PrimeGrid Challenges 2022

[StefanR5R]

tasks pinned to a core cluster...

I have 832 threads running, mostly optimized for cache and affinity.

Sounds like most on the TeAm are deep in the optimization business now. :-)

I wonder how long a single thread takes on one of these tasks?

Certainly depends on what the other virtual machines on the same cloud computer are doing. Would be rather traumatic for a machine if it had 96 single-vCPU VMs all going for PrimeGrid work at the same time. That'd be a rather unlikely scenario though, unless the provider tends to concentrate a customer's VMs on same nodes. (Not sure if any provider is doing that.)

Edit:
tng's take 51 hours for llrCUL:
http://www.primegrid.com/results.php?hostid=1164053&appid=4
http://www.primegrid.com/results.php?hostid=1164288&appid=4
http://www.primegrid.com/results.php?hostid=1166848&appid=4

 

[StefanR5R]

After a long while I am lurking in the SETI.Germany forum a bit again, and found pschoefer's report on the cleanup stage of several past challenges:
Our TeAm had 4 invalidations in the cleanup of the WW challenge of September, 2 invalidations of the PPS-LLR challenge of July, and 1 invalidation of the GFN-17 challenge of January. Most of the other teams didn't have any invalidations, but then again, we were the team with the most results in each of these challenges.
(source)

 

[waffleironhead]

These units (CUL more so than WOO) are really showing the 16mb cache on my 5700g is very lacking versus the 3700x 32mb.
3700x is running 7.5%/hour on woo vs 5700g getting 4.5%/hour

 

[Ken g6]

Day 3 stats:

Rank___Credits____Username
2______8128303____markfw
3______6752119____xii5ku
5______5843551____crashtech
14_____3055710____parsnip soup in a clay bowl
16_____2853878____w a h
20_____2055074____biodoc
22_____1824693____[TA]Skillz
26_____1456625____cellarnoise2
39_____1156478____mmonnin
43_____993969_____Orange Kid
105____308741_____Ken_g6
111____294289_____Lane42
113____286402_____waffleironhead
173____118023_____Mardis
185____109088_____mnelsonx
260____27584___10esseeTony

Rank__Credits____Team
1_____35264535___TeAm AnandTech
2_____21413832___Czech National Team
3_____14827335___SETI.Germany
4_____14688814___Antarctic Crunchers

Alas, somebody rented enough cores to overthrow Mark. But @TennesseeTony showed up.

 

[cellarnoise]

Day 3 stats:

Rank___Credits____Username
2______8128303____markfw
3______6752119____xii5ku
5______5843551____crashtech
14_____3055710____parsnip soup in a clay bowl
16_____2853878____w a h
20_____2055074____biodoc
22_____1824693____[TA]Skillz
26_____1456625____cellarnoise2
39_____1156478____mmonnin
43_____993969_____Orange Kid
105____308741_____Ken_g6
111____294289_____Lane42
113____286402_____waffleironhead
173____118023_____Mardis
185____109088_____mnelsonx
260____27584___10esseeTony

Rank__Credits____Team
1_____35264535___TeAm AnandTech
2_____21413832___Czech National Team
3_____14827335___SETI.Germany
4_____14688814___Antarctic Crunchers

Alas, somebody rented enough cores to overthrow Mark. But @TennesseeTony showed up.

Probably a @TennesseeTony task unstuck from the mud from a year or two ago? Anyway if it finishes it will be a great year! To have "the Foot" as in BigFoot tony back!!!!

Maybe just a sighting though?

Do these work here. This was the most demeaning? I tried.

 

[Markfw]

Back in first place ! posting now, since I am sure its only for a little bit.

 

[StefanR5R]

Reading up on some other stuff, the server troubles during the short PPS-LLR challenge came to my mind:

We have almost killed the PG server in this challenge.

View attachment 64998

They're going to let it run it course, but once the server reaches a point that it can't accept anymore uploads the challenge will end. Which they think might happen early, but they're hopeful it'll make it to the end.

I am shutting down most of my rigs in hopes we don't kill the server. There is no chance anyone will catch us as a team, even if everyone on the team stopped crunching 100%.

With that said, I've set NNT on a lot of my rigs and shutting them down when they complete what they have.

Back then I didn't entirely understand what the precise bottleneck was, and if/how participants could help to avert it.
Now I found this older explanation of the server-side workload for LLR2-style "main task--proof task" verification, and it becomes clear:
PG thread "LLR2 installed on all big LLR projects", message 143359, message 143362, furthermore message 143777, message 145281.

 

[crashtech]

Looks like it's all up to the server.

 

[Markfw]

Well, thanks to the clouds, I am now 3rd. But I think I have that locked up.

Edit: nope, looks like xi5ku will pass me. More cloud ?

 

[Ken g6]

Day 4 stats, kinda late:

Rank___Credits____Username
3______11467064___markfw
4______9594138____xii5ku
6______7957442____crashtech
11_____5239475____parsnip soup in a clay bowl
17_____4044247____w a h
19_____3601814____[TA]Skillz
22_____2812882____biodoc
25_____2196891____cellarnoise2
35_____1690877____mmonnin
44_____1347298____Orange Kid
107____455223_____Ken_g6
109____443875_____Lane42
113____427068_____waffleironhead
181____163842_____mnelsonx
183____148017_____Mardis
223____84424___10esseeTony

Rank__Credits____Team
1_____51674586___TeAm AnandTech
2_____30257070___Czech National Team
3_____26238927___Antarctic Crunchers
4_____21168485___SETI.Germany

Stats in this challenge will jump around a bit, given that one or another team is also bringing single-threaded cloud VMs again.

Yeah, this morning I felt like I'd been hit over the head with a waffle iron, but now I don't. I imagine I may again soon, though.

 

[Skillz]

Well, thanks to the clouds, I am now 3rd. But I think I have that locked up.

Edit: nope, looks like xi5ku will pass me. More cloud ?

xii5ku doesn't run cloud. Those are all his systems, optimized 150%, greased up daily to get the most that those systems can possibly produce doing the work.

 

[StefanR5R]

xii5ku is trying for sure, but I don't think he's got a chance as long as you take care. :-)



(And yes, these are only my home computers.)

 

[Markfw]

As you can see my line is pretty straight. I had a 2 hour outage yesterday while the electrician fixed outlets that had melted wires.

 

[Markfw]

As you can see my line is pretty straight. I had a 2 hour outage yesterday while the electrician fixed outlets that had melted wires.

So, I am going to stop updating "I made 2nd" then "I went down to 3rd", this has happened like 4 times today. Currently 2nd though.

 

[StefanR5R]

But at least the team rank is rock stable. :-D

 

[Skillz]

But at least the team rank is rock stable. :-D

Yep, and poor old me clawing my way up in 14th place.

 

[cellarnoise]

xii5ku is trying for sure, but I don't think he's got a chance as long as you take care. :-)

View attachment 70481

(And yes, these are only my home computers.)

Stefan,

Where or How or did you find or make this chart? You pulling data from the project and creating in your own "Stefan" data tracker? I see that Highcharts.com is used, so you must be doing your own data pulls and stats?

Is it from a site somewhere that is already crunching and tracking these stats?

I only know of this "extra" stats site....

PrimeGrid Challenge Stats

Statistics for PrimeGrid Challenges here. statsgr.id i.e. Stats Grid or statsgrid

statsgr.id

 

[crashtech]

@cellarnoise

PG-Challenge-App (SETI.Germany)

www.seti-germany.de

You have to click on a user, then use the "Vergleich" tab to generate a graph. It's interactive, clicking on the names at the bottom can remove or re-add lines to the graph.

 

[cellarnoise]

@cellarnoise

PG-Challenge-App (SETI.Germany)

www.seti-germany.de

You have to click on a user, then use the "Vergleich" tab to generate a graph. It's interactive, clicking on the names at the bottom can remove or re-add lines to the graph.

Learn something / new tools everyday! Thanks.
Somehow I did not know about this one?

 

[crashtech]

I gave you the wrong link.

PG-Challenge-App (SETI.Germany)

www.seti-germany.de

 

[cellarnoise]

I gave you the wrong link.

PG-Challenge-App (SETI.Germany)

www.seti-germany.de

Figured it out I think. After I got past a lot of hardware pron? . Or maybe I first clicked on the wrong / old browser tab.... Again?

 

[Ken g6]

Day 5 stats:

Rank___Credits____Username
2______14463944___markfw
4______12817204___xii5ku
6______9949760____crashtech
10_____7266981____parsnip soup in a clay bowl
12_____6810678____[TA]Skillz
16_____5420304____w a h
22_____3491880____biodoc
25_____2783249____cellarnoise2
34_____2177659____mmonnin
45_____1738979____Orange Kid
108____572736_____Lane42
109____569886_____Ken_g6
110____566151_____waffleironhead
122____506562___10esseeTony
187____218620_____mnelsonx
197____178915_____Mardis
268____59012______SlangNRox

Rank__Credits____Team
1_____69592527___TeAm AnandTech
2_____38430572___Czech National Team
3_____32517343___Antarctic Crunchers
4_____27107510___SETI.Germany

I don't know about you, but I'm feeling boxed-in my Lane.

 

[Orange Kid]

Thanks for the stats and voting.

 

[StefanR5R]

I tested the leading edge of both on a Haswell CPU.

llrCUL: "zero-padded FMA3 FFT length 2400K" -> 18.75 MB FFT data size
llrWOO: "zero-padded FMA3 FFT length 2304K" -> 18.0 MB FFT data size​

FFT lengths have now advanced to 2560K at llrCUL (20 MBytes) and 2400K at llrWOO (18.75 MBytes).

 

[Markfw]

FFT lengths have now advanced to 2560K at llrCUL (20 MBytes) and 2400K at llrWOO (18.75 MBytes).

what will that do to things ?