would make sense to develop the generic OpenCL first then modify to the various specific models, not that I know anything about OpenCL.

[ChelseaOilman]

I'd never noticed the long tasks for long caches before, it does make some sense but there are other mechanisms in boinc specifically for this eg rosetta has a target length.

Well maybe because of the Pentathlon Yoyo decided to come out with all these days long ECM tasks to cut down on their server load. All the shorter ECM tasks I was getting were replaced with really long running ECM CN and ECM RU tasks. In my case all they accomplished is me switching over to seiver tasks in the preferences and I aborted all the long running tasks. So the Pentathlon is over for me.

[wbiz]

Thanks anyway for the huge boost in points you gave us. While long tasks are frustrating its the long or non-existent checkpoints that really hack me off, just had to reboot one machine for non-boinc reasons but that was instantly 23 cpu-hours down the pan. Obviously some tasks are difficult to checkpoint and I don't expect one-minute checkpoints for every task but I'd hope for at least one-hour checkpoints.

There are a lot of complaints the way some projects treat the boinc volunteers for many different reasons and I can often see why. Communication for some projects is appalling with months going by without hearing anything from the project personnel, other projects are the opposite and have fantastic support and communication but unfortunately the competitive point system pushes volunteers to not be selective about who they donate their computer time to. The boinc team themselves stay clear of ethics, while I don't blame them I don't think is entirely right for what is a massive community project.

[UBT - Timbo]

Thanks anyway for the huge boost in points you gave us.

While long tasks are frustrating its the long or non-existent checkpoints that really hack me off, just had to reboot one machine for non-boinc reasons but that was instantly 23 cpu-hours down the pan. Obviously some tasks are difficult to checkpoint and I don't expect one-minute checkpoints for every task but I'd hope for at least one-hour checkpoints.

Hiya

It can be very frustrating, crunching tasks and finding that ones hardware crashes and work to date is lost. One would have thought that checkpointing would have been implemented across every BOINC project and at reasonably regular times...but I guess this doesn't happen due to the extra stress it puts on the project servers (as every checkpoint must be reported back to the project, so limiting such communications reduces the need for network bandwidth, as otherwise, this might be a significant cost for certain projects) ADMIN EDIT

And of course longer crunching times worsens the problem, if something untowards happens (either due to the tasks, or the host environment).

(And speaking of which, my work PC is now back-online and should be contributing more to the Yoyo and PrimeGrid "runs" ).

There are a lot of complaints the way some projects treat the boinc volunteers for many different reasons and I can often see why. Communication for some projects is appalling with months going by without hearing anything from the project personnel, other projects are the opposite and have fantastic support and communication but unfortunately the competitive point system pushes volunteers to not be selective about who they donate their computer time to. The boinc team themselves stay clear of ethics, while I don't blame them I don't think is entirely right for what is a massive community project.

It is clear that some projects have a very professional team of admins and support staff on hand to keep every one happy. And of course some projects are run on a shoe-string budget and on lower quality hardware. In the end, it is down to each volunteer to decide who they want to donate their CPU/GPU cycles to, but it is obvious that some projects could do a lot better to assist the volunteers with feedback and general information that can help get the best from all the contributions.

regards
Tim

[wbiz]

due to the extra stress it puts on the project servers (as every checkpoint must be reported back to the project, so limiting such communications reduces the need for network bandwidth, as otherwise, this might be a significant cost for certain projects).

Are you sure about that? Some projects do a near realtime checkpoint and the recommended setting is 1 minute iirc, that would be a massive stream of data which I've never seen happen.

Perhaps there is some cross-talk between disk_interval and checkpoint interval with what boinctui calls checkpoint time?

[UBT - Timbo]

Are you sure about that? Some projects do a near realtime checkpoint and the recommended setting is 1 minute iirc, that would be a massive stream of data which I've never seen happen.

Perhaps there is some cross-talk between disk_interval and checkpoint interval with what boinctui calls checkpoint time?

Hiya

I stand corrected !!

It's always been my understanding that checkpoints did involve BOINC Manager sending data back to the project.

However, looking into this a bit, there are "checkpoints" and there are "checkpoints".

So, most checkpoints do save data to the local drive, so that in the event of a system hang or crash (or some other issue) when the host reboots, BOINC Manager simply restarts crunching the tasks from when the previous checkpoint was saved locally. This can be set within BOINC Manager Preferences.

On reflection, I think I was getting my facts a bit cock-eyed, as there have been some projects (such as CPDN), where some data (aka a "trickle") was regularly uploaded to the project during the crunching, so that intermediate credits could be awarded prior to the entire task being completed.

And this is where I think I got my wires crossed. Sorry about that !!

regards
Tim

[wbiz]

Yes, you are right, there are both types and a clash in terminology in places.

At last, some non Cunningham Number tasks are popping up, my sympathy goes to Cunningham and his team who did these things by hand but some of these guys had shortcut tricks up their sleeves that were never published.

[UBT - Timbo]

Hiya

So, the current Yoyo stats from here:

https://formula-boinc.org/equipes_v3.py ... &projet=59

are currently showing the following (as of time of writing this):

Code: Select all

Posn	Member			FB Pts	Credits (this year)	Credits (3rd May stats)
1	UBT - wbiz		25	275,418			140,478
2	UBT - Timbo		18	86,683			32,433
3	Stephen Hunter		15	61,0811			15,992
4	CaffeinatedSloth	-	0

So, Timbo is now in 2nd place, but wbiz is nearly 200,000 credits ahead !

I've also had to limit the number of Yoyo tasks I crunch concurrently, as my 16 core work PC keeps locking up if it tries to crunch even 2 tasks, as the memory requirement can reach over 3GB per task !! (and yes, I will need to increase the onboard RAM from 8Gb to run more Yoyo tasks )

regards
Tim

[wbiz]

On the 4-core PC I have 16GB dram since I had memory problems with 8GB on the first FB challenge I did.

The three Raspberry Pi4's have 4GB, two are running 4 cores and the other is running 3 cores because I'm using that one as my desktop.

I haven't switched the 2 core PC on, the throughput on yoyo is even slower than the Pi's and it is nowhere near as energy efficient as everything else.

Everything is running comfortable temperature-wise, the PC's fans are almost at idle speed and temperature is 72C, the Pi's are around 68C with passive cooling.

Half way through, I thought time would fly by as I've been busier than normal maintaining/restoring/rescuing gardens - I'm not a gardener, I just like physical graft.

[wbiz]

Bonus day on the marathon yoyo@home ecm tasks, but the Repunit Numbers tasks are coming through which are even longer than Cunningham Numbers, however I missed an opportunity to bunker as RU task only use a small amount of memory.

As I have yoyo tasks that are not going to complete during the bonus day I have suspended them to memory and switched to numberfields@home for the sprint to regain the place that Boinc Confederation took but Boincstats look like they are about to overtake as well at a much faster pace.

[UBT - Timbo]

Hiya

I've just switched my GPU over from PrimeGrid to NumberFields - would have done it sooner, but I was in Liverpool all of yesterday (for Eurovision) and didn't get home until v. late

As for yoyo? Diespite what their server says here:

https://www.rechenkraft.net/yoyo/server_status.php

Code: Select all

Application	unsent	in progress
Cruncher ogr	0	4
ecm		12,011	161,902
Siever		8,168	8,301
M Queens	1,612	11,183
ecm P2		3,796	21,953

I cannot get any tasks to fill my bunker and I have just one task crunching

I suspect there is a high demand for tasks and with another 5 days of this run to go, I think a lot of participants are just keeping their bunkers full which then limits how many tasks are available.

It is also a shame that the various ECM sub-tasks, such as hc, cn and ru, cannot be chosen by participants in their yoyo preferences. So, those members with older hardware have to put up with running longer, more memory intensive tasks, which in my case causes my host to lock up, due to memory requirements.

regards
Tim

[wbiz]

It will certainly pay to bunker yoyo because the 100% bonus day is yet to come but high memory and infrequent checkpoints will make bunkering difficult.

Switching to yoyo on bonus day may be futile because the long tasks will limit how many will complete in time.

NumberFields doesn't mention GPU on its applications page but I did notice others with cuda suffix in their stats pages.

If you are firing up your GPU on Numberfields I may as well swap back to the yoyo slog if I can get tasks, I think our Obstacle Course position is fairly safe but you never know.

I've not ventured across the water to Eurovision but been out and about and could see Liverpool from a distance most of the time, I try to have a quiet life,

[wbiz]

NumberFields gone down for maintenance, requesting 3600 seconds.

[UBT - Timbo]

NumberFields gone down for maintenance, requesting 3600 seconds.

Hiya

I've got no issues with downloading NumberFields tasks right now.

Maybe it's just a transient error somewhere in the internet?

regards
Tim

[UBT - Timbo]

HI all

I just uploaded my bunker of NumberFields and Yoyo tasks from my work PC.

I doubt it will do much to improve our overall standings, but it's something

As from midnight tonight, it'll be just Yoyo and Einstein@home tasks that will count towards our team total.

(And on Thursday 18th May at 23:00 UTC, the next FB Sprint will start, so if anyone has some spare capacity, please join in as it would be very helpful - see below for details).

viewtopic.php?p=134879#p134879

regards
Tim

[wbiz]

Just what we needed was a boost on the sprint to prevent us from ending up second to last in the overall standings, BoincPoland and BoincStats both looked like they could have overtaken us on the sprint.

We are unlikely to overtake or get overtaken on the Cross Country (Einstein), a few UBT members have been helping us a lot.

I'm a bit worried that the Obstacle Course (yoyo) could have a lot of bunkered points in other teams (eg TSBT) which will knock us down the table on Bonus Day.

With any luck I'll be swapping my computers around before Thursday, the core 2 duo will be off boinc and swapped with one of my other computers, I doubt I can do it before the Pentathlon ends there are some network changes that need to take place as well.

I doubt I'll ever go for a two week challenge again, its been interesting but overall its just a long slog and stops me being able to make changes to my systems.

[UBT - Timbo]

Hi all

The stats for the Pentathlon make for some interesting reading:

https://www.seti-germany.de/boinc_penta ... +Team.html

So, here's where we stand (at day/time of posting):

Obstacle Run - yoyo@home - ongoing until 19th

Started well, with some good credits on the 4th and 5th days, but days 1-3 and 6 onwards have been pretty flat with no significant credits earned. The run continues until the 18th at 23:59pm UTC, so more effort here would be useful. Currently in 20th place.

Javelin Throw - SRBase - run ended

A few good credits on days 1 and 2, but nothing of significance since, maybe due to an overlap with other GPU based projects? Ended in 29th place.

City Run - PrimeGrid - run ended

Many credits earned on most days, with more earned on days 4 and 5. A good overall effort, but still only achieved 28th place overall.

Sprint - NumberFields - run ended

Good effort on 1st and 3rd 3 days but I bunkered my day 2 credits (and uploaded before the deadline on 15th at 23:59pm UTC) to boost day 3 efforts. Achieved 25th place.

Cross Country - Einstein@home - ongoing until 19th

Good effort so far and run continues until 19th May. 26th place so far, but likely to be overtaken soon.

We've never done very well on the Pentathlon and this year continues the trend

regards
Tim

[wbiz]

I didn't see the Einstein overtake coming

[wbiz]

100% Bonus day on the Obstacle Course (yoyo ecm/ecm2).

We've immediately been overtaken on the Obstacle Course by LITOMYSL debunkerering.with a one million credit jump - wow!

This is far too tactically complicated for me, its the third time I've chosen the wrong tactic, reminds me of the gambling rule "don't chase your losses"

[UBT - Timbo]

100% Bonus day on the Obstacle Course (yoyo ecm/ecm2).

We've immediately been overtaken on the Obstacle Course by LITOMYSL debunkerering.with a one million credit jump - wow!

Hiya

It is a shame that very few UBT members want to take part in these team challenges...but I recognise it isn't the sort of thing some people want to do and of course BOINC is not exactly the highest priority in peoples lives these days.

So, actually competing against some other teams is always going to be tough and of course some teams have significantly more computing power at their disposal.

This is far too tactically complicated for me, its the third time I've chosen the wrong tactic, reminds me of the gambling rule "don't chase your losses"

The one thing that has never been implemented within BOINC Manager is a user-defined scheduler, so you can actually "program" when the host can be used and for which projects - that would then allow everyone to get the most out of these time and project based challenges and so, once set you can then just forget about when you need to switch projects and leave the host in "automatic mode".

regards
Tim

[ChelseaOilman]

Not that it will make much difference but since I passed 10 million points in the Siever subproject last night and got my red star I switched back to ECM tasks. I have another 5 million points to go to reach my goal of 25 million total points in Yoyo.

[wbiz]

Forgot about the end of the Pentathlon because I was re-configuring for FormulaBoinc

We held on to our somewhat disappointing 27th place overall

Obstacle Course (yoyo@home) 19th
Javelin Throw (SRBase) 29th
City Run (PrimeGrid) 28th
Sprint (numberfields@home) 25th
Cross Country (einstein@home) 28th

Total credits over the two weeks of 9,639,811 (688,558 per day)