PrimeGrid World Space Week Kickoff Challenge (part 7 of the 2016 PrimeGrid Challenge Series)

Gilthanis

Gilthanis

[H]ard|DCer of the Year - 2014
Joined
Jan 29, 2006
Messages
8,729
October 3rd-4th. (24 hours)
SGS-LLR

Team page: [H]ard|OCP

This is part 7 of 9 challenges. It is CPU only. Our team is currently in 12th place for the series.

Current team standings are: Challenge 5 is still in cleanup phase and is not reflected below
25 - RFGuy_KCCO
101 - Coleslaw
105 - pututu
142 - HighTech67
268 - tjmagneto
285 - brilong
392 - travanx

Tips:
You must log into your account at PrimeGrid to change your project preferences so that you are only crunching the correct sub project. In this case SGS-LLR.
Intel boxes tend to perform better. The HT probably needs turned off since these are not sieve work units. However with how short these are, you may not notice much. Do your testing in advance.
Intel chips tend to out perform AMD chips
PrimeGrid work will stress hardware harder and so is more sensitive to overclocking. Therefore make sure you have tested their work units on your rigs before the challenge to make sure they are stable.
These challenges only count work units downloaded during and completed before end of the challenge. So, no bunkering.

Since these work units are really small, you should run with zero cache. You don't want to have work sitting around because the first person to complete the work unit gets credit for the Prime when found. SGS has the highest probability of finding a prime.
 
For this challenge, the G34's should finish on time. With a large amount of cores, they are worth running for the short time it lasts.
 
Ugh. Am having major issues with the SSH/SOCKS proxy I use at work. Not sure if it's my hosting provider or if it's the IT department finally getting wise. If I can't get this figured though it could mean little to no output from me; which is especially bad after talking a bit of smack after the last challenge.

See... the whole "under estimating a serious competitor" comments from some of the winning team bit struck a nerve. I didn't under estimate anything as I didn't make a conscious effort to try in the first place; simply switched over most of my systems to PG and that was that. Never disabled HT, which apparently was a huge mistake, nor did I babysit any of the systems (which was also a mistake). Kind of want to lay waste to everyone this time; but if my proxy is borked then I'm screwed. :(
 
Hope you're able to get it figured out, so you can join again. I'd love to see you blow everyone away. :D
 
You have a decent chance of finding a prime with the SGS tasks, probably more than with any other PG project, so that's a little added bonus to participate in this challenge. I have found 4 SGS primes myself. If fastgeek is able to participate in this challenge, I wouldn't be surprised if he found that many in 24 hours.
 
fastgeek said:
Ugh. Am having major issues with the SSH/SOCKS proxy I use at work. Not sure if it's my hosting provider or if it's the IT department finally getting wise. If I can't get this figured though it could mean little to no output from me; which is especially bad after talking a bit of smack after the last challenge.

See... the whole "under estimating a serious competitor" comments from some of the winning team bit struck a nerve. I didn't under estimate anything as I didn't make a conscious effort to try in the first place; simply switched over most of my systems to PG and that was that. Never disabled HT, which apparently was a huge mistake, nor did I babysit any of the systems (which was also a mistake). Kind of want to lay waste to everyone this time; but if my proxy is borked then I'm screwed. :(
Click to expand...
Maybe I shouldn't have posted the thread in the first place from the last challenge . Sometimes ignorance is bliss. I read it as a "friendly challenge" from them and recognizing that you are a threat to them and wanted to pass that info. IIRC, Z didn't say anything about you. I think Z seems like an ok chap but perhaps other team members that may have gone overboard.

Anyway, there are more PG challenges now and years to come.

To other [H]ardCrunchers, this is a one day challenge. If you are reading this thread, a few hours of contribution can help our team a lot. We lack number of participants unlike other teams. See stats below from previous challenge. If you decide to stress test your rig, this is a good time to take advantage of it and perhaps find a few primes as RFGuy has just said. Let see if we can get double digit participation!

upload_2016-9-19_20-41-15.png
 
pututu said:
Maybe I shouldn't have posted the thread in the first place from the last challenge . Sometimes ignorance is bliss. I read it as a "friendly challenge" from them and recognizing that you are a threat to them and wanted to pass that info. IIRC, Z didn't say anything about you. I think Z seems like an ok chap but perhaps other team members that may have gone overboard.
Click to expand...

No worries at all. Zune didn't say anything and have nothing against him in the least. :) A few others did poke the bear, jokingly or not; now the bear wants to have some fun. ;)
 
Ok, this is a bit strange. Test the SGS WUs on i7-2600K at 4.2GHz (see sig below). No CPU intensive program running in the background.

HT is ON but I varied the number of threads. I'm seeing peak output (ppd) with 6 threads. I'm not 100% sure if this is right, but will do a few more tests (HT off) since SGS has short WU.

Will be good if we all can post some stats. Thanks.

upload_2016-9-21_7-29-59.png
 
Am fairly certain my proxy issues are due to network problems in my building... with any luck they'll clear up soon. If things stabilize, I'll try to grab some WUs and see how they work out. :)
 
pututu said:
Ok, this is a bit strange. Test the SGS WUs on i7-2600K at 4.2GHz (see sig below). No CPU intensive program running in the background.

HT is ON but I varied the number of threads. I'm seeing peak output (ppd) with 6 threads. I'm not 100% sure if this is right, but will do a few more tests (HT off) since SGS has short WU.

Will be good if we all can post some stats. Thanks.

View attachment 8067
Click to expand...
Looks like I better switched off the HT. Output is 19K ppd.

Less than 7 days to go for this short one day challenge.
 
*looks around nervously*

.... I don't want to jinx anything .... but think the network problem might actually resolved. Will know more tomorrow. Fingers crossed. :) If things go well I might even have some more Broadwell servers to throw at this since it's just a one day challenge.
 
Decided to do a quick test while the getting is good. ;)

Good Ol; Thing02 will churn out 64 of these WUs every 4m12s - 4m40s. That's with two non-intensive projects running in the background; WUProp and Monkeys.

Will hopefully have more data points soon. Oh, and this time I'm making sure to turn off HT and that no other intensive projects are accidentally left on. Lesson learned.
 
Hope everyone is ready for this short run! I've got my main systems up, configured and running. Time permitting I'll throw a lot more at this Monday morning.

For those that are interested in rough average times per WU:

E5-2640v3 = ~835 seconds
E5-2680v4 = ~840 seconds
E7-8867v3 = ~860 seconds
E5-4650v1 = ~1,010 seconds
E5-4640v1 = ~1,160 seconds
E5-4640v2 = ~1,315 seconds

There are some more; but that will do for now... I should've left an hour ago. :p
 
So far it is

2 - fastgeek
35 - RFGuy_KCCO
74 - pututu
179 - Coleslaw
209 - HighTech67

We still have 20 more hours to go.
 
Averaging the 50 WUs or so, here are my run time per WU.

i7-2600K 4.2GHz ~ 729 secs (12 mins 9 secs)
E5-2695v2 3.0 GHz ~ 1204 secs (20 mins 4 secs)

Update: being lazy to turn off the HT on E5-2695v2 but decided to take the extra effort. Now the run time is about 977 secs (16 mins 17 secs) with this headless system. Yes, I've to carry monitor and keyboard around. I need to figure out booting into a BIOS remotely in the future from another PC withing Win 7 environment. If anyone know of a free software or a good guide, please let me know. Thanks.
 
Last edited:
On a couple systems I followed the suggestion from Gilthanis of simply halving the available CPUs that BOINC can use vs. screwing with the BIOS. Might be leaving a little on the table, but not a huge amount. Just create a cc_config.xml and pop it in C:\ProgramData\BOINC (by default at least) and then tell BOINC to read the config files.

Code: 
<cc_config>
   <options>
      <ncpus>12</ncpus>
   </options>
</cc_config>

Just a thought... whole lot easier than messing about with the BIOS!
 
Hmmmm. I'm not so sure now. Am looking at two boxes that I think are setup the same in regards to BOINC (meaning 100% usage) where one is set in the BIOS to not use HT and the other is using the cc_config file to only use the actual number of physical cores. The OSes aren't 100% identical though (2012 vs 2012 R2) and I'm also not completely certain about the resource usage. None the less....

System using 2012 w/ cc_config = ~2,100s on average (without cc_config and using 24 "cores" took 5,000-6,000 seconds!)
System using 2012 R2 and BIOS = ~1,800s on average

These boxes are using dual X5680 CPUs. Those chips used to be pretty beastly; but now they're seriously out performed by a Sandy Bridge E5-2620. Same number of cores... yet lower clock speeds + new tech makes mincemeat out of the old X's. (~1,280s per WU for the 2620)
 
Team is doing well in third place, within spitting distance of second! Come on, folks! These WUs go fast! Pause your other projects for all of eight hours and lets have at it!

Code: 
1    Aggie The Pew    2085282.32
2    Czech National Team    1402082.44
3    [H]ard|OCP    1311997.61
4    SETI.Germany    1158250.69

As for the race between Zune-san and I? Well, I'm not counting on anything until the clock ticks over to 11:01AM PST. Presuming my proxy doesn't crash (pounds on wood) I may be OK. Just upped the buffer on my big boxes just in case. Had them set to getting WUs as needed to maximize finding Primes; but realized that was a risky move for the race. Currently have 242,554.51 points on him....... fingers crossed things stay stable. :D Oh, and that he doesn't have a massive data dump coming up with five minutes to spare. :p

Code: 
1    fastgeek    [H]ard|OCP    1175413.51
2    zunewantan    Aggie The Pew    932859.00
3    Scott Brown    Duke University    407517.13
 
fastgeek, no cc_config needed. Just change the preferences from within the client. Options > Computing Preferences > Computing tab > Use at most X% of the CPUs.
 
1 - fastgeek
35 - RFGuy_KCCO
78 - pututu
140 - Coleslaw
239 - HighTech67
 
Anyone know Jathanis? He has been crunching PG for some time already. If J can switch to SGS, we will have six participants. Still about 3.5 hours to finishing line...
 
fastgeek, you're being mentioned here again. It appears that our team is being acknowledged in this challenge, all from your humongous contribution.
 
Heh. Yup. I've been watching that thread. While it's not over until it's over, am fairly certain that Zune can't pull out ~40% of his total output from the systems that haven't reported in recently. Guess we'll know in 135 minutes. :)
 
Could have sworn [H] was in second earlier; however if that was true, the Czech team snagged it away at the last second as we finished 3rd in the Team Challenge.

As for the Individual Challenge; the results of this event and the TRP Sieve event ought to silence the WTFgeek stuff. With a gap of ~556K points between us, Zune found himself quite thoroughly beat for the lead.
 
Just saw this. Congrats fastgeek for beating Z-san the second time. Not an easy opponent, I've to admit.
 
Jathanis is one of our long time members. I haven't seen him in the forums lately. His PG output is probably GPU based.

So until cleanup is finished and final results are in, it looks like:

1 - fastgeek
36 - RFGuy_KCCO
78 - pututu
135 - Coleslaw
240 - HighTech67
 
Looks like you're right that Jathanis is running PPS project. Well, still end up with five participants, no improvement.

fastgeek, the next PG challenge is a mix of CPU and GPU and as always GPU is the way to go.
 
You must log in or register to reply here.
Back
Top