FormulaBOINC

[TennesseeTony]

Good one Bill.

Yeah, I sorta have a steady diet of GPU tasks now.....and re-enabled GPU on one machine (bunkering mishap). (face-palm) Not sure I can make up the incredible 12M lead of EVGA (totally sure, actually, that I cannot). Great luck, and good job, EVGA!

 

[ao_ika_red]

@ao_ika_red, credits per task are fixed, but vary between WU types. My last 20 tasks got 77,250 (1x), 137,850 (6x), 157,200 (2x), 252,750 (8x), or 277,950 (3x) credits. IOW credits are mostly at the order of 0.15...0.25 M per task. When OcUK are for example 1.5 M behind us, they just need six tasks to catch up. (Go OcUK go! ) On the other hand, tasks take typically 8 or 10 hours on good cards, and this makes hourly updates of the sprint scores a bit jumpy.

Edit: Actually, while baseline credits per task are fixed, the project server gives extra credit if the result is reported within 48 hours, and even more if within 24 hours.

Thanks a lot for your lengthy explanation, Stefan. Actually this is all my fault in reading OrangeKid's post. I've mistaken our overall fb points as our sprint credit. Really sorry about that.

 

[Orange Kid]

Hanging on to second. With no bunkers to drop at the end, this one will come down to who gets those last WU's in under the wire. Hoping it won't be as close as the last one.
As a side note, hope all that are in Nates path keep safe.

Formula Boic......................Top Five
1

Overclock.net ....................788
2

TeAm AnandTech .............715
3

OcUK - Overclockers UK ..698
4

Sicituradastra. ...................440
5

Crunching@EVGA .............431

GPUGrid....................Marathon
1 Sicituradastra. ...................25 ...........3,176,550,925
2 XtremeSystems .................18 ...........3,007,230,750
3 Ars Technica ......................15 ...........1,525,676,250
4 France .................................12 .............929,196,250
5 LinusTechTips_Team .........10 ..............903,960,400
6 TeAm AnandTech .................8 ..............815,653,475
7 Crunching@EVGA ................6 ..............600,916,625
8 Overclock.net ........................4 ..............598,061,600
9 team.se ..................................2 .............490,085,975
10 US NAVY .............................1 .............397,868,900

GPUGrid..................Sprint...............Day One
1 Crunching@EVGA .............25 .............22,432,425
2 TeAm AnandTech ..............18 .............13,784,125
3 OcUK - Overclockers UK ....15 .............12,119,000
4 XtremeSystems ..................12 ...............9,902,275
5 Sicituradastra. .....................10 ...............9,135,675
6 Overclock.net ........................8 ...............7,453,675
7 Team Norway ........................6 ...............3,297,375
8 meisterkuehler.de Team ........4 ...............3,110,250
9 France ....................................2 ...............2,934,875
10 Ars Technica ........................1 ...............2,674,875

GPUGrid...................Sprint............Day Two
1 Crunching@EVGA ................25 ............46,550,850
2 TeAm AnandTech .................18 ............27,116,425
3 OcUK - Overclockers UK .......15 ...........25,391,250
4 XtremeSystems ......................12 ...........20,015,675
5 Overclock.net .........................10 ...........18,055,000
6 Sicituradastra. .........................8 ............16,359,400
7 meisterkuehler.de Team .........6 ..............6,508,725
8 France .....................................4 ..............6,212,325
9 Team Norway ..........................2 ..............5,572,125
10 Ars Technica .........................1 ...............5,355,775

 

[Orange Kid]

Are you kidding me!!!!!!!!!!!!! I'm an old guy despite my nick on here and you people are gonna do me in
Another squeeker....only a couple of large tasks ahead of OcUK for a second spot.
Congrats to all.
Congrats to EVGA for kicking all our arses.


Formula Boinc...................Top Five
1

Overclock.net .....................790
2

TeAm AnandTech ..............718
3

OcUK - Overclockers UK ...698
4

Sicituradastra. ....................437
5

Crunching@EVGA .............433

GPUGrid.....................Marathon
1 Sicituradastra. ....................25 .........3,184,207,400
2 XtremeSystems .................18 ..........3,015,325,900
3 Ars Technica ......................15 .........1,528,539,375
4 France ................................12 ............932,675,150
5 LinusTechTips_Team .........10 ...........905,198,850
6 TeAm AnandTech ...............8 ............830,962,925
7 Crunching@EVGA ...............6 ............624,898,125
8 Overclock.net .......................4 ............608,331,800
9 team.se ................................2 ............491,375,100
10 US NAVY ...........................1 ............398,708,750


GPUGrid..................Sprint...............Day One
1 Crunching@EVGA .............25 .............22,432,425
2 TeAm AnandTech ..............18 .............13,784,125
3 OcUK - Overclockers UK ....15 .............12,119,000
4 XtremeSystems ..................12 ...............9,902,275
5 Sicituradastra. .....................10 ...............9,135,675
6 Overclock.net ........................8 ...............7,453,675
7 Team Norway ........................6 ...............3,297,375
8 meisterkuehler.de Team ........4 ...............3,110,250
9 France ....................................2 ...............2,934,875
10 Ars Technica ........................1 ...............2,674,875

GPUGrid...................Sprint............Day Two
1 Crunching@EVGA ................25 ............46,550,850
2 TeAm AnandTech .................18 ............27,116,425
3 OcUK - Overclockers UK .......15 ...........25,391,250
4 XtremeSystems ......................12 ...........20,015,675
5 Overclock.net .........................10 ...........18,055,000
6 Sicituradastra. .........................8 ............16,359,400
7 meisterkuehler.de Team .........6 ..............6,508,725
8 France .....................................4 ..............6,212,325
9 Team Norway ..........................2 ..............5,572,125
10 Ars Technica .........................1 ...............5,355,775

GPUGrid................Sprint...............Final
1 Crunching@EVGA ................25 .............73,329,550
2 TeAm AnandTech ................18 .............41,546,550
3 OcUK - Overclockers UK ......15 .............41,032,600
4 Overclock.net ........................12 .............28,176,625
5 XtremeSystems .....................10 .............27,856,700
6 Sicituradastra. .........................8 ..............24,572,350
7 meisterkuehler.de Team .........6 ..............12,633,400
8 France .....................................4 ...............8,934,950
9 Team Norway ..........................2 ...............8,593,825
10 Ars Technica ..........................1 ...............7,820,625

 

[TennesseeTony]

Thanks for the stats-keeping OrangeKid, your job is much harder than mine (weekly stats)!

I wish the tasks were more plentiful, but well done to all teams!

EVGA came in 3rd across all leagues, so high praises to them and their electric companies.
The TeAm came in 6th across all leagues,
and 7th belongs to our league 3 friends at [H]ard|OCP,
OcUK at the 8th place position,
OCN 11th.

 

[bill1024]

Thanks for the stats-keeping OrangeKid, your job is much harder than mine (weekly stats)!

I wish the tasks were more plentiful, but well done to all teams!

EVGA came in 3rd across all leagues, so high praises to them and their electric companies.
The TeAm came in 6th across all leagues,
and 7th belongs to our league 3 friends at [H]ard|OCP,
OcUK at the 8th place position,
OCN 11th.

Everyone who folds or crunches is #1 in my book.
Congratulations TAAT

 

[Smoke]

Great job, Orange Kid! I really enjoy keeping up with all the stats and your reports.

 

[StefanR5R]

Here is a look back at this year's overall Formula Boinc points in our league, thanks to https://web.archive.org/web/*/http://formula-boinc.org/.

Spoiler: February 1, 2017

1   Overclock.net            414   286 %
2   Sicituradastra.          334   230 %
3   Ars Technica             262   181 %
4   OcUK - Overclockers UK   258   178 %
5   meisterkuehler.de Team   228   157 %
6   Team Norway              220   152 %
7   AMD Users                215   148 %
8   The Scottish Boinc Team  155   107 %
9   TeAm AnandTech           145   100 %   <--
10  XtremeSystems            125    86 %

Spoiler: May 10, 2017

1   Overclock.net            551   183 %
2   OcUK - Overclockers UK   479   159 %
3   Sicituradastra.          435   145 %
4   meisterkuehler.de Team   369   123 %
5   TeAm AnandTech           301   100 %   <--
6   Ars Technica             278    92 %
7   AMD Users                252    84 %
8   Team Norway              230    76 %
9   XtremeSystems            200    66 %
10  Crunching@EVGA           195    65 %

Spoiler: June 11, 2017

1   Overclock.net            639   169 %
2   OcUK - Overclockers UK   539   142 %
3   Sicituradastra.          409   108 %
4   TeAm AnandTech           379   100 %   <--
5   meisterkuehler.de Team   374    99 %
6   Ars Technica             265    70 %
7   AMD Users                263    69 %
8   Crunching@EVGA           245    65 %
9   Team Norway              238    63 %
10  XtremeSystems            212    56 %

Spoiler: September 22, 2017

1   Overclock.net            764   114 %
2   TeAm AnandTech           672   100 %   <--
3   OcUK - Overclockers UK   657    98 %
4   Sicituradastra.          436    65 %
5   Crunching@EVGA           392    58 %
6   meisterkuehler.de Team   382    57 %
7   Ars Technica             323    48 %
8   AMD Users                259    39 %
9   Team Norway              221    33 %
10  The Scottish Boinc Team  192    29 %

Spoiler: October 14, 2017

1   Overclock.net            796   113 %
2   OcUK - Overclockers UK   714   101 %
3   TeAm AnandTech           707   100 %   <--
4   Sicituradastra.          432    61 %
5   Crunching@EVGA           423    60 %
6   meisterkuehler.de Team   397    56 %
7   Ars Technica             328    46 %
8   AMD Users                263    37 %
9   Team Norway              232    33 %
10  XtremeSystems            187    26 %

 

[StefanR5R]

Another look at current Formula Boinc standings:
League 2, points split by those earned in sprints, and those earned in marathons.

Pos. Team                      Sprints        Marathons       overall
------------------------------------------------------------------------
1    Overclock.net            239  ( 69 %)   557  (155 %)   796  (113 %)
2    OcUK - Overclockers UK   256  ( 74 %)   458  (128 %)   714  (101 %)
3    TeAm AnandTech           348  (100 %)   359  (100 %)   707  (100 %)
4    Sicituradastra.           70  ( 20 %)   360  (100 %)   430  ( 61 %)
5    Crunching@EVGA           215  ( 62 %)   206  ( 57 %)   421  ( 60 %)
6    meisterkuehler.de Team   117  ( 34 %)   280  ( 78 %)   397  ( 56 %)
7    Ars Technica              33  (  9 %)   295  ( 82 %)   328  ( 46 %)
8    AMD Users                 22  (  6 %)   241  ( 67 %)   263  ( 37 %)
9    Team Norway               44  ( 13 %)   188  ( 52 %)   232  ( 33 %)
10   XtremeSystems             28  (  8 %)   159  ( 44 %)   187  ( 26 %)
11   The Scottish Boinc Team   28  (  8 %)   154  ( 43 %)   182  ( 26 %)
12   US NAVY                   41  ( 12 %)   114  ( 32 %)   155  ( 22 %)
13   BOINC.SK                  26  (  7 %)    80  ( 22 %)   106  ( 15 %)
14   SETI@Netherlands          32  (  9 %)    73  ( 20 %)   105  ( 15 %)
15   Portugal@Home              7  (  2 %)    95  ( 26 %)   102  ( 14 %)
16   LinusTechTips_Team        20  (  6 %)    76  ( 21 %)    96  ( 14 %)
17   team.se                   18  (  5 %)    50  ( 14 %)    68  ( 10 %)
18   IBM                       27  (  8 %)    34  (  9 %)    61  (  9 %)
19   France                    19  (  5 %)    40  ( 11 %)    59  (  8 %)
20   SwissTeam.NET              0  (  0 %)    58  ( 16 %)    58  (  8 %)
21   Boinc UK                   0  (  0 %)    49  ( 14 %)    49  (  7 %)
22   U.S.Air Force              8  (  2 %)    26  (  7 %)    34  (  5 %)
23   UserFriendly.org          10  (  3 %)    15  (  4 %)    25  (  4 %)
24   Hungary                    6  (  2 %)    18  (  5 %)    24  (  3 %)
25   CANAL@Boinc                0  (  0 %)    19  (  5 %)    19  (  3 %)
26   Brasil [SETIBR]            0  (  0 %)    10  (  3 %)    10  (  1 %)
27   BOINC RUSSIA               1  (  0 %)     6  (  2 %)     7  (  1 %)

Clearly, just 5 out of the 27 teams actually took notice of Formula Boinc sprints.

Crunching@EVGA and TeAm AnandTech got about as many sprint points as marathon points (EVGA a bit more in sprints, the TeAm a bit more in marathons). All other teams got much more marathon points than sprint points.

 

[Orange Kid]

Very nice look at the FB stats.
Appreciate the time it took.

 

[StefanR5R]

And here is what the teams of League 1 have been up to so far:

Spoiler: sprint points + marathon points

Pos. Team                 Sprints  Marathons  overall
------------------------------------------------------
1    Gridcoin                346       759     1105
2    SETI.USA                250       475      725
3    Czech National Team     248       406      654
4    L'Alliance Francophone  170       452      622
5    SETI.Germany             88       403      491
6    UK BOINC Team           120       147      267
7    BOINC@Poland             29       220      249
8    Planet 3DNow!            33       205      238
9    USA                      86       121      207
10   Rechenkraft.net          65        96      161
11   Team China               47        81      128
12   Dutch Power Cows          4       107      111
13   BOINC@AUSTRALIA          14        92      106
14   Team 2ch                 34        59       93
15   BOINC.Italy              12        76       88
16   Boinc@Taiwan              5        77       82
17   Russia                   10        47       57
18   The Knights Who Say Ni!   8        34       42
19   Russia Team               3        35       38
20   The Planetary Society    16        20       36
21   Ukraine                  18        18       36
22   SETIKAH@KOREA             2        32       34
23   Canada                    0        25       25
24   BOINCstats                2        18       20
25   BOINC Synergy             6        14       20

I have difficulties to tell which teams ran sprints actively, and which ones scored high in sprints without taking the slightest notice of the event.

As for Gridcoin, they have at least one user who knew of sprints and took part in it. But I am not so sure of his team mates.

I suppose some SETI.USA users are active in sprints. CNT have an active thread about Formula Boinc sprints in their forum. So do AF. UK Boinc team have active threads for each individual sprint. Last time I checked, there was no active thread about Formula Boinc in SETI.Germany's forum.

 

[GLeeM]

Thank you StefanR5R for the wealth of information

 

[ao_ika_red]

I have difficulties to tell which teams ran sprints actively, and which ones scored high in sprints without taking the slightest notice of the event.

As for Gridcoin, they have at least one user who knew of sprints and took part in it. But I am not so sure of his team mates.

I suppose some SETI.USA users are active in sprints. CNT have an active thread about Formula Boinc sprints in their forum. So do AF. UK Boinc team have active threads for each individual sprint. Last time I checked, there was no active thread about Formula Boinc in SETI.Germany's forum.

If there's nothing explosive from Sicituradastra in near future, we'll be guaranteed for entering League 1 next season. And based on your data, we'll have a chance to get into midfield battle because most of League 1 teams are not as interested as us in FB sprint (ceteris paribus).

 

[ao_ika_red]

It's TN-Grid.

 

[Orange Kid]

hmmmmmmmmm......CPU only

 

[ao_ika_red]

hmmmmmmmmm......CPU only

Does it need extra script? Because I believe it has separate AVX and SSE2 app.

 

[crashtech]

Hmm, what to do.

 

[StefanR5R]

Does it need extra script? Because I believe it has separate AVX and SSE2 app.

Projects with multiple application versions will successively send each version which the host supports. Performance of each version is tracked, and eventually you will receive the best performing one almost exclusively. You can check the outcome of these performance measurements locally in client_state.xml (or some other file?), or on the project web site, host details page, "application details" link.

This automated procedure can be circumvented by a custom app_info.xml file. That way you would define what application your host "supports". But writing such a file is a bit involved.

 

[ao_ika_red]

Thanks Stefan. Another thing to learn.
edit:
My Athlon only get 1 SSE2 task, remaining tasks will run on AVX. It's very good.

 

[ao_ika_red]

A bit off FB-Sprint topic.
On Milkyway@Home I've got tens of "cancelled by server" tasks last night, and then I found this on mw@h forum:

Hey All,

We will have to clear the workunit queue for both separation and nbody. This will cause many invalid workunits. We apologize for this, as this is the only way to clear the system for it to return to normal operation.

Thanks for your continued support,
Jake and Sidd

 

[iwajabitw]

A bit off FB-Sprint topic.
On Milkyway@Home I've got tens of "cancelled by server" tasks last night, and then I found this on mw@h forum:

Hey All,

We will have to clear the workunit queue for both separation and nbody. This will cause many invalid workunits. We apologize for this, as this is the only way to clear the system for it to return to normal operation.

Thanks for your continued support,
Jake and Sidd

They were having a bunch of server issues and had to take it off line the other day. Jake was trying not to mess up the tasks, looks like it didn't go a planned.

 

[Jondi]

Does it need extra script? Because I believe it has separate AVX and SSE2 app.

I wouldn't bother, I'm getting AVX SSE2 & FMA .
On my Haswell machines there's nothing in it between the three, runtimes are all about the same.

 

[StefanR5R]

Formula Boinc sprints are popular. Far too popular, relative to the throughput of Leiden's, GPUGrid's, and TN-Grid's work generators.

 

[ao_ika_red]

They were having a bunch of server issues and had to take it off line the other day. Jake was trying not to mess up the tasks, looks like it didn't go a planned.

No wonder a bunch of my completed tasks reported as error this week.

I wouldn't bother, I'm getting AVX SSE2 & FMA .
On my Haswell machines there's nothing in it between the three, runtimes are all about the same.

Any improvement is a good improvement for my little machine. I guess your machine is using Linux because there's no FMA apps listed for Windows.

I've run TN Grid overnight, I want to report how it behaves on my Athlon.
* 1 task per module, AVX tasks need 1h42min up to 1h44min to run.
* 2 task per module, AVX tasks need 1h48min up to 1h50min to run.
* 1 task per module, SSE2 task needs 1h52min to run (I only got 1 SSE2 task so the margin of error is pretty high with this one).

 

[TennesseeTony]

That's not terrible, Red, only slightly slower than some more modern hardware. //thumbsup//