BOINC

Gilthanis said:
Distributed Data Mining just became my 25th BOINC project to reach 1 million points!
Click to expand...

Dang, I only have 10 projects with 1 million points or more. I have lots of catching up to do! :D

Congrats, Gilthanis! Very impressive!
 
I've done some crunching with F@H and have been giving thought to giving BOINC a try; even more so since noticing Muscular Dystrophy on Coleslaw's badge.

My little boy (just turned one three weeks ago) has Spinal Muscular Atrophy - Type II. For those who don't know what SMA is, because I sure didn't until about four months ago, it's a motor neuron disease similar to ALS but manifests in infants and young children. SMA is the number one genetic cause of death for infants. Even worse, it can sit "dormant" in children who seem fine for years; in fact some variants don't manifest until well into adulthood. (Exactly like ALS) Anyway, if you wish to know more you can visit many sites, such as http://www.curesma.org/ for more details.

Here's a bright spot though. SMA was listed by the National Institutes of Health as the neurologic disease closest to finding a cure. Remember how I mentioned that it's a Motor Neuron Disease like ALS? If they cure, or even make great progress with SMA, that information should be invaluable to ALS as well.

In the past I wasn't terribly particular about what project I crunched for and, as mentioned, usually did things with F@H when I did. At this time I'd like to focus on things that are more geared towards motor neuron diseases. The choices in BOINC are overwhelming and the whole thing is 100% foreign to me. Hoping someone can steer me in the right direction.

Thank you all for your time,
Jason
 
fastgeek, glad to have you on board. The disappointing news is that WCG finished the HCMD project. https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,33625

So, for now the closest things I would think of would be the protein research projects. Those are:

GPUGrid - Despite the name, they now periodically have CPU work including ARM.
POEM - Has GPU work but run low at times. CPU work rarely runs out. Please explain your hardware for potential headache prevention.
Rosetta - CPU only including ARM
CAS (runs out of work often) - CPU only
SIMAP (closing doors to BOINC at end of the year for in house research only) - CPU only including ARM
IGEM (just opened its doors so may have some starting blips) - CPU only
Najmanovich Research Group (currently down for server upgrades and technical issues) - CPU only
Volpex (but don't make that your first stop. I can explain more if you are curious) - CPU only

However, if you are interested more in DC'ing in general, let us know the hardware and we will give further guidance. We have various challenges we contribute to. Some span across multiple projects. Also, not all projects support GPU's. So, if you got any not being used, we could always point them elsewhere.
 
Last edited:
Thanks for your reply. :) I noticed POEM and am going to see if I can't throw some resources at it and maybe Rosetta and another. For the most part I'm CPU only. Have quite a few 4P E5-4xxx systems sitting around and even a couple new Haswell servers. The number comes and goes; but suffice to say there's a lot of hardware sitting idle where I'm sitting right now... might as well see if I can make some use of it.

BTW, what's the story with Volpex?

Just saw all your edits... I'll wait for your reply before doing anything else. :)
 
Volpex works different than most BOINC projects. They try to run their CPU work units in parallel with other systems. So, timing is everything and thus it is a learning curve compared to how the rest of the BOINC projects act and behave. It isn't a bad project. But, it does not have regular work. The good news is that you can connect to it and let it do its own thing when the time comes. I just don't want new BOINC'ers to go their expecting traditional work.

Your choices are good and both of those tend to have plenty of CPU work. They are also popular and have active project admins/moderators that update users. You will find that not all projects have that. I also want to stress that you will find that projects score differently. So, don't get confused if one projects is getting more points than another. Some people look at BOINC's over all world totals and then loose focus on the science. Don't let that discourage you.

POEM has GPU issues right now. Last I checked it was mostly for users with multiple nVidia cards in the same box. But, if you are CPU based, then you should have little trouble.

I also recommend you attach each machine to WUProp. It is non-CPU intensive so will run in the background not using very much resources and does not use up a CPU thread. It is also a project where you can get valuable information on resources and results at all of the projects. Give it a shot.

The biggest piece of advice I can give you is to also sign up for BAM! at BOINCStats.com and get familiar with the site. BAM is an account manager and makes it easy to manage multiple hosts and multiple projects from one simple location.
 
Heh, I kind of put 2+2 together after posting that you were indeed Mr. Slaw. :D I was looking at the guide earlier and am heading there again now. This one box has 80 cores and not all of them are being used... seem to recall something about a config file to fix that. (And my god am I happy to see this works with SOCKS!)
 
fastgeek said:
Thanks for your reply. :) I noticed POEM and am going to see if I can't throw some resources at it and maybe Rosetta and another. For the most part I'm CPU only. Have quite a few 4P E5-4xxx systems sitting around and even a couple new Haswell servers. The number comes and goes; but suffice to say there's a lot of hardware sitting idle where I'm sitting right now... might as well see if I can make some use of it.
Click to expand...

@fastgeek, sorry to hear about your son's diagnosis. You definitely came to the right place for suggestions on how best to contribute to BOINC.

If you have 4P E5-46xx servers "sitting around," that will be an excellent addition to any project. I made a good run with POEM CPU work units and it was stable for me. I wished their GPU project was as stable.
 
Hello brilong. The SMA diagnosis is pretty heart breaking. It's highly unlikely that he will ever crawl; let alone walk. That is, unless science comes up with a solution... then... who knows. Anyone who is a parent should be able to empathize a little. It's bittersweet seeing other kids his age starting to walk and to explore. His older brother (6.5 years apart to the day) has a pretty hard time with it too as he wants his little brother to run and play with him. I know it could be worse though. Much worse.

Gilthanis,

I'm having some issues with the 40/80 core 4P system. It'll start out @ 100%, or damn near, then it drops down to ~85% and stays there. No heat issues that I'm aware of and the server doesn't seem to be complaining. Did modify the cc_config file last week. Not sure WTF is going on.

Over the weekend the following were busily humming away:

E5 5680 (x2)
E5-4607 (x4)
E5-4640 (x4)
E5-4640 v2 (x4 this is the problematic system mentioned above)
E5-2660 v3 (x2)
E5-2690 v3 (x2)

All running 2008/2012 R2. Not sure how long all of these can stay on; but will do my best. Gilthanis mentioned a USB bootable image... will have to look into that. :)

Right now I'm sharing the resources available between POEM, WCG and iGEM. I honestly don't understand people who provide resources to things like prime and SETI. To each their own of course! While I like "competing" with people for points I simply don't understand why some projects are so successful for, which IMHO, are extremely unlike to yield any kind of real world use. Once again... strictly JMHO. Their computers, their bills, their kicks! :D
 
Well...Primes are used quite a bit in society. Mostly encryption. So realistically, those are putting out more real world results than the mass majority of projects out there. With that said though, I agree the chivalry and good feelings come from bio/medical sciences. Those debates can go on for hours. However, if you project selection doesn't use certain hardware and you got it sitting cold...might as well put it to use.

As far as your problem child, I'm not sure what the problem is without more info. When it starts to get idle cores/threads, are there any scheduler requests going on looking for work? Check the Event Log in BOINC. Feel free to post a good chunk of it in here so we can see what it is actually doing. I would prefer to see the first 30 lines from a clean start and then about 10-20 lines from when it starts or fails to request work with cold cores/threads.

Edit: also, check to see if those cold cores are tied up at all. I think Grandpa had an issue with CAS doing that. http://casathome.ihep.ac.cn/forum_thread.php?id=17571

Edit2: Try limiting the number of work units your system will request from each project to say 40 or something like that. Then see if you can keep it fed between the three projects. It may be something stupid like the CAS issue. I am interested in knowing though which project has the limitation. Perhaps suspending all projects but one to narrow it down? You can limit the number of units using a cc_config option or possibly even app config file.
 
Last edited:
fastgeek said:
I'm having some issues with the 40/80 core 4P system. It'll start out @ 100%, or damn near, then it drops down to ~85% and stays there. No heat issues that I'm aware of and the server doesn't seem to be complaining. Did modify the cc_config file last week. Not sure WTF is going on.
Click to expand...

Visit http://boinc.fzk.de/poem/prefs.php?subset=global and make sure you have something like this:
u7vb4.png


I seem to remember certain projects had "64" in the number of processors. I removed this and set it to 100% of the processors on the next line.
 
It would be odd for that setting to be overriding just one of his clients. If you change that setting, BOINC will eventually trickle it down to everything including other projects IIRC. Your manual settings within the client "should" supersede these. However, stranger things have happened.Can't hurt to give it a go.

Edit: also, this setting would override the hosts settings if changed just as the hosts settings should override those at the project if changed later.
 
Last edited:
Gilthanis said:
It would be odd for that setting to be overriding just one of his clients. If you change that setting, BOINC will eventually trickle it down to everything including other projects IIRC. Your manual settings within the client "should" supersede these. However, stranger things have happened.Can't hurt to give it a go.
Click to expand...

You're right, but he's trying 80 jobs and it might be limiting to 64. I specifically remember seeing 64 mentioned somewhere months ago and I removed it because I also have an 80-thread workhorse. :)
 
Doesn't hurt to try. It would be good to confirm. If that does end up being the case, I will add it to the guide section.
 
Gilthanis said:
Here is a good read and may give some insight on how the developers/donors are trying to level things a bit better.

http://boinc.berkeley.edu/trac/wiki/CreditGeneralized
Click to expand...

Hmmph while it sounds better than the current point system, it still does not address the fact that Intel CPU's get more credit than AMD CPU's do on some projects when in reality the AMD CPU's are faster than the Intel's are and should receive more credit than the Intels.

I am very happy to see he pretty much agrees that the ASIC miners do not belong in the same arena as CPU's and GPU's and there credit belongs in its own class, (not part of the mainstream Boinc). I believe he sees the handwriting on the wall that it will do some real damage to Boinc and his ability to get Grants.

He may actually be a little more savvy than his competitors. ;)
 
I'll check the CPU resource settings on the 80 core system for POEM. Know that when that box switched to iGEM the usage went up to 100%. Would be pretty embarrassing if I missed something so basic! :eek:
 
I added weak account keys for projects we had individual project threads for on the HardOCPtest account. That way if anyone wanted to contribute anonymously or wanted to test some work units without using their own accounts they could. Let me know if you need the account key for any other project via PM.
 
Okay, I'm disappointed... I looked throught the whole DC subforum, and see nothing about the new 970/980 video cards... My 750ti has been doing 100K ppd on Primegrid fairly easily, but I know the Maxwell core hadn't had much love yet on other projects - now that it's the new king of the hill, I expect that to change soon.

I'm going to be looking at any BF ads on 970's in about 2 months. Anyone else have Maxwell plans? Or even AMD plans, as those R9s should start dropping quick?
 
Jathanis said:
Okay, I'm disappointed... I looked throught the whole DC subforum, and see nothing about the new 970/980 video cards...
<snip>
Anyone else have Maxwell plans?
Click to expand...

I added the EVGA GTX 970 to my wish list on Amazon as they're currently out of stock. In my multi-GPU servers, the standard single fan cooler works better than the double or triple-fan setups. I'm not sure what others have experienced, but I can place two of them side-by-side without temp issues. As soon as I replace one with a double-fan unit, temps go through the roof (running GIMPS or BOINC GPU apps).

I was specifically looking at this model. What do you guys think?
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Those are designed to suck air through the card and out of the case where the others just blow the air off of the card and back into the case thus recycling hot air. That is why the one you are linking too are typically better. That is the route I would recommend.
 
Preliminary Change Log 7.4.22 -> 7.4.23


boinccmd: accept [IPv6_addr]: port as --hostname arg
API: Fix a bug in boinc_get_opencl_ids() which incorrectly returned CL_INVALID_DEVICE on some systems. This would happen on hosts with multiple GPUs if:
&#8226; the host has GPUs from two different vendors with different OpenCL platforms (e.g., NVIDIA and AMD), and
&#8226; the second platform queried has more GPUs than the first one, and
&#8226; the requested GPU is a higher numbed one in the second platform.
Mac installer: fix a bug which prevented localizing BOINC installer dialogs.
Mac Installer: If we have a valid code signing identity, code sign the BOINC client and Manager as well as the installer and uninstaller. OS X's software firewall can interfere with RPCs between the client and Manager. Signing them may make this less likely to be a problem.
client (Unix): when get vbox version, make sure it's not an error msg.
MGR: Fix a problem where OS X's software firewall can cause RPC authorization to fail.
Click to expand...

http://boinc.berkeley.edu/dev/forum_thread.php?id=8378&postid=56370

Mac only.
 
I didn't feel like making a whole guide just yet on BOINCStats. But, I figured this might be helpful to some. Especially for those having trouble with the many challenges we are doing from here. From http://http://boincstats.com/ , logon and in the upper right corner you will see a link called » Add widgets
This will let you add item to your personal view like in the image below.

boincstats_zps611b37e9.png
 
Last edited:
http://boinc.berkeley.edu/dev/forum_thread.php?id=8378&postid=57198#57198

Details can be found in the link above.

Windows and Mac have version 7.4.26 while Linux has 7.4.25

This one caught my eye:
client: If CUDA driver 6.5 or later is installed, prevent use of NVIDIA GPUs with Compute Capability < 2.0 and show explanation in Event Log and Notices.

client: let a MT job run even if it uses more than max # CPUs. Suppose the user fetches an 8-CPU job, then changes their prefs to use 6 CPUs. Let the job run anyway.

client: detect Windows 10
Click to expand...
 
Last edited:
Last edited:
BOINCStats pulled VGTU from the active list. I pulled it from the all inclusive list.
 
Does anyone here know if there is a limit to the number of projects one can have active in BOINC at a given time? Is there a limit to the number of projects BOINC can reasonably control? What is the largest number of projects you've ever had active at one time?

I'm betting Gilthanis can answer these questions, since he has run a very large number of projects.

I am asking simply out of curiosity at this point. I currently have BOINC loaded up with 47 active projects and am experiencing no real issues. The client seems to be scheduling work shares between projects correctly and every project that I do not have set to "No New Tasks" is getting work. I'm just wondering if I will eventually hit a wall as I continue to add projects? Will BOINC choke on me at some point?
 
RFGuy_KCCO, I do not believe there is an official limit. I do know that BOINC has troubles with its learning curve and when you get to a lot of projects it can have difficulties balancing run times. However, I have had pretty much all of the active projects at BOINCStats and a few others not listed there on a few machines. It seems to do OK if you aren't too worried about what projects are running and when. There may also be differences between the generations of the client, so I really can't remember if there was much issue with any specific one. I will say that people have reported having better experiences running more than one project at a time. Some say run times are better as well as stability. I have no idea on that one. I do know that running more than one project can sometimes conflict. Especially when you get into the virtualbox projects and multi-threaded projects.
 
Okay, since it's been a bit dead around here, I'll start the new year off with a few small upgrades - first, order just placed for an x5670 to replace my i7-950, so four more threads for BOINC. Second, I have a bid in for a 680 to pair up with my existing 680 - so I'll finally get to move into the 1M PPD club!

Once I get this done, I'm hoping to throw together a fourth box using my family room flat panel with another x5600 Xeon paired with a mobo that has room for a few more future video cards... But I wan't be able to add more GPUs until I get a good solar quote! :D Luckily for my power bill, I'll be removing my rock solid old 9800GTX+ to add the 680, so minimal power increase.

Of course, I'd also love to retire these two old C2Q systems and upgrade them to hex core Xeons as well - 24c/48t all at 4GHz+ would be a hell of a lot more fun than my current 12c/16t setup... All in good time.
 
Beal@home has been retired at BOINCStats as the website is now a gaming website.
 
Any recommendations for a budget graphics card? It's been cold in SE PA. I have a need for a space heater in my house. Primarily interested in BOINC BioMed projects (GPUGRID, Rosetta, POEM). It will be going into a low-power Ubuntu 12.04 box so a card without additional power plugs is desired. The 750ti looks to be the crowd favorite but the price is a bit higher than I want to spend. I would like to keep it under $75. Are there any BOINC Biomed projects out there that run on AMD cards? FAH is an option also. I have a used GT 210 sitting in a box. Not sure if it's too old to be of any use ...
 
Last edited:
Last I checked AMD cards still ran good at POEM. That is about it in the BOINC world for bio sciences. As far as nVidia, the 650Ti was recently on sale at Newegg, but it typically needs a power connector.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814127793 $90 - $30 MIR = $60 It does come with the power cable molex converter if that is needed....

There is another MSI 650 TI deal from Fry's if you have one nearby http://slickdeals.net/f/7590930-msi...-gddr5-video-card-37ar-ac-email-code-frys?v=1 but you would have had to use the email from their mailing list to get the awesome deal on it.

Then there is the current deal on the 750 Ti which is right at the $100 after MIR http://www.newegg.com/Product/Produ..._mmc=AFC-C8Junction-_-cables-_-na-_-na&cm_sp= $130 - $30 MIR = $100

I honestly wouldn't spend money on any nVidia cards less than the 650Ti. Many of the lesser cards have started to become obsolete. I still run them because I have them. I don't buy them unless they are super cheap. I'm not an AMD fan, so can't give a lot of suggestions on their stuff. Especially in that price range.

If you are willing to run other non-bio/medical projects on the AMD cards, then speak up as there certainly are. Reference my list here in the guides section: http://hardforum.com/showthread.php?t=1801908
 
Last edited:
Thanks Gil and brilong for the suggestions. I'll check out the 750ti. That seems like the best way to go.
 
You must log in or register to reply here.
Back
Top