PrimeGrid Races 2018

[Ken g6]

Day 2 stats:

Rank___Credits____Username
1______2250013____xii5ku
7______1004037____emoga
14_____692997_____crashtech
36_____277952_____Tejas
40_____259584_____biodoc
54_____195030___10esseeTony
61_____186354_____Ken_g6
70_____165233_____Howdy2u2
94_____113097_____Orange Kid
108____101786_____markfw
126____76020______zzuupp
170____45616______indydude, Son of crashtech
196____30777______Kiska
241____12806______SlangNRox

Rank__Credits____Team
1_____7228087____Czech National Team
2_____6373480____SETI.Germany
3_____5497489____Aggie The Pew
4_____5411309____TeAm AnandTech
5_____3919648____Sicituradastra.
6_____1519143____BOINC@Poland
7_____1338956____BOINC@MIXI

With graph from SETI Germany:

We're so close to passing the rats (Aggie the Pew). This has to be the most users we've ever had, but can we get a little more...More...MORE?!

 

[Markfw]

If I get my new 2700X done tonight, I will use it as a OC test....

 

[crashtech]

...This has to be the most users we've ever had, but can we get a little more...More...MORE?!

 

[Tejas_2]

Appreciate the stats Ken. Blowing the dust bunnies out of a few old boxen, got 2 up and running hopefully another one or two later this evening.

 

[emoga]

It won't be much, but I'll switch some ryzen rigs over.

 

[crashtech]

I'm trying to awaken a 2P LGA771 board just to see if it's got any fight left in it. All eight cores might equal an overclocked i5-2500.

 

[Markfw]

44 cores going doing ~100k, adding 16 more soon I hope

 

[Markfw]

60 cores going now.

 

[indydude345]

@crashtech Yes, I restarted the client after making the change. this is my app_config:

<app_config>
<app_version>
<app_name>llrSR5</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
</app_config>

 

[TennesseeTony]

Ah RATS. Now I have to contribute more, since everyone else is, in hopes of catching Aggie of course. 120 threads incoming.

 

[zzuupp]

Hey, if we finish this race in 4th, maybe we can be known as 7447?

?? Let's be Boeing ??

Ken, have we ever beaten the rats? Seemingly waaay back when the rats and the cows were the teams to beat. FBOW, the cows aren't around anymore.

 

[Markfw]

120 threads now

 

[crashtech]

@crashtech Yes, I restarted the client after making the change. this is my app_config:

<app_config>
<app_version>
<app_name>llrSR5</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
</app_config>

Probably just turn SMT back on. The reporting of the run times is buggy, you can turn off networking long enough for a few to finish to see how long the WUs really take.

 

[StefanR5R]

So I'm waking up, check the forum, and see that everybody was busy adding more gear and tweaking...

So I added my Haswell laptop now. Didn't use it for this PG challenge until now because it tends to crash a lot. Its i7-4900MQ is specified for 47 W TDP, but the laptop's cooling system can't really sustain that. Therefore I fixed the processor clock to 2.2 GHz with ThrottleStop, which results in 35 W CPU package power.

Also added an IvyBridge-E, 6 cores @4.2 GHz. This pulls four times the power compared to the laptop, but brings only twice the throughput. I am currently running just 1 task at a time on it; need to check if 2 at a time are better. Might be, with all that RAM bandwidth and ample L3 cache.

 

[biodoc]

For my E5-2690 v2 (Ivy Bridge EP; 10 core, 20 threads) single socket and dual socket machines, I've been running them with HT off and 1 x 10 on the single socket and 2 x 10 on the dual socket. Both are on 64 bit linux.

For an average of 13 tasks, I see 3244 seconds per task which is 43.7K PPD on the single socket and double that on the dual socket.

I switched the single socket to 2 x 5 and am seeing task times of 5467 seconds (average of 20) which is 51.9K ppd. I'll switch the dual socket to 4x5 now.

 

[biodoc]

Convert HH:MM:SS to seconds tool.

 

[StefanR5R]

If I get my new 2700X done tonight, I will use it as a OC test....

Keep an eye on the validations; the LLR applications are sensitive.

--------

I am noticing that there is some variability between the currently available SR5-LLR work units:

Performance monitoring based on run times alone remains imprecise. Here is another distribution of credits from my latest validated tasks, downloaded between 15 June 20:00UTC and 16 June 09:00 UTC:

--------

Switched the IVB-E from 1 task x 6 threads to 2 tasks x 5 threads. While I have too few tasks from before and after (and none validated yet), the rise of CPU package power from 145 W to 155 W, and the progress rate of a task which was 33 % complete when I switched configs, make it seem that throughput increased about proportionally to power consumption.

--------

The TeAm rank is progressing well so far:

 

[Kiska]

It's one way. Though you need to be there to control the supply of tasks.

The other way is to use the "--get_old_tasks" command line in boinccmd
That should generate a list for the last 241 hrs

 

[Kiska]

Sample boinccmd --get_old_tasks:

task llrSR5_299591386_1:
   project URL: http://www.primegrid.com/
   app name: llrSR5
   exit status: 0
   elapsed time: 4649.136311 sec
   completed time: 16-Jun-2018 21:46:03
   reported time: 16-Jun-2018 21:46:03
task llrSR5_299591387_1:
   project URL: http://www.primegrid.com/
   app name: llrSR5
   exit status: 0
   elapsed time: 6042.214588 sec
   completed time: 16-Jun-2018 22:05:24
   reported time: 16-Jun-2018 22:05:24
task llrSR5_299591384_1:
   project URL: http://www.primegrid.com/
   app name: llrSR5
   exit status: 0
   elapsed time: 7662.311695 sec
   completed time: 16-Jun-2018 22:12:12
   reported time: 16-Jun-2018 22:12:12

 

[StefanR5R]

Now that's useful. Would be even more useful if this went further back than just 1 hour.

 

[Markfw]

OK, Ken, where are day 3 stats ??? getting impatient

 

[Ken g6]

OK, Ken, where are day 3 stats ??? getting impatient

Derp! This is a very inconvenient challenge because stats time is also dinnertime.

Day 3 stats:

Rank___Credits____Username
1______3562341____xii5ku
6______1683793____emoga
12_____1242860____crashtech
30_____558559___10esseeTony
33_____506115_____Tejas
38_____469922_____biodoc
50_____344385_____Howdy2u2
54_____329334_____markfw
66_____286622_____Ken_g6
101____174655_____Orange Kid
124____135473_____zzuupp
176____73386______Kiska
212____45616______indydude, Son of crashtech
240____27216______SlangNRox

Rank__Credits____Team
1_____11998330___Czech National Team
2_____10125247___SETI.Germany
3_____9440283____TeAm AnandTech
4_____8755858____Aggie The Pew
5_____6343473____Sicituradastra.
6_____2401445____BOINC@Poland

Wow! Third and potentially rising!

 

[Tejas_2]

Many Thanks to Ken for the stats!

 

[crashtech]

Outstanding, way to go TeAm!

 

[Orange Kid]

Thanks for the stats.