Info 11th BOINC Pentathlon 2020

[mopardude87]

Immense cooler!
Btw that web page, the sizes in inches & mm don't match, I don't know which one is right but I hope you've looked at its dimensions on the makers site.
e.g newegg says 4.79" is 163mm, but it's actually 122mm, or the other way around 163mm is 6.4"! lol. A big difference!
I would contact them, but I don't have a newegg account.

Oh the size will not be a damper on this rest assured. I almost went aoe but apparently that cooler and the NH-D15 stomp about everything right till like a 380mm aoe or a loop? bit much i would think for something i intend on not overclocking.

 

[Assimilator1]

As long as you haven't got a tight fit with their measurements in inches you'll be fine .
Cool & quiet is good btw!

 

[mopardude87]

As long as you haven't got a tight fit with their measurements in inches you'll be fine .
Cool & quiet is good btw!

My case fits up to a 190mm.

Are you a human?

www.newegg.com

 

[StefanR5R]

Jeeper's bulletin #9:

Marathon: "The teams on the long distance have currently sorted themselves. Even the squabblers of the last few days are largely at peace."

Cross Country: "What is happening here is still inscrutable. The overall standings could be decided here and so it seems to be noticeable that not all teams are showing their true abilities yet. The gaps in the TOP 6 are not yet insurmountable, so the situation changes by the hour."

Overall Standings: "TSBT took the lead again and relegated SG to #2. The Germans would certainly be satisfied with that in the end. For P3D (#3) I am not so sure. The weakness in Cross Country is not new, but the results are still missing to make up for it. Right behind them is TAAT (#4), who will surely score points with the Javelin."

And not least, one of the Three Good Questions: "Who screws up at the Javelin?"

 

[StefanR5R]

The Sprint project was announced:
Ibercivis BOINC
https://boinc.ibercivis.es/ibercivis/

This is a fairly new project (actually, a rebooted project), and a medical one.
Good thing that we already have a TeAm there.

Edit: Post #1 has been updated with some basic technical info. More to follow on the go.

 

[StefanR5R]

Sprint — Ibercivis BOINC:
RAM requirements of this project are very low.
If your computer has not enough RAM for Rosetta@home, you can run this one instead.

Note, only results which are reported after May 16, 00:00 UTC (Friday May 15, 20:00 EDT / 17:00 PDT) will count towards the Pentathlon.​

To add the project to BOINC, copy+paste https://boinc.ibercivis.es/ibercivis/ into the bottom field "Project URL:" of the "Tools/ Add project..." dialogue:

​

Then click [Next >] and enter account details.

After you created your account, visit our TeAm page and click [Join this team].

The application does not support so-called checkpoints. This means, if you suspend computing, or reboot the computer, previously running tasks will restart at 0 % progress; all prior computation is lost.
To avoid progress loss during suspension of computing, use boincmgr's advanced view, "Options" -> "Computing preferences..." -> "Disk and memory" tab, and switch on "✓ Leave non-GPU tasks in memory while suspended".
To avoid progress loss if you shutdown and restart the computer, only suspend or hibernate the computer, if possible.

 

[Fardringle]

Work is VERY intermittent on ibercivis, and there currently aren't any tasks available at all, so it should be an interesting sprint..

 

[StefanR5R]

At this time, just some replica of error returns are available.
It's now in the middle of the night in Spain. Maybe an admin will dish out some new tasks later in the day.

I wonder if there will be an admin on duty over the weekend, with the sprint happening on Saturday – Monday.

 

[StefanR5R]

Sprint — Ibercivis BOINC
Just last night, an application update was published.
(message board, applications)

Apparently, checkpointing has been implemented. I will update post #1 later today.

 

[StefanR5R]

PS,
as usual, the announcement of the last discipline (this time, the Sprint at Ibercivis BOINC) also contains the documentation of project elections/ selections.

At all Pentathlons before, teams could vote for projects in advance, and the organizers would create the schedule based on the most popular projects, with constraints such as using only quorum-1 projects for sprint in mind. — This year, the team voting was omitted, and the organizers chose all projects on their own.

About the project choice
Unlike for the first ten Pentathlons, all projects were chosen by the organizers this time. We tried to balance out many aspects: Research fields, CPU vs GPU, projects with Pentathlon experience, but also new projects.

Rosetta@home for the Marathon was an easy choice. The project would have been chosen again by the teams in the traditional project choice process for sure and occured as the Sprint project for the last two times. In addition, the Pentathlon could also support COVID-19 research this way.

We wanted to include a second project from the biological-medical field, but the selection of suitable projects in this field is small, especially as we wanted to save World Community Grid for next year and only had it in mind as a backup project (that's why we didn't rule out last year's projects from the beginning). First we considered GPUGRID, but their problems with the disk space filling up too quickly worsened. Fortunately, we could finally win the just recently restarted project Ibercivis BOINC for the Sprint.

As we wanted to balance out the NVIDIA-only GPUGRID with an AMD-GPU-friendly project, we initially considered MilkyWay@home (that's why the limit of only one GPU project was dropped). Unfortunately, this project didn't think they are currently in the right shape, and they didn't reply to our inquiry until very late in the process. Instead, we chose Amicable Numbers as the GPU project with the least Pentathlon experience to ensure some variety.

For Javelin Throw, another project with a quorum of 1 had to be chosen. As neither Rosetta@home (Marathon) nor Cosmology@Home (on admin's request) were available, this had to be another math project. As we wanted to choose a CPU-dominated project and some Pentathlon experience is very helpful for the Javelin Throw in order to reduce the possibility of tactical decisions being rendered void by server problems, the only remaining project except last year's projects NFS@Home and yoyo@home was NumberFields@home, even though it is not a CPU-only project anymore.

Finally, the always very popular astrophysics projects should not be neglected. We chose Universe@Home to keep the Pentathlon from becoming too GPU-heavy.

PPS,
Ibercivis BOINC introduced a limit on tasks in progress at some point during the last several hours: The quota is 5 tasks per active logical CPU, while considering at most 64 CPUs per host.¹ (That is, a host with ≥64 active logical CPUs is given at most 320 tasks in progress.)

--------
¹) By now, most of my own computing capacity is in hosts with quite a few more than 64 logical CPUs. :-/
Yet another project which requires workarounds for fast and/or wide hosts.

 

[StefanR5R]

I added more technical info on Ibercivis to post #1 (runtimes, workunit replication, quota on tasks in progress...).

 

[Fardringle]

The devs uploaded new work along with the new Ibercivis application and there are currently about 250,000 tasks available.

Hopefully the runtime estimates on tasks for the new Ibercivis application are wrong, or bunkering for the sprint will be difficult. The previous application ran for about 15-19 hours on my 3900X. The new one is estimating about 53 minutes per task. If that is true, it means the 5 active tasks per cpu makes it necessary to have multiple clients on each PC in order to stock up more than a few hours of work. I'll keep an eye on it and see if the runtime estimates stay low, or start climbing up to times similar to the previous application version.

 

[StefanR5R]

Hopefully the runtime estimates on tasks for the new Ibercivis application are wrong, or bunkering for the sprint will be difficult.

The initial estimate is wrong, and bunkering for the sprint is difficult. (Or inconvenient, at least.)

 

[Fardringle]

I will say however, that the limit of 5 tasks per physical CPU is definitely not functioning right now, so if you want to bunker a big load of work, now is a good time before they actually put that limit in place (and these tasks don't expire until 5/21, so they won't time out before the race ends).

I set BOINC on my 12 core, 24 thread 3900x to ask for 2 days worth of work on Ibercivis, and got 512 tasks. I only should have received 60 if the 5 per core rule was working...

 

[StefanR5R]

I will say however, that the limit of 5 tasks per physical CPU is definitely not functioning right now,

Have they switched it off again? It definitely was enforced at the time of my writing.

.
.
.

Everything is possible. Tomorrow it might become a GPU project, for all we know. /s

 

[biodoc]

I set up Ibercivis on my 3950X several hours ago and could only pull 160 tasks. 158 out 160 are short tasks (~1 hours each confirmed) and 2 tasks are predicted to finish in 12 hours. I'll set up the next bunker soon taking into account those 2 long running tasks (not suspending).

 

[StefanR5R]

Sprint — Ibercivis BOINC
Just last night, an application update was published.
(message board, applications)

Apparently, checkpointing has been implemented. I will update post #1 later today.

This application update is also meant to fix SIGSEGV if the Linux kernel does not allow classic vsyscalls. (source)

 

[StefanR5R]

Hopefully the runtime estimates on tasks for the new Ibercivis application are wrong,

The initial estimate is wrong,

Update to my statement: The initial estimate seems to be reasonably accurate now.

Ibercivis BOINC introduced a limit on tasks in progress at some point during the last several hours: The quota is 5 tasks per active logical CPU, while considering at most 64 CPUs per host.

I will say however, that the limit of 5 tasks per physical CPU is definitely not functioning right now,

Have they switched it off again? It definitely was enforced at the time of my writing.

The quota has been turned off again indeed.

 

[StefanR5R]

Jeeper's bulletin #10:
"The organizers were in a cold sweat to organize a suitable track for the sprint. And when all seemed already lost, it succeeded after all. Let's see how our brave pentathletes master these challenges. Only those who believe in themselves and for whom Mount Everest is only a small hill will be successful. And whoever thinks that a small climb is already the Matterhorn will fail."

 

[TennesseeTony]

Fellas we have decided to participate in the 4th Javelin throw for NumberFields, instead of skipping it as originally planned. It starts in 4 hours, and runs for 24 hours after the start.

 

[StefanR5R]

Earlier today I fetched more Ibercivis tasks, now with their updated application version. Initially it looked as if the runtimes of these tasks were more uniform now. But I was mistaken. I've got many which run between 1.5 h...less than 2 h (on my server CPUs with middle-of-the-road processor clocks), but there are also some tasks which set out to take 1.5 days.

Needless to say: Later into the 3-day sprint, task runtimes need to be watched carefully.

 

[StefanR5R]

With pschoefer's announcement from 47 minutes ago, the contest calendar is now complete.

                             Tu We Th Fr Sa Su Mo Tu We Th Fr Sa Su Mo
                    02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
     Rosetta@home   __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄   14 days Marathon
    Universe@home    _ __ __ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄                         5 days City Run
 Amicable Numbers                         _ __ __ ▄▄ ▄▄ ▄▄ ▄▄ ▄▄          5 days Cross Country
  Ibercivis BOINC                                     _ __ __ ▄▄ ▄▄ ▄▄    3 days Sprint
NumberFields@home   __ __ __ ▄▄                                           1 day  Javelin Throw
                                   __ __ __ ▄▄                            1 day
                                               __ __ __ ▄▄                1 day
                                                  __ __ __ ▄▄             1 day
                                                           __ __ __ ▄▄    1 day
                    02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
                             Tu We Th Fr Sa Su Mo Tu We Th Fr Sa Su Mo

 

[TennesseeTony]

Thanks to all who helped with this 4th Javelin Throw at Numberfields. We have met our goal, and now sit in 4th place, 70 points. Two teams appear poised however to run the 5th day, and both may overtake us. There is nothing to be done about this though, as only the middle of the 5 throws gets counted. If you are still running NumberFields, you may now switch to Ibercivis, or Rosetta.

 

[Endgame124]

Thanks to all who helped with this 4th Javelin Throw at Numberfields. We have met our goal, and now sit in 4th place, 70 points. Two teams appear poised however to run the 5th day, and both may overtake us. There is nothing to be done about this though, as only the middle of the 5 throws gets counted. If you are still running NumberFields, you may now switch to Ibercivis, or Rosetta.

Curses!! I literally just set up number fields, as I thought I could assign my r250x to it.

 

[TennesseeTony]

I know how you feel! The Pentathlon can cause hair loss!