Christmas Race 2010

If there are any questions from the WCG folks about whether you are "legit", hopefully you can point them to the poll here as "proof" that you are the newly voted rep.
 
metallicafan said:
Easy fellas. I waited quite a while to make this post until someone from our team had actually heard from hito, but i guess I didnt get us entered as I assumed he would. Sirmonkey - the vote you linked it still open for a few hours. ;)

I have submitted a support email to WCG and posted in the forums asking if an Admin can enter us. The challenge is percentage increase and allows late entrance so if we can still get entered by tomorrow it shouldnt effect the outcome.

If we dont get entered for the challenge, for those of you that I disappointed, please accept my apology. I appreciate all of the people who have stepped up to help, but no hard feelings if you decided to run f@h.

Either way I will still participate and I will keep track of our teams participation manually and post it here. It wont be on the official WCG stats page but that matters very little to me anyway. ;) Thanks again.
Click to expand...


nah not completely blaming you. only about 25% of it. rest of it goes to the guy that went inactive on the team and didnt have the common sense to transfer the team to some one else. this should of been ironed out well before this thread was made though. would of been nice to give them a run for their money in the competition but joining late just takes all the fun out of it.
 
Actually from the way I read it, it sounds like late entry is the best way. Its a percent increase therefore before the start we have a lower baseline (fewer people doing work) Then a few days in we destroy by adding a lot of people that usually do f@h as well as general ramps by others. They wouldn't know what hit them. That is why I waited to switch my computers over. It was actually worse to add the clients before the race due to the fact it would increase our baseline thus lowering our increase.

I personally like the points races better but, this is my take on the race.
 
Well part of this is mine to own. I should have been aware that the Captain of WCG was unresponsive and requested Admin of the team as I am in most other Boinc Projects.
I fully support the elected Captain of WCG metallicafan and encourage him to contact the Admins to make the switch.
 
rmdashrrootsplat said:
Just go to the projects tab and press the "No New Tasks" button for the WCG project.
Click to expand...

Thanks for the info. I've currently set this as it appears like the team may not get into the contest this year and I would like it to get all currently downloaded work finished before I move back to F@H. If however something does get worked out and the [H] team does enter this year I'll fire them back off. :)
 
Nice bump over the last couple of days......Hopefully we'll get into the contest.
 
We are in. The race is on. :D

I'll look more in depth at everything when I get off work tonight. :)

Crunch [H]ard.

Sorry again for the late entry. Like I mentioned earlier (and as Eric eluded to) the late entry really wont hurt us but still my apologies for not being on time.

Looking forward to see what some of you can do with those hex cores!
 
thank you metallicafan. good to see you were able to speed up the process of becoming captain!!! I just removed the no new tasks and turned in 22 WU's :D ya thats right i was saving them til the very last minute..
 
got my laptop going. will switch some CPUs over at some point as well
 
Good job metallicafan!!!

Glad to see we are in this.
 
metallicafan said:
We are in. The race is on. :D
Click to expand...

Nice work getting that taken care of.
thumbsup.gif



I'll go convert the 980 over now.
 
rflcptr said:
I'm excited! :D
Click to expand...

Me too!

Now for the noob questions - how do I know that it's working right? I see 12 jobs running, which makes sense on an Intel hex-core. The clean water ones estimate 2:50:xx to complete, and the childhood cancer around 5 hours. Are those appropriate? Also, I see the CPU utilization in task manager looks like a sawtooth on all cores, varying from 30 - 100%. Is that normal?

Where's the best stats pages to look at, and how often do they update?
 
Punchy said:
Also, I see the CPU utilization in task manager looks like a sawtooth on all cores, varying from 30 - 100%. Is that normal?
Click to expand...

Punchy, I don't have any CPUs with hyper threading but all three of mine (Q6600, some quad xeon, and some non HT hex xeon at work) are pegged at exactly 100% without any saw toothing or spiking in general. There are 4 or 6 processes in task manager, each one runs at 25-22 (Edit: or ~16 lol) with my other programs taking up the left over %.

To me, it seems like WCG doesn't play nice with Hyper Threading?
 
Last edited:
What happens is a unit runs on a thread of the processor. Therefore my i7 has 8 WUs running at the same time. (Versus running one unit on all the threads). This keeps the processor running at 100% all the time.

Now for the estimated time, until you do a few units the time jumps around. After about ehh lets say 5 to 10 WUs it is pretty accurate. Thats why at first it could say its 5% through for a while.
 
Got the following working on this for us, just a small contribution:

Q9450 - 3.5 Ghz
Athlon X2 5000+ - 2.6 Ghz
Phenom II X6 1090T - 4.0 GHz

sooo, 12 cores, pretty small considering we have SR-2's here now
 
Punchy said:
Me too!

Now for the noob questions - how do I know that it's working right? I see 12 jobs running, which makes sense on an Intel hex-core. The clean water ones estimate 2:50:xx to complete, and the childhood cancer around 5 hours. Are those appropriate? Also, I see the CPU utilization in task manager looks like a sawtooth on all cores, varying from 30 - 100%. Is that normal?

Where's the best stats pages to look at, and how often do they update?
Click to expand...

the clean water ones take roughly 3-3.5 hours on my phenom II 940 @ 3.4ghz. but dont trust the times til you complete about 15-20 WU's. the program averages out your completion time the more WU's it does. the cpu usage may fluctuate periodically. i suggest in the settings to set under advanced/preferences/processor usage "While processor usage is less than 25%, set that to 0.


stats: http://boincstats.com/
just look up your name then on the right side click the magnifying glass.
stats update every 24 hours on WCG's site, but im not sure about on the boincstats page. also note WCG uses its own point system so your WCG points will not match what you see on boincstats since it uses the correct boinc point system.


W.Feather said:
Got the following working on this for us, just a small contribution:

Q9450 - 3.5 Ghz
Athlon X2 5000+ - 2.6 Ghz
Phenom II X6 1090T - 4.0 GHz

sooo, 12 cores, pretty small considering we have SR-2's here now
Click to expand...


physical cores are better then logical cores in boinc so your 12 cores will probably compete fine against an sr-2 running 24 threads.
 
sirmonkey1985 said:
the clean water ones take roughly 3-3.5 hours on my phenom II 940 @ 3.4ghz. but dont trust the times til you complete about 15-20 WU's. the program averages out your completion time the more WU's it does. the cpu usage may fluctuate periodically. i suggest in the settings to set under advanced/preferences/processor usage "While processor usage is less than 25%, set that to 0.


stats: http://boincstats.com/
just look up your name then on the right side click the magnifying glass.
stats update every 24 hours on WCG's site, but im not sure about on the boincstats page. also note WCG uses its own point system so your WCG points will not match what you see on boincstats since it uses the correct boinc point system..
Click to expand...


ya they do, but clock speed wise only the 1090T keeps pace with the SR-2s




physical cores are better then logical cores in boinc so your 12 cores will probably compete fine against an sr-2 running 24 threads
 
sirmonkey1985 said:
the cpu usage may fluctuate periodically. i suggest in the settings to set under advanced/preferences/processor usage "While processor usage is less than 25%, set that to 0.
Click to expand...
I'll have to give that a try, because right now it seems to be horribly inefficient at using 12 "cores", averaging about 70% CPU utilization overall (even though 12 projects are running).
 
So I am wondering though

On an i7 if i have HT turned on i can compute 8 WUs simultaneously. Would it be fast to turn off HT and run 4 units twice as fast?
 
Punchy said:
I'll have to give that a try, because right now it seems to be horribly inefficient at using 12 "cores", averaging about 70% CPU utilization overall (even though 12 projects are running).
Click to expand...


i think it has to do with the logical cores and it not being able to use them correctly due to the cache being over loaded(stupid intel and their damn shared cache).

i know there were a few non F@H projects(not sure if WCG is one of them) but they showed a performance increase by not using the logical cores. since that gives the physical cores 100% of the core/cache to use instead of sharing it with the logical core. its the one thing most people forget is that the logical core doesnt give you 100% increase in performance. it really only gives about 60% while taking away 20-30% from the physical core thread. but you could try it both way's and see what happens.
 
70% hmmm..

I have my dual E5530 setup running 16 projects and its running 100%
 
Punchy said:
Also, I see the CPU utilization in task manager looks like a sawtooth on all cores, varying from 30 - 100%.
Click to expand...

Check for a discrepancy of CPU usage between the 'Simple View" and the "Advanced View", specifically if the "Simple View" is set to 60%.

My hex CPU utilization was weird too until I changed those two settings (or changed them back).
Make sure the 'Simple View' CPU usage is set to 100% and the 'Advanced View" CPU usage is set to '0'. When I did that all 24 threads started running at 100%.

Your times look the same as mine and I am still looking for another stats site. I believe WCG updates twice a day.
 
Okay, I fixed it. In the advanced preferences there was a box for "use at most x% CPU time" and it was set to 60% (I guess that's what KMac was referring to, and 60% must be the default). I set it to 100 and a few seconds later the sawtooth went away and it's running at 100%.
 
nitrobass24 said:
On an i7 if i have HT turned on i can compute 8 WUs simultaneously. Would it be fast to turn off HT and run 4 units twice as fast?
Click to expand...

I doubt that would work. It seems that BOINC scales differently than FAH because it is not a true SMP process. An i7 w/ HT enabled running 4 units will have 4 threads at 100% and 4 threads at idle.
The curious thing I noticed is if you set it up like that, BOINC seems to rotate the thread in use so the CPU heats up across the entire chip rather than all in one spot.
 
thats actually windows rotating it and not boinc. windows automatically moves the highest load program to an idle thread if something requests processing power from that core. kind of annoying but at least it works the way its suppose to.
 
nitrobass24 said:
So I am wondering though

On an i7 if i have HT turned on i can compute 8 WUs simultaneously. Would it be fast to turn off HT and run 4 units twice as fast?
Click to expand...

Nitro - Leave ht enabled. You will gain a good amount of performance from it! My i7 does much better with HT.

Also Puncy - Yeah you got it man. For whatever reason BOINC (on some systems) defaults to 60% CPU utilization. Everyone else who has just installed BOINC you might want to check this setting (this is the [H]ardforum :D). This setting can be found in the Advanced View under Preferences. There is also a setting on how many CPUs (cores) to use.

BOINC uses a single threaded WU and just starts one Wu for every thread available.
 
Punchy said:
Me too!
Where's the best stats pages to look at, and how often do they update?
Click to expand...

I use FreeDC stats. Just keep in mind that this site will use BOINC scoring whereas WCG has its own points system (in which the official scoring will be done) that is roughly 7x the BOINC scoring system.

Here is our team page on the WCG page including scoring graphs, etc. Quite a nice rise in those graphs!
 
OK, I am BOINC'ing now. I'll switch an SR-2 over tomorrow, although I don't think SR-2s are the way to go with this since it does not take advantage of SMP for a single project at all. It all looks too linear for the way my farm is set up. Great for older hardware i suppose, but my farm is too optimized for F@H to be nearly as effective in this.
 
74.41% (1,008,473 / 578,219)

Just look at that baseline average of 578,219! Relatively high compared to the other teams at the top.

We are doing 1,008,473 points as of right now quite a ramp up lets see if all these added hex cores and everyone that his helping out can get us to #1. :)

Just though id add a picture of an interesting set of WUs. Sooo many CEP2 haha.
Untitled.png
 
lol i havent gotten a single CEP WU.. :D

wcg.png



and those of you that look closely you will see how off the times are on estimated time vs actual run time on the WU. thats because i've only run about 15 WU's but the times are getting closer. at the start they were all over 8 hours.


oh by the way, those of you asking about the WU sizes and download/upload sizes check this link out. scroll to the bottom.
http://www.worldcommunitygrid.org/help/viewTopic.do?shortName=minimumreq
 
musky said:
OK, I am BOINC'ing now. I'll switch an SR-2 over tomorrow, although I don't think SR-2s are the way to go with this since it does not take advantage of SMP for a single project at all. It all looks too linear for the way my farm is set up. Great for older hardware i suppose, but my farm is too optimized for F@H to be nearly as effective in this.
Click to expand...


Your SR-2s will pound through work units and crank out the points. I wouldn't be shocked if you came close to equaling the output of my whole farm.
 
You must log in or register to reply here.
Back
Top