Formula BOINC 2023

[ChelseaOilman]

But this would of course have been an ideal place to set up an "all in one" page where team members could chat about anything - and of course it would mean there was no point having THIS forum...

You mean like Discord where BOINC, Gridcoin, BOINCGames, PrimeGrid, UKBOINCTeam and others are located? I check there more often than here which is typically a ghost town.

[UBT - Timbo]

You mean like Discord where BOINC, Gridcoin, BOINCGames, PrimeGrid, UKBOINCTeam and others are located? I check there more often than here which is typically a ghost town.

Hiya

Well, we've got a Facebook page, a Twitter account and a Discord page...as well as this place.

BUT: They all rely on members spending a few minutes and actually join up with one or more...there's no "automatic" way of signing up members to a single project-wide "website" that accumulates every members email address and allows founders to keep members interested/informed etc.

And of course, to put it bluntly, many people cannot be bothered to sign up for some 3rd party website and of course many, many people signed up for BOINC and they didn't put in enough effort and dropped out for many and various reasons.

And then there's also the problem of keeping each and every social media "page" updated, keeping stuff fresh and generating interesting topics and subjects to keep people involved and hopefully acquiring new members to take part.

Yes, this forum has seen better days, especially when it first started up with multiple posts from many members every day. But since then, other social media platforms have sprung up and each one has it's own "features and benefits" and ultimately this has meant members have become accustomed to only checking in on their preferred platform, rather than spending their time on multiple websites.

And from an admin point of view, there's only so many hours in a day...

regards
Tim

[ChelseaOilman]

I'm not a fan of Facebook or Zuck and never really got into Twitter. The thing with Discord or Slack is it's like instant messaging. You can carry on a conversation. I still follow Slack because my old team is there. Discord is the best platform for talking to people that are active in BOINC. And you can keep up with the latest info there. Think of it like your phone. If you want to contact someone quickly, would you email them or send them a text.

[UBT - Timbo]

Hi all

The next FB Sprint starts tomorrow (Friday) at 5am (UTC).

So, if you have some spare processing time, please check below to see which project has been randomly selected:

https://formula-boinc.org/sprint_v3.py? ... &year=2023

regards
Tim

[ChelseaOilman]

I'm in the process of shutting things down in scorching hot Texas. Heading to the much cooler temps in the Colorado Rockies at 9,000 feet elevation. I'll have 4 computers when I get there and may build another during the summer. My output will be greatly reduced.

[wbiz]

yoyo@home for the sprint

One of my old favourites (ARM friendly)

[wbiz]

Had to bomb yoyo's ECM P2 tasks, they take too much memory.

[UBT - Timbo]

I'm in the process of shutting things down in scorching hot Texas. Heading to the much cooler temps in the Colorado Rockies at 9,000 feet elevation. I'll have 4 computers when I get there and may build another during the summer. My output will be greatly reduced.

Hi Carl

No worries.

Have a safe journey up to Colorado and thanks for your continued support - it is much appreciated.

And don't worry about your "output"...as you know *every* contribution, no matter whether it is large or small, really helps the team.

regards
Tim

[UBT - Timbo]

Had to bomb yoyo's ECM P2 tasks, they take too much memory.

Hiya

Yup, one has to be a bit careful to ensure the tasks that you choose to process are within the capability of each hosts hardware.

But it is the same for everyone, although more recent PCs will have a slight advantage in terms of on-board memory and CPU processing ability.

regards
Tim

[wbiz]

Thought I was by myself but Chris has just stepped in - good man.

The team is also getting some points from Alan and Gordan who are running yoyo routinely at the moment.

[UBT - Timbo]

Thought I was by myself but Chris has just stepped in - good man.

The team is also getting some points from Alan and Gordan who are running yoyo routinely at the moment.

Hiya

Sadly, and despite your efforts and those of Chris and others, as of now (just gone midnight, after 19 hours of this Sprint, UBT is in 15th place and way behind the 14th placed team:

Code: Select all

Team			Credits	Posn
The Scottish Boinc Team	865,138	1
Rechenkraft.net		676,667	2
Planet 3DNow!		457,878	3
Ukraine			304,020	4
H.P.C.N.Swiss		229,231	5
BOINC.Italy		219,431	6
BOINC@AUSTRALIA		213,722	7
LITOMYSL Boinc Team	193,831	8
L'Alliance Francophone	188,126	9
AMD Users		176,011	10
Czech National Team	99,332	11
SETI.Germany		97,091	12
SETI.USA		86,741	13
The Final Front Ear	82,831	14
**UK BOINC Team		27,337	15**
meisterkuehler.de Team	10,106	16
BOINC Synergy		9,973	17
BOINC.BE		7,481	18
Crunching@EVGA		3,283	19
Crystal Dream		1,822	20
TSC! Russia		624	21
Canada			84	22

I've allowed my multi-core Xeon host to take part (and limiting it to 10 concurrent tasks due to memory limitations), but each task is taking about 14,000 seconds (a shade under 4 hours) and each task earns just 252 credits. So my 10 hosts are earning 630 credits per hour.

So, over 3 days (72 hours) each of my CPUs can process 18 tasks, so that 180 tasks @ 252 credits = that's about 45,000 credits I can contribute (and assuming they are all validated in time !).

And working back TSBT's credits - their 865,138 credits earned in 19 hours = 45,533 credits per hour....equal to abut 180 active hosts (if my maths is correct and based on each task earning the same 252 credits !!)

So, it might take some more firepower to get UBT into the Top 10 on this Sprint !

regards
Tim

[UBT - Mikee]

Hi

Still here, but much reduced firepower due to real life, not that I had much compared to other wanyway!

Late to the party - added 14 cores, but as they take 4 hours each and wingman required, may not have that much impact.

Mike

[wbiz]

but as they take 4 hours each

If that were only the case, there are tons of long tasks that land as well, they range from 4 hours to 39 hours. I only get the long tasks on my Intel computers. When I started off the sprint with the initial download one computer only got long tasks and another computer only got short task which was weird, had the same with the last yoyo sprint as well.

For ecm tasks the points range from 250 to 900 roughly.

I'm playing whack-a-mole with the 39 hour tasks.

[stephenhunter]

Late to the party again. Been away sailing.
Will try to catch up, but Yoyo doesn't seem to agree with me

[UBT - Timbo]

Hi

Still here, but much reduced firepower due to real life, not that I had much compared to other wanyway!

Late to the party - added 14 cores, but as they take 4 hours each and wingman required, may not have that much impact.

Mike

Hiya Mike

Hope you are well and maybe enjoying the extra tourism in your neck of the woods !! (courtesy of messrs McElhenney & Reynolds )

As you know, every little bit helps on these sprints so, a few extra cores is always welcome !

The quorum (of 2) might be a pain, but the normal ECM tasks have a quorum of 1 and they take maybe fewer/less resources that the P2 tasks - but it's the same for everyone so, we can't really grumble (to much ).

regards
Tim

[UBT - Timbo]

but as they take 4 hours each

If that were only the case, there are tons of long tasks that land as well, they range from 4 hours to 39 hours. I only get the long tasks on my Intel computers. When I started off the sprint with the initial download one computer only got long tasks and another computer only got short task which was weird, had the same with the last yoyo sprint as well.

Hiya

I've had ONE much longer ECM task that took 37,720.25 seconds and earned me 670 credits. All the rest have been around the 13,000 second mark.

For ecm tasks the points range from 250 to 900 roughly.

I'm playing whack-a-mole with the 39 hour tasks.

So, maybe it's just the luck of the draw? As I don't think there is any subterfuge going on but I cannot be 100% sure of course

regards
Tim

[UBT - Timbo]

Late to the party again. Been away sailing.
Will try to catch up, but Yoyo doesn't seem to agree with me

Hiya

No worries, but glad you are on board.

I have found that there are times even where I've limited the number of tasks being processed, where my multi-core host just stops responding to anything else I try to do on it (other than BOINC)...so there is clearly some fairly intensive processing going on, that saps CPU bandwidth and uses lots of the main memory.

Maybe you should just limit how many concurrent YoYo tasks can be run - starting with 1 tasks running on just 1 CPU and see how it goes? Then increase it to 2 and see if you notice any sluggishness or other effects? Keep adding to the number of running tasks and see when things get much more sluggish. (Use Windows Task Manager to monitor CPU and Memory performance? - or similar on Linux).

regards
Tim

[wbiz]

The more the merrrier!

I've put my Intel's on Siever only, a bit more risk of needing validation but a 39 hour ecm task won't end in time for the end of the sprint. Generally the more ecm tasks you do, the more you get landed with the long tasks from my observations (but its easy to come to the wrong conclusions).

I don't pretend to fully understand the scoring systems (FB or BG) but as far as I can see we are a no-hoper of getting points out of this yoyo sprint. HOWEVER, we are on two points in the yoyo marathon so any yoyo credit gained helps us maintain those points.

Or are my wild guesses completely out of line?

[UBT - Timbo]

Hi all

Sadly, my (remotely located) work host is giving out

Code: Select all

Client error	Compute error

error messages on completed/returned tasks, some of which have taken many thousands of seconds to "complete".

(The stderr log shows "No heartbeat from core client for 30 sec - exiting")

I suspect the host has "locked up" and I cannot do a remote reboot

So, I doubt I can continue to contribute until the end of the Sprint...and if I do it will be by luck more than anything.

regards
Tim

[wbiz]

Painful isn't it!

[ChelseaOilman]

Left Texas saturday at 6am for the 1,000 mile drive to Colorado. Drove strait through and arrived at the Colorado home at 11pm same day. I had one computer already here so just fired up some yoyo tasks. Should be able to get a few in before the sprint ends. 2990WX system with 32C/64T. Looks like I got almost all short tasks. Aborted the few long tasks I received.

[stephenhunter]

I managed to get a few tasks done, in amongst the frequent lock up's!
Hoping that's the last we will see of YoYo

[wbiz]

I managed to get a few tasks done, in amongst the frequent lock up's!
Hoping that's the last we will see of YoYo

YoYo s a bit more entertaining than switching on and just watching the eleccy meter race round, but it does make you pull your hair out at times.

Back calculating everything I only achieved 63% of the credits I would have got if I'd started off with the correct choices. I remember the same conclusion last time I was fighting Yoyo, one day I'll write some notes - that's when I figure out somewhere I can put notes so I can find them when I need them.

We got back to 15th on the sprint after a bit of a lapse but we've gone down a point on the Yoyo marathon - those kilt wearing haggis eaters with the dirks and sporans threw a caber right past us (I'm half Scottish and partner is Scottish)

[stephenhunter]

[UBT - Timbo]

Hi all

The chosen project for this weekends Sprint is SiDock.

You can check how well we are doing here:

https://formula-boinc.org/sprint_v3.py? ... &year=2023

regards
Tim

[ChelseaOilman]

I'm out for this sprint. Started to download tasks and realised there was only LOOOONG tasks available. Aborted them and went back to Denis.

[stephenhunter]

Same here, it's too painful.
I will finish the few got first.

[wbiz]

I lost a week somewhere, I thought I had another weekend without challenges. Weirdly I changed one computer to SiDock last night, I'll watch its beard grow.

[UBT - Timbo]

I'm out for this sprint. Started to download tasks and realised there was only LOOOONG tasks available. Aborted them and went back to Denis.

Hi Carl

Yup, the SHORT tasks seems only to be available for Linux running on ARM CPUs.

The LONG tasks are for the above hardware as well as Windows and Linux on ARM using say AMD/Intel CPUIs.

https://www.sidock.si/sidock/apps.php

regards
Tim

[wbiz]

I'm out for this sprint. Started to download tasks and realised there was only LOOOONG tasks available. Aborted them and went back to Denis.

Hi Carl

Yup, the SHORT tasks seems only to be available for Linux running on ARM CPUs.

The LONG tasks are for the above hardware as well as Windows and Linux on ARM using say AMD/Intel CPUIs.

https://www.sidock.si/sidock/apps.php

regards
Tim

But the SHORT tasks are estimating 45 hours on my ARMs, tasks are 32-bit.

The estimates for Intel are a bit weird, coming up with 7 days on one computer and 39 hours on another and I think both are overestimates, I'm guessing and hoping they will end up finishing at 30 to 31 hours, only time will tell.

[ChelseaOilman]

Long tasks are fine for the marathon but projects chosen for a 3 day sprint should have short tasks. Just my $0.02.

[UBT - Timbo]

Long tasks are fine for the marathon but projects chosen for a 3 day sprint should have short tasks. Just my $0.02.

Hi Carl

I do agree with you to a point.

And I did already answer this query on the FB forum as per here:

FYI: There is something else that also needs to be considered: Many projects are no longer active or have limited numbers of available tasks. And those projects that do have lots of tasks have mostly been chosen already this year, so whilst the project selection process is random, there are only so many projects that can be chosen that will be suitable for selection on a Sprint.

So, it is a juggling act to ensure that only those projects that are actually active (when the selection process takes place) can be considered (and hence inactive projects are not considered for a Sprint).

However, some good team and individual members credits *are* being earned across the board on this Sprint so it's not ideal but it is at least helping the project with a significant increase in uploaded and completed work units, even if the time to complete each task is longer than some other project tasks. Plus the Sprint is over 72 hours so, some members with higher spec'd CPUs could be returning more tasks within the period.

https://formula-boinc.org/sprint_v3.py? ... &year=2023

HOWEVER, I do recognise that many of these credits might be being earned by regular credit contributors to this projects tasks, rather than just the members who have switched to this project just for the Sprint.

regards
Tim

[ChelseaOilman]

SiDock would have been a good project to speculate on and bunker prior to the sprint project being announced.

There's 13 sprints left and only 5 projects that haven't been chosen yet.
DENIS@Home
Gerasim@home
LHC@home 1.0
Moo! Wrapper
PRIVATE GFN SERVER
Two of those I'm currently doing and plan to keep running all summer as long as tasks are available.

Einstein has already been chosen twice. With more sprints than unchosen projects left there will be more duplicates. If I think of it I may bunker the projects I'm doing a few days just prior to a sprint being chosen just in case I get lucky.

[ChelseaOilman]

Next sprint is only 2 days. Why is that? Better not be SiDock again! LOL

[wbiz]

Next sprint is only 2 days. Why is that? Better not be SiDock again! LOL

Boinc Games sprint runs the day after FB so quite a long session adding the two together.

Just had my electricity bill, ten percent higher than my target value, I'll blame it on the washing machine.

[UBT - Timbo]

Next sprint is only 2 days. Why is that? Better not be SiDock again! LOL

Hiya

Actually it is THREE days as it starts at 00:01 on 7th July and ends at one minute before midnight on the 9th. So, that's 71 hours and 58 mins long (call it 3 days !!).

regards
Tim

[UBT - Timbo]

Boinc Games sprint runs the day after FB so quite a long session adding the two together.

Hiya

Well, there is about an 8 hour gap between the end of the Sprint and the start of the Boinc Games - so, maybe enough time to re-configure any bunkers and/or project preferences.

I wonder if BG will choose the same project?

regards
Tim

[UBT - Timbo]

SiDock would have been a good project to speculate on and bunker prior to the sprint project being announced.

Hiya

The actual Sprint results on SiDock were pretty good and a few teams got some great credits, so although the tasks were (in some cases) quite long, lots of work units were completed within the Sprint window.

There's 13 sprints left and only 5 projects that haven't been chosen yet.
DENIS@Home
Gerasim@home
LHC@home 1.0
Moo! Wrapper
PRIVATE GFN SERVER
Two of those I'm currently doing and plan to keep running all summer as long as tasks are available.

You should also note that the "possible" candidates list for each Sprint *does* change, depending on whether any advance warning is given as to whether a project (or it's website) will be undergoing maintenance.

And also, if it can be seen that a project has very few tasks available, then that will mean they get removed from the contender listing shown.

Einstein has already been chosen twice. With more sprints than unchosen projects left there will be more duplicates. If I think of it I may bunker the projects I'm doing a few days just prior to a sprint being chosen just in case I get lucky.

Indeed, it is even possible that there might be triplicate or quadruple instances where a project is chosen multiple times...and that is simply due to the limited number of projects that have lots of available tasks. This also means that if anyone wants to speculate on one or more projects, that the fact a project has not been chosen yet, is *not* a guide to the fact that previously unchosen projects have a higher chance of being selected.

So, you can imply from that that any project that has few tasks available will not be entered into the "random project selector".

regards
Tim

[UBT - Timbo]

Hi all

The chosen project for this weekends Sprint is MilkyWay@home.

You can check how well everyone is doing here:

https://formula-boinc.org/sprint_v3.py? ... &year=2023

and task availability can be seen here:

https://milkyway.cs.rpi.edu/milkyway/server_status.php

Please note that this project only issues CPU tasks now - no more GPU tasks are available.

regards
Tim

[ChelseaOilman]

Well that was painful. Milkyway on CPU is like watching paint dry. I'm used to getting millions of PPD on GPU tasks. Glad it's over.

On a side note, BOINC Games sprint starts tomorrow. Hopefully MW isn't chosen.

[UBT - Timbo]

Well that was painful. Milkyway on CPU is like watching paint dry. I'm used to getting millions of PPD on GPU tasks. Glad it's over.

Hi Carl

Yup, MW no longer has any need for GPU hardware, so it only has CPU tasks, which just happen to take ages. I think they will lose a lot of crunchers now, as there is no real benefit using CPUs on their tasks.

On a side note, BOINC Games sprint starts tomorrow. Hopefully MW isn't chosen.

I doubt it will be MW, as that project was chosen for the last BG Sprint, back in April:

https://www.boincgames.com/sprints.php

regards
Tim

[wbiz]

On the starting blocks for the sprint tonight

[UBT - Timbo]

On the starting blocks for the sprint tonight

Hiya

Excellent.

The project has been randomly selected and all should be well so, I'm keeping my fingers crossed there will be lots of interest.

BOINC.Italy has recently been beefing up their team on FB, as 3 new members registered this week (though this only affects the individual rankings, as team stats are based on the entire team membership and not just those members who register).

https://formula-boinc.org/forum/viewtop ... 3096#p3096

But more team members joining implies more individual crunchers on the Sprints.

regards
Tim

[wbiz]

Sprint Cancelled due to technical problems, shit happens ¯\_(ツ)_/¯

Onwards and upwards - to infinity and beyond

[UBT - Timbo]

Sprint Cancelled due to technical problems, shit happens ¯\_(ツ)_/¯

Onwards and upwards - to infinity and beyond

Hiya

Sorry, for the late update, but yes, this weeks Sprint (#13 - unlucky for some? ) has had to be cancelled as the back-office stats collection/display functions did not appear to be working both last night and this morning as well.

So, sincerest apologies for not advising this Sprint had been cancelled sooner and hope not too many have been inconvenienced by this.

And sadly, I had to go out earlier and I just got back; I've taken a look in the FB forum and only one person has whinged so far.

Hopefully the next one will be OK.

regards
Tim

[UBT - Timbo]

Hi all,

I've restarted this weekend Sprint on a "test" basis for 3 hours as from 7pm BST tonight.

And as of a few minutes ago, the Sprint #13 page is actually now showing team and member names (where before it showed no names at all).

https://formula-boinc.org/sprint_v3.py? ... &year=2023

So, maybe a gentle "switch it off and then back on again" (ie reset) has worked?

I will keep monitoring and maybe if after another hour it still works OK, then I will extend the Sprint by the missing (20 or so) hours (when no stats seemed to be collected).

22nd July @ 8pm BST UPDATE: The FB website Sprint #13 stats are now working properly (or so it seems).

The Sprint test is now extended until at least 7pm UTC tomorrow night and with a view to continue for another 48 hours after that, so that the full 72 hour Sprint duration will take place.

regards
Tim

[wbiz]

Almost immediately before it was cancelled, I spent half an hour trying to manually drag 2.5 days worth of tasks out of Universe which I had a good laugh at afterwards. I've learnt that with validated projects its best to load tasks up early as they have better chance of the buddy completing those in time. Plus Universe being flaky on issuing task I wanted a good cache,

After it was cancelled I decided to let them run anyway so there should be points continuing to come across.

Its provided some entertainment after the longer slog of BG.

[UBT - Timbo]

Almost immediately before it was cancelled, I spent half an hour trying to manually drag 2.5 days worth of tasks out of Universe which I had a good laugh at afterwards. I've learnt that with validated projects its best to load tasks up early as they have better chance of the buddy completing those in time. Plus Universe being flaky on issuing task I wanted a good cache,

After it was cancelled I decided to let them run anyway so there should be points continuing to come across.

Its provided some entertainment after the longer slog of BG.

Hiya

Wow - many thanks for your commitment and steadfast approach to the Sprint !

On the FB forum, just after the official start, member "entity" reported issues of downloading tasks and getting none. So I allowed one of my hosts to try and I had the same problem.

Eventually, about 3½ hours later (after the start) I got some tasks and these have been slowly processed...such that as of now, I have 40 tasks awaiting validation and ZERO credits received so far as none of my other 28 reported tasks have been validated (as the quorum is 2).

I suspect that I will wait a long time for my wingman to upload their tasks, as a few FB members may have cancelled/aborted many tasks and I will have to wait until these tasks are re-issued before receiving any credits, and this may not happen until well after the Sprint has ended.

And it has been fun trying to get the FB website working again... (And to be honest, I really do not know *why* it is working now...maybe I was just lucky ? I'll have to try the lottery this weekend and see if my luck continues )

regards
Tim

[wbiz]

I don't envy your position, you are damned either way on most decisions and kevlar is so expensive. I think you've learnt to make a decision and just let the rumble happen. I don't mind people moaning and complaining, its best to know what people think but getting personal is a no-no with me and should be on all of the internet.

With free speech comes responsibility, free speech does not give you the right to harass, abuse, belittle or be derogatory to others.

Suggesting Sebastian resign was funny, Sebastian isn't an employee or a committee member, he is an owner and as you quite rightly pointed out he has done a great service over many years, he has already done his "above and beyond".

Anyone can start their own site and I'm sure they would soon see things differently from the other side of the picket line.

[UBT - Timbo]

Hiya

Thanks for your kind comments.

My skin is actually a bit tougher than kevlar and I got it for free from my parents

I have also been lucky to have faced adversity and come through it without any scars, but gained a lot of experience in the process, so I can handle cowards and losers who think they can hide themselves away and spout off, thinking they can get away with it.

I feel very sorry for the less able who live in fear of such abuse and for those who cannot stand up to it, who suffer as a result. Some people cannot cope with this sort of abuse and this can result in suicide, when it gets too hard to take any more

Sadly, abusive cowards are totally irresponsible and do not realise the damage they can cause.

regards
Tim

[UBT - Timbo]

Hi all

23rd July @ 9pm BST UPDATE: The FB website is continuing to behave itself, so the Sprint will go the full 72 hours, but obviously the original deadline has been extended as a result of the original problem - so the 13th Sprint will finish tomorrow (24th July) at 17:59pm UTC.

regards
Tim

[wbiz]

A consistent 14th for the team with eleven members contributing.

I noticed that the winning team TSBT had a member who would have been top every day by himself without the rest of the team, unfortunately he has his computers hidden but it would be interesting to see what he uses. Was the google search on a go-slow over the weekend lol?

[Woodles]

Not an absolute indication but at least points in the right direction.

WCG lists the number of installations of their Boinc program.
Chelsea has installed WCG 84 times
I've installed WCG 24 times
You've installed WCG 2 times
and PecosRiverM has installed WCG 1,144 times

If that's once per host then you'll see why he scores a lot of points

[wbiz]

and PecosRiverM has installed WCG 1,144 times

If that's once per host then you'll see why he scores a lot of points

I find five computers running boinc enough of a headache, he deserves the points

[wbiz]

14th Sprint - gerasim@home

[UBT - Timbo]

14th Sprint - gerasim@home

Hiya

Thanks for staying up late and announcing this

I've got my work host crunching some tasks - and I have a mixture of both CPU and GPU tasks - the GPU tasks earn more credits per task than the CPU ones (which is quite normal).

regards
Tim

[wbiz]

I've got my work host crunching some tasks - and I have a mixture of both CPU and GPU tasks - the GPU tasks earn more credits per task than the CPU ones (which is quite normal).

regards
Tim

Aye, its a bit lonely, not a lot of support for the task, if others have had the same trouble registering with gerasim it doesn't surprise me, I had to opt out the last gerasim sprint as I couldn't get registered. I tried the day before this sprint and it still wouldn't work, then I managed to succeed after this sprint started.

So now I've broken my gerasim@home cherry, it looks like a reasonably generous project for credits and doesn't stress the CPU's much, servers have been brilliant and I love the additional project controls for number of tasks and downloads, validation is also a breeze. It might become my credit-grabbing project for x86.

There doesn't appear to be a massive difference in credits for GPU vs CPU unlike some projects, hence my CPU's have been dipping in and out of the points boundary, I was chuffed to be 8th for a while, short-lived illusions of grandeur may have happened

Both Free-DC and Boincstats are very slow in picking up credits for this project (once per day?) which is weird as FB is getting the hourly updates fine.

x86 only so my Pi's are picking up some small but rare pickings on Rosetta Beta.

[UBT - Timbo]

Hiya

On CPU tasks, I'm getting 3x the Claimed Credit being awarded as Granted Credit, but this is a lot lower than the GPU credits.

Plus the CPU based credits are variable and proportional to the length of CPU time each task takes.

So, ranging from 42 Granted credits for a task that takes 1,500 seconds, up to 434 credits for a task taking 15,500 seconds !!

The NVidia (opencl_nvidia) GPU credits seem to be fixed/constant with processing times of between 361 seconds up to 1,650 seconds all claiming 50 credits, but being Granted 500 Credits

But GPU tasks do not appear to be available 24/7, so they are either in short supply, OR they being downloaded by people filling up their caches...leaving no more for anyone else?

But at least the tasks are now available on various platforms, as there used to be only tasks for CPUs on Windows OS hosts.

https://gerasim.boinc.ru/users/viewApps.aspx

regards
Tim

[wbiz]

Looking at Gerasim stats to calculate average credits per second per core for decic tasks

Windows CPU - 0.05
Nvidia GPU - 1.63
Amd GPU - 0.69
Linux CPU - 0.22

My 4790S on linux at stock speed (3.2GHz) are getting 1.81 per CPU = 0.23 per core [edited]

I guess the Windows figures are low because there are a higher percentage of Windows computers running boinc in the background and/or are not on 24/7, there are probably other factors, perhaps laptops are more windows, higher powered 24/7 CPU's are more likely to be linux(?) etc.

[wbiz]

The first time and probably the last

[UBT - Timbo]

The first time and probably the last

Hiya

Don't worry - there's still a few more Sprints to go and the Marathon carries on until the end of December, so plenty of time to accrue more FB points (and assuming you continue crunching, of course !!).

regards
Tim

[wbiz]

Never mind how I worded it, it sounded like a goodbye statement, which its not.

I plan to carry on with sprints this year and probably longer.

My main focus in between sprints will be biology/medical projects ie WCG, Denis and Rosetta (SiDock and I have never got on very well), sometimes those projects have been sparse so I had a little bit of opportunistic fun overtaking Stephen while he's off in GPU mega-credit land.

Hopefully electricity prices will come down further and more members will get involved in the sprints.

[wbiz]

just over 24 hours until next sprint

24th Aug 23:00 (UTC) - 27th Aug 2023 22:59 (UTC)

[wbiz]

SRBase? Who is that????

[UBT - Timbo]

SRBase? Who is that????

Hiya

It is this project:

https://srbase.my-firewall.org/sr5/index.php

(If you do not have an account already, then register via their website FIRST, noting the invitation code on their homepage, and then afterwards, you can add the project to BOINC Manager as per the normal method).

Also:

There is a known issue crunching some Linux tasks.

Check their message board for more details:

https://srbase.my-firewall.org/sr5/foru ... hp?id=1790

rebirther, the SRBase admin, knows about this and is working towards fixing it. However, do not be surprised if you run Linux hosts and issues arise - so maybe set up a VIrtual (Windows) "box" and you should be OK.

regards
Tim

[wbiz]

Well that was weird, I already had an account but no credits???? Didn't have it on any of my boinc computers.

I have some credits now lol

I don't do virtual boxes, I'll just keep a hammer nearby to smash anything that goes weird. I can't see the linux post you are talking about - its probably staring at me!

[UBT - Timbo]

Well that was weird, I already had an account but no credits???? Didn't have it on any of my boinc computers.

I have some credits now lol

I don't do virtual boxes, I'll just keep a hammer nearby to smash anything that goes weird.

Hiya

Excellent - glad you've got some tasks ! I saw that your SRBase account was created some time ago (on 27 Apr 2023), so maybe you set it up when the previous SRBase Sprint took place (27th-30th April 2023)...though perhaps you couldn't get any Linux tasks back then?

Looking back through the posts on this thread, you had some issues getting SRBase tasks, maybe due to an invalid site certificate?

see the following:

viewtopic.php?p=134811#p134811

I can't see the linux post you are talking about - its probably staring at me!

I edited the link to the message board, as the issue relates to how Linux suspends SRBase tasks (and it implies the issue might be within the Linux "wrapper"). But I guess if you do not suspend tasks, then there's no issue?

Also, the GPU application earns a lot more credits - around 6,500 for crunching the TF (cuda100) tasks, which take about 2,100 seconds on my GTX2080i. The CPU tasks earn a lot less, just 8 credits for the "Sierpinski Base - short v0.28" tasks, which take about 6 minutes each.

regards
Tim

[wbiz]

I shouldn't have any tasks getting suspended, I have a very long time set for task switching because some projects have long checkpoints and it can end up going in a circle of gaining progress and losing it again if tasks start switching,

Well I'm glad the certificate issue resolved itself, I thought it had been a sprint before but didn't get round to checking.

On my CPU's I'm getting about 75 credits an hour per core, I won't be making much difference to the team's outcome on this one. Heavy memory usage as well so its dropping one core per computer.

I'm still resisting the urge to get a GPU up and running.

[UBT - Timbo]

Hiya

Just a quick tip:

Go to your SRBase preferences (on their website) and do NOT allow the "Sierpinski Base - short v0.29" CPU tasks to be downloaded as they earn a PALTRY (FIXED) 8 CREDITS for about 500 seconds of CPU time.

Instead, select the "Sierpinski / Riesel Base v0.29" CPU tasks, as they earn a FIXED 140 credits (per task) and take maybe 5,000 seconds, but they earn (pro-rate) more credits per second. So, nearly double the rate.

EDIT: The Sierpinski / Riesel Base - long v0.28 tasks take another order of magnitude LONGER than the above SR Base v0.29 talks, but they earn more than 10 times the credits (1,700 as opposed to 140).

The GPU tasks earn far more credits - a FIXED 6,500 for about 500 seconds of crunching, so those are ideal if you have a good-ish GPU. Even a cheap and cheerful, older series GPU should be OK, as long as it has the latest drivers installed, as you should still get thousands of credits as opposed to "hundreds" (or just 8 credits) per task.

regards
Tim

[ChelseaOilman]

The GPU tasks earn far more credits - a FIXED 6,500 for about 500 seconds of crunching, so those are ideal if you have a good-ish GPU. Even a cheap and cheerful, older series GPU should be OK, as long as it has the latest drivers installed, as you should still get thousands of credits as opposed to "hundreds" (or just 8 credits) per task.

Yes, the TF tasks on GPU are the way to go on this project. I'm only running tasks on my 4 GPUs, nothing on CPU.

My GPUs are set to use 65% power in MSI Afterburner.

RTX 3070 Ti = 308 seconds
RTX 3070 Ti = 315 seconds
RTX 4070 = 246 seconds
RTX 4070 Ti = 182 seconds

[ChelseaOilman]

And here we go!!!

Project down for maintenance
Please check back in a few hours.

[UBT - Timbo]

And here we go!!!

Project down for maintenance
Please check back in a few hours.

Hiya

Thanks for the heads up on this.

The SRBase admin, rebirther, gave me NO advance warning that this was going to happen !!

So, I'm mightily disappointed that this has occurred...so, I hope it is just a minor unscheduled incident that can be rectified quickly.

His latest post on the message board doesn't mentioned this:

https://srbase.my-firewall.org/sr5/foru ... =1795#9001

So, is this a BOINC Manager error message?

regards
Tim

[ChelseaOilman]

So, is this a BOINC Manager error message?

Maybe it was just a glitch at the site while he was posting new messages. When I refreshed my All Tasks page the message showed and then I went to the Main page and it was there as well. It's gone now so not sure what happened. Maybe he remembered there was a sprint still going on.

[UBT - Timbo]

Maybe it was just a glitch at the site while he was posting new messages. When I refreshed my All Tasks page the message showed and then I went to the Main page and it was there as well. It's gone now so not sure what happened.

Hiya

It could just be a BOINC server message that is auto-generated, when something "happens" on the site. Maybe just a server connection issue or perhaps when one (or more) of the server functions is paused or interrupted?

Maybe he remembered there was a sprint still going on.

Yup, could be - the last I heard from him was on Tuesday this week, and he had no issues about the Sprint being selected...plus he posted on the message board about the Sprint taking place (and that would filter down into active members BOINC Manager "Notices" tab).

Either way, I've put a message on the FB forum, so at least some people will get a heads up about it and can act accordingly.

regards
Tim

[ChelseaOilman]

Haven't heard from Stephen in a while. Looks like he has GPUGrid on auto pilot.
https://stats.free-dc.org/stats.php?pag ... 3&team=672

Maybe he's trying to catch up to me!

[UBT - Timbo]

Haven't heard from Stephen in a while. Looks like he has GPUGrid on auto pilot.
https://stats.free-dc.org/stats.php?pag ... 3&team=672

Maybe he's trying to catch up to me!

Hiya

I've not heard from Stephen for a while now, so I hope he's OK and maybe just taking a leisurely break for a while.

But nice to know that his host(s) is/are still crunching something, rather than being totally inactive.

And he'll take a long time (I guess about 15 months at the current rate) to catch up with your current total on GPUGrid !!

PS: You're just about 25m credits shy of becoming our #1 cruncher on GPUGrid project, as our current #1 member ("GPUGRID Role account") hasn't generated *any* credits to the team for a very, very long time (about 9 years !!).

regards
Tim

[wbiz]

Hiya

Just a quick tip:

Go to your SRBase preferences (on their website) and do NOT allow the "Sierpinski Base - short v0.29" CPU tasks to be downloaded as they earn a PALTRY (FIXED) 8 CREDITS for about 500 seconds of CPU time.

Instead, select the "Sierpinski / Riesel Base v0.29" CPU tasks, as they earn a FIXED 140 credits (per task) and take maybe 5,000 seconds, but they earn (pro-rate) more credits per second. So, nearly double the rate.

EDIT: The Sierpinski / Riesel Base - long v0.28 tasks take another order of magnitude LONGER than the above SR Base v0.29 talks, but they earn more than 10 times the credits (1,700 as opposed to 140).

The GPU tasks earn far more credits - a FIXED 6,500 for about 500 seconds of crunching, so those are ideal if you have a good-ish GPU. Even a cheap and cheerful, older series GPU should be OK, as long as it has the latest drivers installed, as you should still get thousands of credits as opposed to "hundreds" (or just 8 credits) per task.

regards
Tim

Thanks

There isn't a huge difference on my machines between short and normal.

Short tasks last between 200 and 360 seconds, usually 320 to 350 ~= 0.025 credits per second

The normal tasks are about 4200 seconds ~= 0.033 credits per second (30% better)

I filled my cache and had long tasks switched off so can't assess them but will have a play after the sprint ends ready for next time.


EDIT:

Ran out of cache on one computer, not sure what happened there, it must have scheduled all the longer tasks first.

Aborted all the short tasks, and loaded up long3 (6 hours) and long2(16 hours) tasks on both computers.

[ChelseaOilman]

Well we nailed 4th place. Even though FB sprints are open to everyone on the team only 4 of us were running the project during the sprint. Kind of surprising given how good the points are on GPU. I made almost 27 million points in 3 days with just 4 GPUs.

[wbiz]

Well we nailed 4th place. Even though FB sprints are open to everyone on the team only 4 of us were running the project during the sprint. Kind of surprising given how good the points are on GPU. I made almost 27 million points in 3 days with just 4 GPUs.

Yep, you did very well Carl. It was also a very poor turnout for other teams, perhaps the school holiday period didn't help.

It was good to see Mikee throwing in some credits.

Next sprint is 31st August.

[ChelseaOilman]

PS: You're just about 25m credits shy of becoming our #1 cruncher on GPUGrid project, as our current #1 member ("GPUGRID Role account") hasn't generated *any* credits to the team for a very, very long time (about 9 years !!).

Guess I need to take care of that!
I suspect that's why Stephen is running GPUGRID, lots of points!

[UBT - Timbo]

Guess I need to take care of that!
I suspect that's why Stephen is running GPUGRID, lots of points!

Hiya

Stephen has his work cut out for himself though...

He has earned 66,187,500 credits in the last 28 days to increase his total GPUgrid credits total to 207,937,500.

And you are on 1,314,227,961 - so a difference of 1,106,290,461, which at the current rate will take him 64 weeks to reach.

https://stats.free-dc.org/team/ps3/672

And that assumes GPUgrid continues to have available tasks...

I think I'll just sit by the sidewalk and see how this pans out !!

regards
Tim

[wbiz]

10 hours until the next sprint

[UBT - Timbo]

10 hours until the next sprint

Hiya

Getting a little bit impatient are we??

It's all looking good for the randomly selected project for tonight - so as long as no website issues happen (on either/both FB and the project) then all should go well.

regards
Tim

[wbiz]

Getting a little bit impatient are we??

I got the dates wrong and started shutting down projects yesterday, I'm normally quite organised but everything boinc goes bosoms up.

[UBT - Timbo]

Hi all

The project randomly selected for FB Sprint #16 is NumberFields

Good luck everyone !

regards
Tim

[ChelseaOilman]

Seems if I run several projects at the same time the odds of one of them being picked for the sprint are much better.

[UBT - Timbo]

Seems if I run several projects at the same time the odds of one of them being picked for the sprint are much better.

Hiya

Well, the choice of Sprint project is from a small selection, as published here:

https://formula-boinc.org/sprint_v3.py? ... &year=2023

So, the trick (as a participant) is to add long deadline tasks to your cache, from a variety of projects, and to then go offline, allowing your host machine(s) the chance to process as many as possible and once the project has been announced, upload all of the tasks and then only download new tasks from the now selected project.

This is pretty much what TSBT do, across a number of their participating members and various projects. And in truth *ANY* team can do this, if they have sufficient hosts from enough members.

And as of 1pm BST, an hour after this Sprint started, TSBT have already earned 5,766,675 BOINC credits, with the 2nd placed team only earning 38,985 credits

https://formula-boinc.org/sprint_v3.py? ... &year=2023

regards
Tim

[ChelseaOilman]

I have no desire to bunker projects. Frankly it's a big turn off from these sprints. Makes me not even want to participate.

[UBT - Timbo]

I have no desire to bunker projects. Frankly it's a big turn off from these sprints. Makes me not even want to participate.

Hiya

Yup, understood.

Personally, I hate it when people bunker tasks...and one day, I hope I can find a solution to this. I had thought that giving people zero advance warning might be enough to deter people...but the fact is that with so many BOINC projects shuttering, there aren't enough fully active projects, (that can be selected for a Sprint) that are still going AND that have relatively short processing times.

Which means there are fewer Sprints to actually choose from, so hedging ones bets on a number of projects can lead to benefits if a team has enough active hosts available (that allows you to spread your bets across multiple projects).

regards
Tim

[ChelseaOilman]

Personally, I hate it when people bunker tasks...and one day, I hope I can find a solution to this.

This is where PG has an advantage with their challenges, bunkering isn't possible.

[wbiz]

Personally, I hate it when people bunker tasks...and one day, I hope I can find a solution to this.

Winning points for every hour's work would help to curtail the super-bunkering.

If contestants weren't allowed to have anonymous computers then you can obtain the task data and ignore any that were issued before the start of the challenge. Personally I think having computers hidden from view in the challenge is a bit weird, its like a football game with players wearing masks, or more accurately a snooker game where you can only see the pockets.

[wbiz]

For some reason Carl's credits don't seem to be clocking up on the FormulaBoinc sprint?

[UBT - Timbo]

This is where PG has an advantage with their challenges, bunkering isn't possible.

Hiya

True...but PG are the operators /administrators of their own Challenges so they can define exactly when to start counting BOINC credits which can be directly associated with the exact date and time when the specific tasks were downloaded, and no 3rd party (such as FB or BG) has access to this data.

So, FB and BG can only access the starting credits total and to then take a snapshot every hour or so as the total increments.

regards
Tim

[UBT - Timbo]

For some reason Carl's credits don't seem to be clocking up on the FormulaBoinc sprint?

Hiya

I'm think they are:

FB stats currently show UBT has 172,845 BOINC credits.

And DC-stats shows:

https://stats.free-dc.org/team/num/46

Code: Select all

ChelseaOilman	296,010
UBT - wbiz	8,970
UBT - Timbo	4,485
pc1		3,450
Mike		1,035
Gordon Lack	345
(no other members earning credits on this project at present)

So, taking out the lowest credits, (about 18,000), there's still 154,000 credits unaccounted for?

We'll see how it pans out.

regards
Tim

[UBT - Timbo]

Hiya

An hour or so later the FB Sprint stats show UBT has 190,440 credits (as of the time of this posting).

https://formula-boinc.org/sprint_v3.py? ... &year=2023

In the meantime, the DC-stats has not changed

https://stats.free-dc.org/team/num/46

So, it could be an aberration, or that maybe the DC-stats website starts their 24 hour "daily stats" period in a different timezone compared to FB?

regards
Tim

[UBT - Timbo]

Hi again

Meanwhile the TSBT stats on DC-stats are here:

https://stats.free-dc.org/team/num/533

and the member stats are:

Code: Select all

PecosRiverM		3,345,465	
Megacruncher TSBT	2,256,990	
davidBAM		356,730	
Miklos M.		325,335	
SolidAir79		61,410	
Thinboab		1,380
(no other members credited)

Meanwhile on the FB stats, the registered TSBT members have earned:

Code: Select all

PecosRiverM		3,345,465	
Megacruncher		2,669,610
Miklos M.		330,165	
(no other TSBT members registered with FB)

And this ties in pretty well.

So, I don't think the FB totals are wrong, but maybe Carls recent NumberFields credits straddled the time when FB started collecting the Sprint stats?

regards
Tim

[wbiz]

There is something weird. My FB sprint stats are usually ahead or equal to Free-DC yet Carl's are behind?

The other thing is that de-bunkered drops from TSBT usually don't appear in the last-hour, last-3-hours or last-6 hours, I forgot to screen-shot.

[wbiz]

FB appears to have caught up in the end, I just don't follow how it runs ahead of free-dc for some people and behind for others. Overthinking things as usual.

[wbiz]

Tim

FB's sql server has crashed, let's hope a reboot resolves any issues.

[UBT - Timbo]

FB's sql server has crashed, let's hope a reboot resolves any issues.

Hiya

Thanks for the heads up !

Both the forum and the FB stats website are still offline - and sadly I do not have any admin control over the server itself.

So, I can do zero / nada / zilch / bugger all, to bring it back online

And no doubt there'll be some retards who'll complain as soon as the FB forum resumes.

Also: the WU-Prop website (where Sebastien is also admin) also has issues.

https://wuprop.boinc-af.org/

*AND*

There are "500" code errors on the L'alliance Francophone website (also admin by Sebastien)

https://www.boinc-af.org/


Edit (13:24 BST):

I've directly messaged two of my contacts on the L'AF team (via the Numberfields project), as well as putting up messages on both the NumberFields message board and the BOINC Games forum, so hopefully someone can nudge Sebastien and he can resolve whatever issues have occurred on all 3 platforms (FB/WU-Prop and L'AF).

regards
Tim