FormulaBOINC

[crashtech]

Go, Kiska, Go!!!!

 

[Jondi]

Go, Kiska, Go!!!!

Indeed, I think the rest of you can take 5, he's got this covered.

 

[StefanR5R]

I have a power meter on a power strip which feeds some of my PCs. During usual DC events it hurts to look at the meter because the consumption is so high. Now at the Leiden sprint, it hurts to look at the meter because consumption is so low!

 

[Orange Kid]

Day one and we are in first.
The hard part to this sprint is getting enough work to keep the machines busy, so the most creative team to do that will be the one who win this one.

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

Overclock.net................... 784
2

TeAm AnandTech ............703
3

OcUK - Overclockers UK .673
4

Sicituradastra. ..................434
5

Crunching@EVGA ...........399


Leiden Classical...............Marathon
1 TeAm AnandTech ...........25 .........8,379,382
2 SETI@Netherlands .........18 .........6,804,599
3 XtremeSystems ...............15 .........5,232,765
4 meisterkuehler.de Team ..12 .........3,122,558
5 Overclock.net ...................10 .........1,877,339
6 Brasil [SETIBR] .................8 .............785,608
7 AMD Users ........................6 ............746,093
8 Crunching@EVGA ............4 ............530,689
9 team.se .............................2 ............450,127
10 OcUK - Overclockers UK 1 ............261,645


Leiden Classical...........Sprint.............Day One
1 TeAm AnandTech .............25 ............207,082
2 OcUK - Overclockers UK ..18 ............141,031
3 Overclock.net ....................15 ..............93,949
4 SETI@Netherlands ...........12 ..............42,612
5 Crunching@EVGA ............10 ..............22,333
6 meisterkuehler.de Team .....8 ..............17,601
7 Hungary ..............................6 ...............2,020
8 AMD Users .........................4 ..................797
9 team.se ...............................2 .................739
10 Ars Technica .....................1 .................426

 

[Orange Kid]

Oooops, day two finds us in the number two spot. Guess some are having better luck getting work than others.
Still a close race, so who knows the tide may change.
OcUK has also picked up 4 points in the marathon, so congrats to them on that.

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

Overclock.net ...................784
2

TeAm AnandTech ............696
3

OcUK - Overclockers UK .685
4

Sicituradastra. ..................434
5

Crunching@EVGA ............397

Leiden Classical...................Marathon
1 TeAm AnandTech .............25 ............8,592,649
2 SETI@Netherlands .............18 ...........6,847,088
3 XtremeSystems ...................15 ...........5,232,765
4 meisterkuehler.de Team ......12 ...........3,143,038
5 Overclock.net .......................10 ...........1,949,861
6 Brasil [SETIBR] .....................8 ...............785,756
7 AMD Users ...........................6 ................746,123
8 OcUK - Overclockers UK ......4 ................555,804
9 Crunching@EVGA ................2 ................537,347
10 team.se ...............................1 ................450,792

Leiden Classical...........Sprint.............Day One
1 TeAm AnandTech .............25 ............207,082
2 OcUK - Overclockers UK ..18 ............141,031
3 Overclock.net ....................15 ..............93,949
4 SETI@Netherlands ...........12 ..............42,612
5 Crunching@EVGA ............10 ..............22,333
6 meisterkuehler.de Team .....8 ..............17,601
7 Hungary ..............................6 ...............2,020
8 AMD Users .........................4 ..................797
9 team.se ...............................2 .................739
10 Ars Technica .....................1 .................426

Leiden Classical...........Sprint.............Day Two
1 OcUK - Overclockers UK ....25 ...........403,313
2 TeAm AnandTech ...............18 ...........394,890
3 Overclock.net .......................15 ..........155,909
4 SETI@Netherlands ..............12 .............78,274
5 meisterkuehler.de Team ......10 .............37,583
6 Crunching@EVGA ................8 ..............26,511
7 Hungary .................................6 ................3,489
8 team.se ..................................4 ................1,270
9 Ars Technica ..........................2 ................1,045
10 AMD Users ...........................1 ..................827

 

[StefanR5R]

Well, I am helping out a bit anyway. I got a message from the Leiden server that I though was funny:

9/27/2017 4:53:47 PM | Leiden Classical | (there was work but it was committed to other platforms)

It's not so funny once you've seen it a few times. It means the server has work, but won't give it to you because of your operating system.

The real reason is not your operating system though, but the infamously slow pace of Leiden's work generator or scheduler.

In fact, it's even weirder. Until now I thought "platform" refers to the three (four?) application types which Leiden has: Windows x86-32, Linux x86-32, Linux x86-64. But...

Looks like they are matching up computers by OS and processor types... Intel core to Intel core, xeon to xeon, amd to amd... (OCN post)

...indeed; Intel Xeon/ Windows gets an Intel Xeon/ Windows as wingman, Intel Core/ Win gets Intel Core/ Win, AMD Athlon/ Win gets AMD Athlon/ Win, and so on.

(Besides, Linux x86-64 WUs are systematically just about 1/3 the size of Windows x86-32 WUs.)

Furthermore, we see on the server status page that there are generally a few hundred tasks ready to send, and we mostly receive batches of 16 tasks in a single request–response when we are lucky to receive tasks.

So maybe it goes like this:

• Work generator prepares some WUs for AMD Athlon, running Windows.
• Work scheduler waits for some Athlons on Windows to come along and ask for work. Meanwhile, shrugs off other processor & OS combos.
• Once a bunch of work was sent to matching clients, repeat with a different "platform".

 

[crashtech]

That's so weird. I'd read on the Climate Prediction site that they wanted to trim down the amount of available clients due to the fact that different platforms and OSes give slightly different results. That frankly boggles my mind, math is supposed to be math, but the inherent imprecision of floating point can get handled in more than one way, at least that's how I grasp it as a layman.

 

[StefanR5R]

The Leiden Classical application doesn't even give identical results on same hardware. I may have mentioned it before, but: Some time before the sprint, I received the two tasks of a WU on two hosts of mine which have same CPU, RAM, MB, and OS. (Server hardware which was always super reliable in other DC projects and in "real life" projects.) Validation of this WU came out inconclusive, so the project server generated the third WU and sent it to... another host of mine with same CPU, RAM, MB, OS. When this third task was finished, it and one of the earlier two tasks were marked valid, the other invalid.

 

[Ken g6]

It's gonna be down to the wire...
It's a photo finish...

And it's TeAm AnandTech by a nose! No, by a nose hair! Well, something like that.

1 TeAm AnandTech 25 722,915
2 OcUK - Overclockers UK 18 722,846

Congratulations and thanks to everyone involved!

 

[ao_ika_red]

I guess we won by single task. That's photo finish right there.
Congrats TeAm!!!

 

[Orange Kid]

WOW WHAT A FINISH !!!!!!!!!!!!!!!
Good thing this wasn't an US election, a recount would already be happening.
Congrats on a stellar effort, TeAm. Who'd have ever imagined a finish that close.
Congrats to OcUK for coming from off the grid to 6th place in the marathon and gaining eight points, well played
I'm just going to sit and stare at these numbers for a while

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

Overclock.net ...................779
2

TeAm AnandTech ............703
3

OcUK - Overclockers UK .680
4

Sicituradastra. ..................432
5

Crunching@EVGA ...........403

Leiden Classical..............Marathon
1 TeAm AnandTech ................25 ............8,882,550
2 SETI@Netherlands ................18 ...........6,880,064
3 XtremeSystems ......................15 ...........5,232,765
4 meisterkuehler.de Team .........12 ...........3,165,052
5 Overclock.net ..........................10 ...........2,004,545
6 OcUK - Overclockers UK ..........8 ..............832,584
7 Brasil [SETIBR] ........................6 ...............786,322
8 AMD Users ...............................4 ...............746,237
9 Crunching@EVGA ...................2 ...............541,940
10 team.se ...................................1 ...............450,792


Leiden Classical...........Sprint.............Day One
1 TeAm AnandTech .............25 ............207,082
2 OcUK - Overclockers UK ..18 ............141,031
3 Overclock.net ....................15 ..............93,949
4 SETI@Netherlands ...........12 ..............42,612
5 Crunching@EVGA ............10 ..............22,333
6 meisterkuehler.de Team .....8 ..............17,601
7 Hungary ..............................6 ...............2,020
8 AMD Users .........................4 ..................797
9 team.se ...............................2 .................739
10 Ars Technica .....................1 .................426

Leiden Classical...........Sprint.............Day Two
1 OcUK - Overclockers UK ....25 ...........403,313
2 TeAm AnandTech ...............18 ...........394,890
3 Overclock.net .......................15 ..........155,909
4 SETI@Netherlands ..............12 .............78,274
5 meisterkuehler.de Team ......10 .............37,583
6 Crunching@EVGA ................8 ..............26,511
7 Hungary .................................6 ................3,489
8 team.se ..................................4 ................1,270
9 Ars Technica ..........................2 ................1,045
10 AMD Users ...........................1 ..................827

Leiden Classical...........Sprint.............Final
1 TeAm AnandTech ................25 .............722,915
2 OcUK - Overclockers UK .....18 .............722,846
3 Overclock.net .......................15 .............213,777
4 SETI@Netherlands ..............12 .............115,411
5 meisterkuehler.de Team ......10 ..............62,278
6 Crunching@EVGA ................8 ...............31,555
7 Hungary ................................6 .................5,367
8 team.se .................................4 .................1,282
9 Ars Technica .........................2 .................1,182
10 AMD Users ..........................1 .................1,096

 

[GLeeM]

I didn't join in until half way through Saturday!

A win by 69 points!?! That's exactly how many points geecee had yesterday and today.

 

[StefanR5R]

GLeeM and geecee know how to keep us on the edge of our seats!

 

[TennesseeTony]

The next Sprint is announced in only 12 hours, 1500UTC.

 

[Ken g6]

Oh, no! Looks like someone forgot to select a project again.

 

[Orange Kid]

Ooooops, no new project was picked in a timely manner so the Formula Boinc site broke.
We'll just have to keep checking.

 

[ao_ika_red]

What's wrong with the FB page?

 

[Jondi]

Oh, no! Looks like someone forgot to select a project again.

Ooooops, no new project was picked in a timely manner so the Formula Boinc site broke.
We'll just have to keep checking.

How do you guys know that's the reason?
I know nothing about code.

 

[Ken g6]

I'd suggest somebody post this on the FB forum:

Except that's a little too much time to give them.

How do you guys know that's the reason?
I know nothing about code.

I'm pretty sure they failed to set the projet_sprint variable, leading to "TypeError: 'NoneType' object has no attribute '__getitem__'" when they try to access it.

Edit: P.S. It's "projet", not "project", because it's French.

 

[Ken g6]

The project finally turned out to be GPU Grid. They're out of work, from what I can tell.

 

[Orange Kid]

The projet is GPU Grid

 

[Orange Kid]

WTH. We seen to be posting at the same time today.
I dont think you can bunker this as it is do one get one if I remember correctly.

 

[Kiska]

Welp!
Guess I won't be participating at all this sprint

 

[Ken g6]

Last time I tried this project I got two WUs. I wasn't sure if I could complete them in 24 hours, which gets the best credit, so I aborted one.

 

[StefanR5R]

Obvious similarities to Leiden Classical:

• Not a number theory project, whoohoo!
• Task available only now and then.

 

[crashtech]

Nvidia only, and only 2 tasks at a time, right?