• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

PrimeGrid LLR Races Thread 2016

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
16 hour warning, PG race is about to happen!

This is primarily a GPU race, put your CPU cores to use on something that doesn't have a GPU app (POGS, Rosetta, Universe, WCG etc etc etc)

Make sure you have your preferences set for Proth Prime Search Sieve (PPS-Sieve) and that the correct brand GPU box is checked off to the right, and that up top of the preferences you have checked the 'use GPU' box.
 
TennesseeTony said:
This is primarily a GPU race, put your CPU cores to use on something that doesn't have a GPU app (POGS, Rosetta, Universe, WCG etc etc etc)
Click to expand...

Hmm, is it correct that a PPS-Sieve task takes about 10...20 hours on a CPU, but at the order of 5...30 minutes on a GPU?

[edit:] Indeed. BDW-E @ 4.1 GHz needs about 11 hours, GTX 1070 somewhat over 6 minutes.
 
Last edited:
The race has started! Though I was too lazy to get up when it started, my computers weren't. 😉

4-hour stats:

Rank___Credits____Username
11_____1132656___10esseeTony
64_____262938_____xii5ku
140____107872_____Ken_g6
226____53936______geecee
298____26968______Fardringle
298____26968______Orange Kid

Rank__Credits____Team
6_____2090020____BOINC@MIXI
7_____1934954____Team 2ch
8_____1806856____US Navy
9_____1611338____TeAm AnandTech
10____1513579____Gridcoin
11____1503466____Rechenkraft.net
12____1270867____Team China

Welcome to xii5ku! Did you come over from the F@H race? I don't see you listed there.
 
Ken g6 said:
Welcome to xii5ku! Did you come over from the F@H race? I don't see you listed there.
Click to expand...

Thanks for the welcome; xii5ku is my alter ego. So yes, I came over from the F@H race were I had some good initial results with two large CPU clients. They don't receive suitable WUs anymore since Thursday, but luckily two GPUs arrived just in time to compensate. But seeing the 3-day PG challenge announcement, I thought I use this to dry out the factory smell before putting the cards to F@H duty.
 
Smells like a pair of 1070's that got a slightly late start to me. 😉 Or.... come on man! Whadda ya git! 😀 I always did like watching other people open their Christmas gifts, and seeing what they got. 😀
 
TennesseeTony said:
Smells like a pair of 1070's that got a slightly late start to me. 😉
Click to expand...

One is a 1070 indeed, with a hefty triple-slot cooler (Gainward Phoenix GS). This one works out well so far; it is supposed to go into a PC for daily work where noise would be a concern.

The other one, I am embarrassed to say, is a 1080. This went into a PC which I use in a slowly progressing experiment on whether or not watercooling is a viable alternative to largest-possible air coolers. Considering the cost of the waterblock, I thought I could as well go for the 1080 instead of a 1070 for this one. Unfortunately this card (a reference board, from inno3D) emits considerable coil whine at >= 1.7 GHz GPU core clock.
 
Day 0.5 stats:

Rank___Credits____Username
12_____3529437___10esseeTony
68_____947251_____xii5ku
165____343842_____Ken_g6
208____266309_____geecee
339____117985_____zzuupp
346____114614_____Orange Kid
413____70791______Fardringle
435____57307______waffleironhead
570____13484______GLeeM

Rank__Credits____Team
8_____6506030____Rechenkraft.net
9_____6138591____Gridcoin
10____6050945____US Navy
11____5461020____TeAm AnandTech
12____5016048____Team China
13____4233976____USA
14____4126104____SETI.USA

We've fallen back from where we were earlier. 🙁
 
Some observations with the "PPS (Sieve) 1.39 (cudaPPSsieve)" application:
Code: 
card     power    core        time       power     energy    comment
         target   frequency   per task             per task
---------------------------------------------------------------------------------
1080     none     1.6 GHz     292 s      125 W     37 kJ     reference
---------------------------------------------------------------------------------
1070     none     2.0 GHz     319 s      165 W     53 kJ     same power target
1070     77 %     1.9 GHz     334 s      125 W     42 kJ     same power
1070     65 %     1.8 GHz     348 s      105 W     37 kJ     same energy per task
1070     46 %     1.6 GHz     386 s       80 W     31 kJ     same core frequency

Power was not measured externally; it is taken from HWinfo64 as average over the entire runtime of multiple tasks. How far is this from reality? This is board power; the host's PCs power usage is not included.

Time per task as averaged over several tasks as well.

(Earlier I had the GTX 1080 running cudaPPSsieve at 1.8 GHz and about 190 W (with 120 % power target), but so far I am unable to reproduce this.)

Conclusion: At this particular application, six 1070s throttled to 1.6 GHz are about as fast as five 1080s at stock 1.6 GHz, but consume less energy (if power consumption of the host PCs remains the same). Which is kind of obvious in hindsight.

Would ten 1060s boosted to 1.6 GHz be more efficient than five 1080s? Not if power usage scales similar to TDP.

Edit, conclusion corrected:
  • six 1080@1.6 against seven 1070@1.8 achieve same throughput at same energy efficiency.
  • six 1080@1.6 against eight 1070@1.6 achieve same throughput, but the latter are more energy efficient.
 
Last edited:
Day 1.5 stats (halfway!)

Rank___Credits____Username
13_____10237727___10esseeTony
70_____2774333____xii5ku
188____987703_____GLeeM
194____957364_____Ken_g6
232____711281_____geecee
290____532618_____zzuupp
383____323616_____Orange Kid
462____219115_____Fardringle
476____202260_____waffleironhead
874____6742_______VirtualLarry

Rank__Credits____Team
9_____17876413___BOINC@MIXI
10____17559539___Rechenkraft.net
11____17482006___US Navy
12____16952759___TeAm AnandTech
13____16517900___USA
14____13180610___Canada
15____12388425___SETI.USA

Well, somebody finally passed me. That should only require a 1070, or maybe a fast 980. But overall we've fallen another place. 🙁
 
Seeing a posting at primegrid suggesting to run two tasks per card for better utilization of cards with high core count, I created the file "C:\ProgramData\BOINC\projects\www.primegrid.com\app_config.xml" with the following contents and let the boinc manager re-read the configuration files.
Code: 
<app_config>
    <app>
        <name>pps_sr2sieve</name>
        <max_concurrent>4</max_concurrent>
        <gpu_versions>
            <gpu_usage>0.5</gpu_usage>
            <cpu_usage>1.0</cpu_usage>
        </gpu_versions>
    </app>
</app_config>
(This is valid for 2 cards in the box, times 2 concurrent tasks per card. I specified in the web interface to accept only GPU tasks, no CPU tasks. Actual CPU usage of each GPU task remains way below 1.0 core.)

This improved the mean time between task completions slightly:
GTX 1080 @ 1.6 GHz: 292 s -> 276 s (6 % more throughput)
GTX 1070 @ 2.0 GHz: 319 s -> 290 s (10 % more throughput)

I guess the effect of running two tasks per card would be more pronounced on cards with core counts like Titan's or Fury's.

StefanR5R said:
(Earlier I had the GTX 1080 running cudaPPSsieve at 1.8 GHz and about 190 W (with 120 % power target), but so far I am unable to reproduce this.)
Click to expand...

I tried to fix this by reboot and had it running at 1.8 GHz and 200 W (120 % power target) once more, but only for a short while. Now it is permanently down to 1.6 GHz again. Well, Furmark is always able to bring up 1.8 GHz, but not cudaPPSsieve, neither one nor two instances of it in parallel.

[Edit:] formatting
 
Last edited:
I'm running 4 wu at once on my hd6950. 1 wu at a time was 2300 seconds. 4 at a time takes 8000 seconds each. So, 2000 seconds a WU. I havnt tried 6 or 8...yet.
my app_config.xml
<app_config>
<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>.25</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
</app_config>
 
Day 2.5 stats:

Rank___Credits____Username
14_____16686450___10esseeTony
73_____4827272____xii5ku
162____2019229____GLeeM
200____1594483____Ken_g6
238____1277609____geecee
281____1034897____zzuupp
402____546102_____Orange Kid
473____360697_____waffleironhead
482____350584_____Fardringle
979____13484______VirtualLarry

Rank__Credits____Team
9_____30591825___Rechenkraft.net
10____30376081___BOINC@MIXI
11____29631090___US Navy
12____28710807___TeAm AnandTech
13____27962445___USA
14____22302536___Canada
15____21547432___SETI.USA

12th place looks fairly solid at this point.
 
Mostly-final stats:

Rank___Credits____Username
14_____19686640___10esseeTony
72_____5777894____xii5ku
155____2504653____GLeeM
202____1887760____Ken_g6
215____1783259____geecee
278____1254012____zzuupp
407____647232_____Orange Kid
491____407891_____Fardringle
498____394407_____waffleironhead
1039___13484______VirtualLarry

Rank__Credits____Team
10____36265218___BOINC@MIXI
11____35361790___US Navy
12____34741526___USA
13____34357232___TeAm AnandTech
14____26715175___Canada
15____26340994___SETI.USA
16____23114947___AMD Users

Looks like some people made a surprise decision to make USA great again. 🙄
 
Ken g6 said:
Just thought I'd bump this thread one more time for the final stats: http://www.charleygielkens.nl/pg/index.php?challenge=0&uot=t&year=2016&team=undefined&type=html
[...]
Edit: Those are not the final standings. The last race (Winter Solstice) is final, but hasn't been added to that table yet.
Click to expand...

It still wasn't added, and presumably never will.
Michael Goetz said:
The person who was in charge of that is no longer available, and that list was done on his private resources. We haven't replaced it yet, so 2016 is the latest year with cumulative totals.
Click to expand...
http://www.primegrid.com/forum_thread.php?id=7623#111009
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top