PrimeGrid Races 2018

[ao_ika_red]

Finally, my first ever completed SoB task.

Spoiler: Task 866759723

If there's something wrong with it, let me know.

 

[StefanR5R]

I shouldn't watch what my personal rank is doing; it's motion sickness inducing. Most of the time I'm 15th; Ken's last two stats happened to show me at 12th or 13th, and at this very minute I am at 8th. (Sorry @bill1024, I didn't mean to.) I suppose I will be back down at 12th or 15th or something soon enough.

--------
On an unrelated note, it just occurred to me that it would have been fitting in a way if PG had scheduled the "SoB" subproject for their "year of the dog" challenge.

 

[ao_ika_red]

On an unrelated note, it just occurred to me that it would have been fitting in a way if PG had scheduled the "SoB" subproject for their "year of the dog" challenge.

That's a bit unfair to Chinese people, isn't it? But I get your point. This sub-project named as SoB for a reason.

 

[bill1024]

I shouldn't watch what my personal rank is doing; it's motion sickness inducing. Most of the time I'm 15th; Ken's last two stats happened to show me at 12th or 13th, and at this very minute I am at 8th. (Sorry @bill1024, I didn't mean to.) I suppose I will be back down at 12th or 15th or something soon enough.

--------
On an unrelated note, it just occurred to me that it would have been fitting in a way if PG had scheduled the "SoB" subproject for their "year of the dog" challenge.

I see what you did there lol nice /\5 nice!
Really, my ears are still ringing. 3rd row stage left, right in front of the massive stack.

I do have the bomber that will make a run, looks like 5 hours before the end.
At least I hope it is before the end, it's close

 

[bill1024]

That's a bit unfair to Chinese people, isn't it? But I get your point. This sub-project named as SoB for a reason.

"Year of the dog" a play on "hair of the dog."
Now your messing with a SOB.......Yes?

 

[Ken g6]

Day 5 stats, 1/3 done!

Rank___Credits____Username
11_____2764973____xii5ku
21_____1657109____crashtech
94_____304050_____Ken_g6
124____228154_____biodoc
154____171340_____zzuupp
159____154263_____Howdy2u2
226____57993______Kiska
233____50516______Orange Kid
264____23877.18___Marcellius Ika Permadi

Rank__Credits____Team
3_____18762755___Czech National Team
4_____18010411___SETI.Germany
5_____9626420____Crunching@EVGA
6_____5412278____TeAm AnandTech
7_____4230178____BOINC@MIXI
8_____3684603____BOINC@Poland
9_____3673622____Overclock.net

 

[Orange Kid]

Thanks for the stats.

 

[ao_ika_red]

"Year of the dog" a play on "hair of the dog."
Now your messing with a SOB.......Yes?

I don't know if such expression is existed. Coincidentally, The YoTD challenge happens at the end of 1 month long of Chinese New Year celebration and this year 's zodiac animal is a dog.

I just want to have a bit of joke about SoB because it puts plenty of stress to my laptop but if that puts you and other seasoned PG cruncher off, I'm really sorry.

Thanks for the stats, Ken.

 

[bill1024]

I don't know if such expression is existed. Coincidentally, The YoTD challenge happens at the end of 1 month long of Chinese New Year celebration and this year 's zodiac animal is a dog.

I just want to have a bit of joke about SoB because it puts plenty of stress to my laptop but if that puts you and other seasoned PG cruncher off, I'm really sorry.

Thanks for the stats, Ken.

I did not know it was the year of the dog.
Hair of the dog means to take a drink of the booze you drank last night that gave you the hangover today.
The other hair of the dog is a song.

 

[Howdy]

MOARRRRRRRRRRRR COW BELLLLLLLLLLLLLLL!!!!!!!!

 

[Ken g6]

Also moar stats!

Rank___Credits____Username
8______3415723____xii5ku
23_____2025061____crashtech
98_____367385_____Ken_g6
131____283078_____biodoc
155____216166_____Howdy2u2
163____202659_____zzuupp
220____92678______SlangNRox
252____57993______Kiska
256____50516______Orange Kid
295_______Marcellius Ika Permadi TeAm AnandTech

Rank__Credits____Team
3_____23787854___Czech National Team
4_____22412108___SETI.Germany
5_____11735173___Crunching@EVGA
6_____6735141____TeAm AnandTech
7_____5231275____BOINC@MIXI
8_____5038690____BOINC@Poland
9_____4322887____Overclock.net

I'm having trouble with Marcellius Ika Permadi's stats. His name is too long!

 

[ao_ika_red]

I'm having trouble with Marcellius Ika Permadi's stats. His name is too long!

Fixed, Ken. Your yesterday's stats seemed fine with it, though.

 

[StefanR5R]

As always, this mexxus guy of Boinc@Australia is employing a horde of cloud computers (presumably gratis trial instances), each one single-threaded, and in case of llrSOB, each one apparently capable to complete only one WU before the trial runs out. Everybody has their own unique way to have fun at a math race, I guess.

 

[StefanR5R]

I do have the bomber that will make a run, looks like 5 hours before the end.
At least I hope it is before the end, it's close

You probably know the following already, but nevertheless: If you want to, you can switch already running tasks over to multithreading anytime. Switch "Leave non-GPU tasks in memory while suspended" off (at least for the duration of the following steps), suspend the tasks, change app_config.xml, "Read config files", resume the tasks. Boincmgr's display of #CPUs used in the tasks tab may be wrong, but task manager or top or whatever will show that the tasks are indeed running with the chosen thread count. I have done this many times with different LLR subprojects, and at most it cost a tiny fraction of processing time depending on snapshotting granularity, but I never got abortions or invalids from it.

The LLR subprojects with long runtimes have higher throughput when running multithreaded, even though fewer tasks can be run in parallel on the host then. At least that's true on Intel CPUs; I haven't done comparisons on AMD CPUs myself and don't remember having seen rigorous comparisons on AMD CPUs from others. Since multiprocessing vs. multithreading performance is a cache vs. RAM performance thing, AMD hosts will behave somewhat different from Intel hosts in this regard, but the basic principle should be true on AMD hosts as well.

 

[Ken g6]

Week 1 stats:

Rank___Credits____Username
8______4210002____xii5ku
22_____2455204____crashtech
109____431535_____Ken_g6
123____359614_____zzuupp
142____312937_____biodoc
154____280727_____Howdy2u2
233____114275_____Orange Kid
244____92678______SlangNRox
280____57993______Kiska
324____23877______MarcelliusIkaP

Rank__Credits____Team
3_____28642853___Czech National Team
4_____26876262___SETI.Germany
5_____13749305___Crunching@EVGA
6_____8338846____TeAm AnandTech
7_____6122305____BOINC@Poland
8_____5912682____BOINC@MIXI
9_____4583344____Overclock.net

I finally got the name issue sorted.

 

[StefanR5R]

Poland is increasing its new lead over Japan, and latest stats look like our lead over Poland is slowly melting away. Need to keep an eye on these guys.

 

[Ken g6]

I'm having trouble with Marcellius Ika Permadi's stats. His name is too long!

@ao_ika_red (?), when I said this I was kidding. I meant your username was too long for my stats script (which I fixed), not that you needed to change it.

 

[crashtech]

I'm finding that running Primegrid concurrently with F@H adversely affects ppd even on quad-channel systems, if the GPU is larger than a 1060. I'm discontinuing PrimeGrid on another box today. Seeing these problems puts me back in the mindset of having two kinds of boxes, one for GPU and one for CPU, the former being quad cores, and the latter being 2P with no GPU at all. Maybe Linux will help.

 

[StefanR5R]

Perhaps Linux would help a little with that, but probably not a lot. Actually I even suspect it won't help at all. I currently fold only on Linux, hence don't have a direct comparison. I was dissatisfied with how much PG pulled F@H performance down on my Z270 nodes and removed it there early on. I still have PG on the 3-GPU X99 host with E5-2696v4, running 2 tasks with 10 threads each. I.e. I am occupying 20+3 logical CPUs on this 22C/44T Xeon, and the hit to F@H PPD is minimal.

Since 12 hours now, the two CPU hosts which I have at F@H are not receiving work. I switched them to PG when I noticed it. Thanks to @Kiska's tip to use boinc's Options/ Exclusive applications, they will auto-suspend PG as soon as they receive F@H work again.

Edit,
hmm, OK, those 3 GPUs on my X99 are large (1080Ti), so maybe Linux does help a bit with limiting the negatve llrSOB effect after all.

Edit 2,
I became curious whether I would get 0xa4 work on 24-thread F@H slots, so I bogged down one slot. Immediately got a 0xa4 WU... but 0xa7 WUs on the remaining 44-threads slots too. So, these CPUs are off of PG again.

 

[crashtech]

I'll need to learn different ways of looking at GPU utilization in Linux, do you use a command line utility, @StefanR5R ?

 

[StefanR5R]

Yep.

 

[crashtech]

I might want to quote you in the Linux thread. There's so much information an all these threads! many times I find myself doing searches for things vaguely remembered but lost in the shuffle.

 

[Ken g6]

Day 8 stats:

Rank___Credits____Username
7______5065407____xii5ku
23_____2777285____crashtech
107____520940_____Ken_g6
139____383503_____zzuupp
142____372772_____biodoc
160____311267_____Howdy2u2
237____124392_____SlangNRox
250____119082_____Kiska
255____114275_____Orange Kid
353____23877______MarcelliusIkaP

Rank__Credits____Team
3_____33925394___Sicituradastra.
4_____31284441___SETI.Germany
5_____15898063___Crunching@EVGA
6_____9812804____TeAm AnandTech
7_____7459140____BOINC@Poland
8_____6693201____BOINC@MIXI
9_____5874326____BOINC@AUSTRALIA

Nice ranking, Stefan!

 

[Orange Kid]

Thanks for the stats.
Added the config xml but it doesn't seem to take so am on the slow boat.

 

[ao_ika_red]

@ao_ika_red (?), when I said this I was kidding. I meant your username was too long for my stats script (which I fixed), not that you needed to change it.

Nvm Ken. Ever since I used that username I wonder if it would bring trouble to stats masters here. Once I detected a problem, I changed them on a whim. Unfortunately, it trigggered some unknown problem for me.