BOINC

Looks like the board on my dedicated server at OVH died the other night, they replaced the board and had it back up within about 20 minutes, good service :)
I logged on today to check how things were and was a bit surprised to see the load average just over 4 rather than just over 2 - turns out it now has a 1.86GHz 2C/4T D2550, rather than the older Atom series, so that's a nice upgrade. Recently renewed my hosting after it's been up for a month out of the 3 I paid for originally, so I paid up front for another year at just over £3/month. Excellent price for a dedicated server and I'm running a few bits and bobs off it at the minute along with BOINC.
 
For those that are running the new WCG project - Does the project seem to be stable? Any bad WUs in the production runs?

Looks like the system requirements call for 400MB of RAM. Anyone noticed the RAM usage per WU on these?
 
Mapping Cancer Markers has been running fine on all my machines. I haven't had any bad WUs. RAM usage is < 100MB. Runtimes are 3-8 hrs.

CEP2 is another story. Something is really wrong with that project now. I can't run more than a couple units simultaneously. I started having problems max'ing out my cores when the big WUs showed up. Lots of 'max disk usage exceeded' or 'signal 11' errors. Work units are resetting on my XP box and taking twice as long to run before hitting the 12 hour cutoff. The upload files keep getting larger. I had a 65MB upload ealier today. I saw a forum post recently mentioning they're considering releasing bigger units. That's just crazy. They really need to revisit how that project works.

Focusing on MCM1 now ...
 
Last edited:
Run times can vary. They are still getting the kinks out of the work units last I knew. They were running ok for the most part until the scientists weren't happy with how the results turned out so had put the project back on temporary hold while they make some adjustments. However, as I look at my devices I can see that work has been turned back on. Perhaps I will take a look at the RAM usage a bit further. Time to pause Cosmology again...

Edit: AgrFan, I hear ya. I stopped running more than one CEP2 work unit a long time ago. I got tired of babysitting rigs. Especially when it came to borged machines. Nothing kills a friendly borg more than BSOD's. I have found that some hard drives keep up for a while and later degrade. It happens to all brands, but I find Hitachi to be most finicky. CEP2 is the number one reason I started putting the BOINC data directory on a separate HD when possible. I have had others argue with me non-stop on this, but a hard drive can pass all diagnostics and still not like CEP2. I have a 100% success rate with systems not liking CEP2 being fixed with a new HD. However, wear and tear can eventually bring it down again. I have a whole shelf of HD's pulled from recycled PC's that are now kept solely for BOINC. It is just becoming harder and harder to find spare IDE ports to use them. I will say though that since the longer units came out, I haven't tested the HD replacement solution.
 
Last edited:
I agree with CEP2 - I had to limit my Quads to 2 WU as well, as more gave me the same headache. I have 10 days runtime to get me 1 year, and I will be done with this project for a while...

AgrFan said:
Mapping Cancer Markers has been running fine on all my machines. I haven't had any bad WUs. RAM usage is ~64-96MB per unit. Runtimes are 3-8 hours.

CEP2 is another story. Something is really wrong with that project now. I can't run more than a couple units simultaneously. I never had problems max'ing out my cores until the big WUs showed up. Lots of 'max disk usage exceeded' or 'signal 11' errors. The upload files keep getting larger. I had a 65MB upload ealier today. The uploads are hanging my router. I saw a forum post recently mentioning they're considering releasing bigger units. That's just crazy. They really need to revisit how that project works.

Focusing on MCM1 now ...
Click to expand...
 
I hadn't been paying attention, but YoYo now has the Cruncher OGR app for Android devices.
 
Gilthanis said:
I hadn't been paying attention, but YoYo now has the Cruncher OGR app for Android devices.
Click to expand...

Yeah, I've been running them now on my devices for a few weeks - they give scary long estimated run times at the beginning (65 - 70 hours on my Galaxy S3, ~125 hours on my old Galaxy Tab), but they seem to jump to 33/67% within 3-8 hours and wrap up within 12-18 hours max. I don't think I've had on run for more than 24 hours. I've also noticed that my run times as reported on the Yoyo site are completely screwed up - right now, I have 17 completed for credit - 11 have a run time of exactly 0 secs, the other 6 range from 2.83 - 31.4 seconds, all successful with credit. Not sure if it's the project or something in NativeBOINC...

Just as an FYI, I'm leaving bright and early tomorrow morning for a family vacation to Hawaii, so if you don't see me on here until after Black Friday, now you know why! I have Asteroids & WCG locked on all three Quads, and the systems should be fine unattended for a week... Fingers crossed! Have a happy Thanksgiving, everyone, and I'll see you for the December Christmas Race in WCG! :D
 
I have switched to the official BOINC client on my Bionic and Galaxy S3.

Galaxy S3 has only done one so far. 639.55 seconds 1.27 claimed 11.06 granted
Bionic has only done one so far. 31.73 seconds 0.04 claimed 18.71 granted
 
I see we are now on the list for the Christmas challenge. Just some pointers... CEP2 is very demanding. It uses probably the most Ram, can use a lot of bandwidth, and has a ton of IO's. I like the project in general, but for a challenge it may not be a good candidate to choose. FAAH seems to have the shortest work units at the moment and also the lowest RAM footprint. Unlike CEP2 and MCM, it also supports ARM processors. Last I looked CEP2 didn't score as well as other project as far as points either. For those liking Linux, last I knew the 64bit Linux systems were much better than Windows. However, I will admit that I've been a little rusty on my WCG efficiencies.
 
It appears that WCG takes a few days to get things validated I started a few rigs running WCG on Thursday and Friday and the daily credit has been increasing by about 3,000 a day since then. How long does it take it to get caught up to a steady daily average. ?

And what is the format of the Christmas Challenge, I am assuming that you just run WCG for team HardOCP.

Are there any rules or anything a person needs to know. ?

Is everybody switching everything over to WCG ?

I have seen the Christmas Challenge mentioned a couple of times but no instructions on what we need to do, perhaps a Christmas Challenge thread might be a good thing to do for some of us guys that might not be so involved with the other DC folding world.. :D

http://stats6.free-dc.org/stats.php?page=team&proj=bwcg&team=1411
 
As long as you joined our team, then you are all setup for the challenge. The only one who had to do anything to get that going is the team captain which is metallicafan. Points can take some time now that the options for projects are slim. With a new project, there are a lot of pending validations created since all pc's are considered unreliable for a small time. CEP2 isn't as popular due to its higher demands of systems and longer run times. The FAH work units are smaller and will probably build your RAC quicker. However, the Christmas race is based on points. Only points that are accumulated between Dec. 1st and Dec. 31st will count towards the challenge.

I have already switched the majority of my systems over to it. That way I can hit my 5 year goal on the new project and I already have points waiting to validate when the challenge begins.

For those considering joining but unfamiliar with BOINC, pending validation is a state where the project sends an extra copy of a work unit out so that others hosts can validate the accuracy of your system and therefore giving a reliable status. They will also periodically recheck a machine just in case it later starts producing invalids that aren't caught by the validator.

I agree a separate post would probably be good. I may create one tomorrow when I have more time unless someone else takes the initiative beforehand.
 
Last edited:
Thanks AgrFan. That info will be helpful for establishing this years. Only a few things will need edited from it. I will also try to post the hints and suggestions for WCG when I post tomorrow so others wont have to dig for it.
 
Here is the Christmas Challenge 2013 page:
http://hardforum.com/showthread.php?p=1040413899&posted=1#post1040413899

Sorry for getting a late jump on it this year. Glad there is some interest in doing it! I'll be moving all of my machines over to WCG this week. Grandpa - thanks for contributing! In the past 2-3 days has been mostly enough to even out my average points but I havent run the new project yet.

If there is anyone who has configuration recommendations, tips, tricks, etc for the post just let me know and I'll gladly add them.
 
Does anyone know much about Ripple Labs? https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,35847_offset,0 Apparently it is yet another cryptocurrency. However, I'm not sure if they are just paying you to be on their team or if they actually have you doing a portion of crunching to actually generate or monitor currency transaction. They aren't real open about that part. Everything I have seen so far has been to associate your WCG account with their team and trade points for XRP currency. I'm leary on cryptocurrency to begin with, but this group is now the number 2 team at WCG.
 
Constellation@home has limited work again. Get them while you can. :)
 
I contacted Travis Desell in regards to a few projects that he is the admin on. He responded with:

Hi Colelsaw,

You hit the nail on the head -- I'm trying to find some grad students to take an active role in the other projects. Most of my time is taken up with Wildlife@Home, especially now that it's being funded by the national science foundation.

I do have a student starting up in the spring as a research assistant, and he's already started working on sending out work (that would be Kyle on the forums). He's hoping over break he might get things set up to start having a more steady stream of work.

I also have a student whose getting close to releasing more work units for DNA@Home, finally. We've been collaborating with some people in UND's med school so we're going to move it's focus to looking through human DNA for factors that might influence cancer growth. It's pretty cool. With any luck in a month or so new work should be coming out of that.

Unfortunately I haven't gotten someone in the pipeline for subset sum@home. So I'm not quite sure when that will get back up and going.

What I really need to do is combine the forums for all three projects, since I've been spending a lot of time in the wildlife@home forums.

--Travis
 
Another update from Travis Desell:

No problem! I'll try and keep everyone posted as things progress. I'm hoping over break we'll start sending out work units for DNA@Home again.

--Travis
 
I added another dual core laptop to the farm this AM. Toshiba Satellite A205-S4577

Edit: And apparently also adding a Dell Latitude D520 dual core. :)
 
Last edited:
Don't you love it when additions are random and spontaneous?
 
I especially love when those spontaneous additions are from office inspections and my dept. says get rid of all of this junk. :)
 
Lol, oh but that is certainly not junk! Moar power!!!
 
Nope...I have pretty much an endless list of people in need of laptops..so even if I don't have a spot here for them, they will be put to use somewhere.
 
I wonder if the new GPU project will be more consistent than the current one was...
 
We can only hope. It was frustrating having idle GPU's.
 
In cases like POEM, I would always set another project to priority 0 as a backup. That way if the GPU or CPU was gonna set idle, BOINC would request work from the project with priority 0. Mine rarely sit idle. ;)

The worst part is that there just aren't very many bio/medical GPU projects to choose from. Like right now I'm having trouble with GPUGrid's SANTI work units creating BSODs. It has happened in 3 different boxes with 3 different GPU's so far. AMD cards only get to choose from POEM or go outside of BOINC. Otherwise, you have to look to other science categories.
 
Is the SANTI BSOD's an application problem you think? I was having similar issues and just gave up altogether.
 
I think it is the application. Others have posted troubles with the SANTI work units as well. There are a couple forum posts over there. I don't think it is a fluke that it happens only when running those work units and to get BOINC to quit crashing the system I have to pause it and abort the work units. One machine had GPU's run at all times selected, so to get where I could cancel the work units I had to exit BOINC before it fully loaded and then delete the GPUGrid project folder. That would cause the work units to fail and then it would re-download the normal project files when you first attach. So, those machines have been set to No New Work until a much later date. It isn't the drivers because it is happening on two different sets of drivers. And those machines continue to run just fine on all other GPU projects.

http://www.gpugrid.net/forum_thread.php?id=3578
http://www.gpugrid.net/forum_thread.php?id=3539
 
Asteroids is loading longer work units. These will have credit reflecting the additional run time.
 
I have to preface this idea I just had with the fact that I'm not an IT person at all, and that computers and technology are a hobby for me, albeit a pretty serious hobby of mine. That being said, I was reading a great review of the new Silvermont Atom C2750 and noticed that it not only has low power, but fairly good performance and a fantastic price. Also saw on servethehome.com where they made a node of three of the embedded jobbies with some clever modifying of a BitFenix Prodigy case. This spoke to me, and not only is a clever way to get more cores on more projects (times may obviously be a bit slow, but power usage and cost would be fairly low), but it also led me to look at 3.5" SBC form factor PC's. Looks like AMD is looking at releasing steamroller core versions of their G embedded processor, as well as having their new Opteron X2100 series APU's (replete with steamroller I believe), that and such SBC form factor boards can also come complete with 2nd-4th generation i7's!!!! Asrock also is releasing a BayTrail version as well, which would again be more cost effective versus i7's.

Anyway, my point in this is that a server farm with either BayTrai SBC form factor boards (lots of them) or Atom C2750's could be a realizable goal to increase production without hitting the wallet or the power budget to hard! To be fair to my ignorant self, I've never ever known anything about the embedded computer market, so this discovery was like learning how to ride a bike all over again. Awesome.
 
Sounds interesting, but we would have to see the real world numbers. Currently ARM processors look appealing to low cost power consumption and I've heard arguments on both sides. Some people don't want a room dedicated to mini devices and some people don't believe that the devices are worth running since they take hours where traditional processors or even GPU's could take minutes. What you will also find when convincing people of this type of transition is the genre followers. WCG'ers were filled with GPU naysayers and had heavy resistance. Then when it finally got a project, guess what...those same naysayers were amazed and spouting their findings as if those findings weren't the same at practically every other GPU adopted project. Now, there are tons of Atom and ARM nay sayers. Especially in regards to using phones and tablets. At first it was... "they can't handle the heat" or "nobody will run it because the battery will die too quick" and "nobody will want to remember to run it while it is charging each night". Well the 19,000+ devices running Android goes to show that people certainly will do these things. The devices are just as powerful or more powerful than many devices still crunching WCG. I know my phone outperforms some P4 laptops I have ran. I still even have a couple P3 laptops running in hopes of getting BETA's for the badge.

So, if these Atoms deliver the PPD/power consumption ratios well enough with a desirable up front cost, I could see it being feasible. The problem is that these devices will probably not end up with the masses because they would have to be in products the masses are buying already. That is the success of DC'ing on ARM. Nobody buys a phone or tablet to DC on. The Raspberry Pi devices were for hobbyists and would never have gotten anywhere without phones and tablets.
 
Since Moo! sometimes has issues when it comes to boxes with multiple GPU's and being attached to multiple GPU capable projects, I posted a helpful tip in the BOINC configurations and How To's thread that should help. It could also be used for other reasons such as if you had an ATI 4350 that works at Collatz, Moo!, Primegrid, but doesn't work at POEM. http://hardforum.com/showthread.php?t=1729016
 
Plagiarism@home is sending out test units again. My Q8200 was ok but the Pentium 3's aren't liking it so much.

Edit: Well most of my machines are failing for now. Hopefully he can get it figured out because it would be nice to have more non-cpu intensive apps out there.
 
Last edited:
Well, I say we take a look at the specifics of just how applicable the performance is for the Avaton Atom as well as perhaps Kaveri when embedded or low power options start showing up. At least I think I'll take some for a test drive to see how good they do for our purposes. Folding? Likely not so much, but if the Atom C2750 can do as well as my Opteron 6344, then I think it'll be a good choice for low power computing. I'll take a look though and report back.

File Server

Because images and I don't really get along too well, I provided a nice link full of Virii! Kidding! But I put together a little file server for my business complete with 48 threads...
 
Last edited:
You must log in or register to reply here.
Back
Top