7th Annual BOINC Pentathlon (2016)

We finished 13th position in the POGS (City Run). A solid 1M points ahead of the closest team but 700K points shy of overtaking the next team. At the end of 1st day, we were way back at 21st position but thanks to fastgeek (those xeon E7-8867v3 4p boxens are awesome), we made tremendous improvement. I did a quick stats of team members contribution based on BOINCstats. The numbers are close to SG stats. No matter how big or small, all contributions are welcome! It is still not too late to participate in the Pentathlon.

Key takeaway from this challenge: A lot of bunkering and hoarding of WUs. See next post.

One down and four more to go.

upload_2016-6-12_21-58-23.png
 
Last edited:
After POGS ended, I looked at my validation pending, I still have 22.3% after I stopped crunching about 1.5 days before the run ended.

upload_2016-6-10_22-2-30.png


Out of the 251 WUs pending for validation, there are still 80 WUs completed on the 1st day itself that have not be validated yet. To be honest, there are 5 out of 80 WUs which I did bunkering and still pending!
upload_2016-6-10_22-4-35.png


Anyone is running Universe at home? I'm still on CSG and PG.
 
I'm running all 3 right now. Waiting to see what the last project to be announced is. Plus CSG keeps running out of work, so figured couldn't hurt to pick up a few.
 
For CSG, there is a good suggestion by overclock.net member. Create a batch file and I've ran this yesterday to keep my rigs loaded with CSG WUs. One thing I like about CSG is that it doesn't seem to allow you to bunker up lots of WUs, so I'm currently getting about 14% pending validation rather than 50%+ in most days in POGs. I'm running Windows so I lost some efficiency versus linux. I'm not running Universe@home yet fyi.
upload_2016-6-11_5-53-23.png
 
I'm the opposite. 27 waiting to be validated, 23 Valid in CSG. Been like that since the start.
 
I'm a little late showing up but I started working on PrimeGrid yesterday.
 
Less than 30 minutes before Universe@home portion begins.
 
We have dropped to 18th as the bunkers dropped leaving us in 28th place at Universe.
 
Also, Rosetta@home is declared the Sprint project. Perhaps more of our team will help since it is only a 3 day event for that section.6/15-6/18
 
For those unfamiliar with Rosetta, it is CPU only and each work unit uses ~512MB RAM.
I wouldn't bother with Android on this project as the work units are RAM heavy for those devices typically and the run times may not be so great for the 3 day challenge. I would recommend those devices stay on CSG.
 
Last edited:
Gilthanis said:
We have dropped to 18th as the bunkers dropped leaving us in 28th place at Universe.
Click to expand...
I saw this last night, so decided to fire up Universe. However as of this morning, 100% awaiting validation!
Wondering what is fastgeek next move?
upload_2016-6-12_7-7-8.png
 
Gilthanis said:
For those unfamiliar with Rosetta, it is CPU only and each work unit uses ~512MB RAM.
I wouldn't bother with Android on this project as the work units are RAM heavy for those devices typically and the run times may not be so great for the 3 day challenge. I would recommend those devices stay on CSG.
Click to expand...

I've had a lot of compute errors on my Nexus 7/9 and Rosetta. Not worth it on those devices. It's not just me either. Some tasks have 3-4 errors.
 
As soon as I get my new AC unit in place, I may have some CPU power to contribute. Right now, my GPUs are working pretty hard at F@H trying to keep our numbers up there, but the CPUs are more or less Idle. Mostly i5/i7 at the moment, but I have a couple of Xeons E5-2620s that will be available possibly in August. What kind of CPU hardware would make a dent in these BOINC projects? You'll have to forgive me, my DC experience is mostly limited to S@H and F@H. I have installed BOINC before, and ran it for a few weeks as a possible alternative when there was talk of BigAdv going away, but I wasn't perusing the forums enough to keep a solid interest.
 
Hardware selection typically depends on projects themselves as some will have optimized apps. However, since you are using decent Intel chips, those tend to rock most projects pretty well. August won't help with the current Pentathlon, but will help with other challenges going forward. If you choose PrimeGrid, Hyper Threading should be typically turned off. Most other projects it should be turned on. WCG is close to [H]'s heart and in my opinion the most rock solid and valued science projects out there as it actually has had real world results tied directly to the results we have crunched. However, you will find that WCG is considered a low point payout project in the BOINC world. So, if you decide to stick to just one project, then no biggie. If you want to compare your points between projects, then you will find yourself earning low figures. As far has what hardware, some folks at WCG are collecting efficiency data from members here: World Community Grid - View Thread - Points per watt: Lets try an update or you could always look through the data at WUProp (which everyone should attach to as well) here: Report limit

I tend to suggest Intel chips. If you have ARM devices, they may not put out tons of points per device, but they tend to have the highest efficiency of points/watt. I just don't typically recommend going out and buying ARM devices to create a phARM with. Project selection is a log more limited than standard CPU based projects and they are a lot more succeptable to heat.
 
My only tablet is already running F@H, albeit anonymously because there's no way to tie it to existing accounts, so no worries there. I think WCG is the project I ran for a few weeks, but it was hard to get excited about because I knew less about it and didn't have time for the research. I'll put it up on one of my 8-cores and go from there. It's my understanding that the BOINC client software is all that is needed in a Windows environment. Any suggested add-ons for management, viewing, etc.? I've already got remote access to all machines, so at least I've got that covered.

EDIT Update. Looks like I've got it running on one box. Seems like I had all my info still attached to the account for [H]ardOCP from when I first signed up back in early 2014. Not sure if I have a way to verify that without digging into some tutorials. Does anyone have a good newbie walk-through link that includes stats check, team assignments, and whatever verifications I need to put in to get proper credit for the team?
 
Last edited:
Most references needed can be found here:
All Inclusive DC Project list v.2

That includes team names and I believe I have created a shortcut to the teams page on each project as well.

As far as managing multiple clients. There is a 3rd party program that you can use located here: BoincTasks download | eFMer I personally don't use it, but it is very handy tool. IIRC it is similar to the HFM.net tool in concept. We do not currently have a how to guide on it to my knowledge. Perhaps someone would take the time to create one.

For stats sites/account managers, I recommend using BOINCStats BAM! located at www.boincstats.com You can get a quick setup instruction here: | BOINCstats/BAM!
 
The PrimeGrid portion (City Run) is now complete. We finished in 17th place in that one.
 
Gilthanis said:
Most references needed can be found here:
All Inclusive DC Project list v.2

That includes team names and I believe I have created a shortcut to the teams page on each project as well.

As far as managing multiple clients. There is a 3rd party program that you can use located here: BoincTasks download | eFMer I personally don't use it, but it is very handy tool. IIRC it is similar to the HFM.net tool in concept. We do not currently have a how to guide on it to my knowledge. Perhaps someone would take the time to create one.

For stats sites/account managers, I recommend using BOINCStats BAM! located at www.boincstats.com You can get a quick setup instruction here: | BOINCstats/BAM!
Click to expand...

I just got BoinCTasks setup. In addition to monitoring like HFM it can also control Boinc like suspending, adding projects, etc. More useful to those that multiple PCs as they can all be controlled from one PC.

Adding this into cc_config on the remote PCs was the easiest or me:

Code: 
<options>
<allow_remote_gui_rpc>1</allow_remote_gui_rpc>
</options>
 
We are back up to 14th but we will see how things settle as more people drop their bunkers.We are currently in 10th place at Rosetta. I would love to see that position much higher.
 
Team is currently 8th with Rosetta. I'm hoping we get a few big hitters to surprise us with some rigs on this. It would be nice if we could hit the top 3 in this category.
 
Last edited:
Is there a stats page that confirms I'm donating to [H] currently? I'm in the middle of a domain migration for a customer and just don't have the time to do the research ATM.
 
Last edited:
Gilthanis, as MGMCCALLEY, only WCG so far. I stopped using the Olmec name quite a while ago because people didn't recognize folding vs. forum. Would need a quick tutorial to crunch under anything else. Although it's just an 8-core AMD at the moment. Not sure if it would make a dent in the current contests or not. Gotta say BOINC seems intimidating when compared to F@H. Lots of options, seemingly lots of maintenance. If I had educated myself on the ins and outs before I started working 18 hours a day I might understand it more. Nowadays I'm just exhausted all the time and I have no motivation to read documentation after having poured through hundreds of pages a day already.

Also, I just noticed that even though I limited BOINC to 65% on that CPU, my F@H performance on my GPU took a HUGE hit, to the tune of 420 thousand PPD drop to about 80k on a 980 Ti. Maybe F@H and BOINC on the same rig is a bad idea until I figure out how to properly assign cores to projects across multiple apps and make it stick between reboots. Until then, no new BOINC tasks on F@H machines.

Sooo... tired.... must... drink... coffee......
 
The advantage here is that if WCG is working fine, the other projects in this challenge can easily be added and you can let the client decide which project to call for work on. Since the remaining projects are all CPU based. If you saw a point reduction in FAH, then certainly looking over the setup should be done as it shouldn't have affected FAH like that. You reserved cores back, those should have fed the GPU's. However, I don't know how many GPU's are in that single box. Rule of thumb, reserve a full core per GPU and you should have more than enough in general. 65% should only give BOINC 5 cores to play with on the octacore system.

Edit: I would also look at heat as the system may have throttled back in order to compensate too. Double check if any drivers changed...
 
Last edited:
And if you want to pause WCG until the end of this challenge, Just select WCG and click the No New Tasks button. This will essentially let the current work units finish but tells the client not to ask for more from that project. If you suspend the project, it will stop asking for work on it AND pause all work in your que as well...
 
I did just that, and it was running 5 tasks, which I assume took up 5 cores? And I did find the No New Tasks button, so it should finish up here within 16 or so hours and I'll see if/how my F@H bounces back. I hate leaving a project unfinished, so the WCG tasks will be allowed to finish up and I'll see how it scales with F@H as WCG ramps down. There's only one 980 Ti in that rig, so I'm not sure what's going on. I recall a way to lock F@H to a specific core in an old version or add-on, but I don't see that in the configs now. Is there something similar on the BOINC side?
 
Not built into the client. I think some people will manually do it using process lasso or similar 3rd party application.
 
If you focus on Rosetta, those work units do not need a wingman and is the next project to end in the over all Pentathlon. It is protein based research if that matters. Then Universe@home and finally CSG. CSG keeps having spurts of work. So, once Rosetta is done, just crunch both and let her ride. Just my suggestion anyways...lol
 
We just dropped to 9th in the Rosetta portion.
We are down to 18th at CSG
And we are up to 21st at Universe today
 
Last edited:
Since last Sun, I switched all my rigs to Rosetta (yes I did some bunkering) hoping that we can finish this sprint run with a credible position. On the last day, I can switch back to either CSG or Universe and hoping that I can get at least 50% or more WU validation....
 
When you switch, attach to both. CSG has had trouble keeping a supply of work units. If you want to set CSG to a higher priority, that would work best for trying to get work from there more often...
 
Looks like my PPD drop may have been heat or HDD related. Got a new AC unit put in and testing the theory now. If it's heat, it should be solved, if it's drive, I'll need to image it to another or a small SSD. Hopefully I'll have WCG finished up soon and start on some other projects to help you guys out.
 
Last edited:
Well... the good news is that if it were HDD related, it may be as simple as not running CEP2 (assuming you were). CEP2 is IO intensive.
 
Oh I've got a stack of 128 GB SSDs that I pulled from some laptops a while back. I should really be using them for things like folding. Reduces power usage and speeds up reboots, etc. I just had regular hard drives in those rigs because at one point I intended on selling them. I actually think it's temp related, though, because after I put the new AC in and the temps dropped into the 70s it seemed to clear up the problem. It's currently exceeding 630k PPD on a 980 Ti which is about as good as that rig has ever done. It's possible that if I get that kind of improvement on more of my cards that I go up 10% or more PPD overall. Would love for that to be true. Thanks for all the info. When I get some free time I'll peruse and dive into another project. The current contest is for Rosetta?
 
You must log in or register to reply here.
Back
Top