• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

WCG/boinc (not only): help needed

ChristianVirtual

ChristianVirtual

[H]ard DCOTM x3
Joined
Feb 23, 2013
Messages
2,561
Help !

I try to run WCG on a hex core 1650v2 / CentOS and have it sofar running.

But I get a notice from BOINC that my current settings do not allow tasks from this project. To fix I can change project preferences.

Now admitting that I'm less knowledgable person on this earth with respect to BOINC and WCG: what / where I have to do something.

Second question: any good guidance to connect to BOINC client via SSH tunnel ?

Thanks
CV
 
Last edited:
wait, might be false flag and I do some mapping cancer marker already with 13% ...

Captain, watch me that I'm in the right team ...
 
next question: seem only using one core the the 6c/12t I have in that box -> how to get the other one working too ? Or is WCG one-threaded mainly and i need to assign more tasks
 
This page will allow you to choose the apps in WCG:
https://www.worldcommunitygrid.org/ms/viewMyProjects.do

There is a Device Manager too linked on that page for setting up different apps for separate systems:
https://www.worldcommunitygrid.org/ms/device/viewDevices.do

The WCG team seems to be named HardOCP. None one with the [H] in the name.
http://stats.free-dc.org/stats.php?page=team&proj=bwcg&team=1411
https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=BP5XNJBR9N1

I see you on the correct team:
https://www.worldcommunitygrid.org/stat/viewMemberInfo.do?userName=ChristianVirtual

I would make sure its set to use 100% of the CPU cores, 100% of the time and to crunch while in use. There is an option for all 3.

It will download more units for all the other cores to work on depending on the queue allowed. It might get one the 1st time and download more a bit later. Looks like you've returned 8 units already.
 
mmonnin said:
Looks like you've returned 8 units already.
Click to expand...

Last Result Returned (UTC)
spacer.gif
8/26/14 14:07:33 [22,765+ hour(s) ago :oops:
 
ok; now we are talking ... and all cores are sweating ... :)

Screen Shot 2017-04-01 at 13.02.48.png


Now I need to figure out how to get my iOS monitor being able to read BOINC client ...
 
based on the wingman concept what is a good way to reduce the queue / what is a fair size of the queue ... don't want to pile up assignments which take more time and a wingman suffer from delayed validation ?
 
another question on the credits:

when I see the credits it seems that the result from my wingman and myself get added and then 50:50 distributed. Is that right ? Nothing like QRB in WCG/boinc ?

Is there a way to predict the credits I get create ? In FAH term there are project definitions with credits and with the TPF I can calculate. How is the credit for WCG defined ?
 
Some projects have quorum or 1 and some have more. It's just a part of crunching. You're a new computer so everything might get validated at WCG for a bit not eventually you'll complete enough work to 'pass the test' of stability and you won't need validation at WCG. I wouldn't worry about making your own queue set to reduce validation time.

Each get the same, full credit. There is no QRB in WCG or really in BOINC. There are a couple projects or apps that give double credit it returned within a day (GPUGrid) and some apps that give a bonus if it uses a lot of memory or finds a prime. Like 10% bonus.

Some BOINC projects have a set Point per WU so its always the same. Others use this system called CreditNew which uses your FLOPs and completetion time to grant credit. It's on its 3rd revision and there are talks about a 4th. WCG is supposed to be something like 25 credits an hour per CPU. I think its done in Windows as the VINA apps get more credit in Linux as VINA based apps run better in Linux.
https://boinc.berkeley.edu/trac/wiki/CreditNew

WCG was around for awhile before it joined BOINC so it has a point system already setup and produced more points than the typical BOINC CPU project. So points on the WCG site are 7x what get reported to BOINC stats sites. An FYI if you notice the difference.
 
Sorry ChristianVirtual, I was on the road all day yesterday at work and went straight to the in-laws after. Thanks mmonnin for chiming in.

To answer the question on que, I typically run mine at 0 so I don't have any more work in there then what I need. That way if a system goes down, then I don't have work units missing deadline. Unless there is a challenge you are bunkering for or there is a shortage of work units for the sub-project you are focused on, there really isn't a need for most to have a cache. However, if you are limiting your data transfers to a certain point of the day, then a cache is probably necessary. You can set that in the BOINC client itself. With BOINC open in the advanced view. Click Options > computing preferences > Computing tab > (at the bottom under Other) change store at least X days of work > change store up to an additional X days of work. To run with no cache of course you would just put 0 and your client will pull work as it needs.

Our team's page is here: https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=BP5XNJBR9N1

If anyone else is joining please create an account from this link: http://join.worldcommunitygrid.org/?recruiterId=338542&teamId=BP5XNJBR9N1 It will assign you to our team automatically as part of the sign up process.

Also keep in mind there is Pending Verification and Pending Validation. New users can get confused on them. Validation is more for validating the computer is reliable (but still running good work units) and the other is verifying the work unit processed by reliable hosts. Pending validations can occur on any project regardless if they normally have wingmen or not. Pending verification is only on projects that need a double checker ... aka wingman.
 
The credits can be confusing as mmonnin explained. Rest assured though that faster machines still earn more than slower machines. Just keep in mind WCG does not have optimized apps so newer tech like AVX doesn't really come into play much. So, AMD competes fine with Intel in terms of points.
 
Oh...and I've never messed with SSH tunnels, so I cannot help you there. Perhaps someone else on the forums can chime in?
 
no problem, all running right now

But my last question for today:

For whatever reasons I just got myself a hosted server with GTX 1080 and a 6700 Intel ...

The learning curve was steep and new system was up and running in no time with mix of WCG and GPUgrid ; now based on FAH experience its good to leave one core/thread per GPU. How is that in BOINC ? Right now I run the GTX on GPUGrid one full load on WCG.

Screen Shot 2017-04-01 at 23.50.56.png


Good idea or better reduce the load from WCG in this case ?
 
Gilthanis said:
Oh...and I've never messed with SSH tunnels, so I cannot help you there. Perhaps someone else on the forums can chime in?
Click to expand...

I guess I will figure that out myself as I do that with FAH already; can't be that different (right now I'm half-protected with firewall and fixed IP); but eventually SSH tunnel will be the way into my remote boxes
 
The # of CPU cores to leave open for GPU projects depends on the projects I have found. Some of it depends on OpenCL vs CUDA. Amicable numbers I ran several GPU tasks per GPU, you can do that btw, and still loaded up 8 tasks on my 3770k. They used very little CPU usage and took the CPU usage they required with the CPU tasks taking the rest. Einstein @ Home is another story. I can only use 4 CPU cores while running 2x E@H tasks per 2 GPU. I still do run 8x tasks on 4 CPU cores by using process lasso. E@H doesn't fully utilize the GPU but 2x at once puts it much closer to 100% all the time.

GPUGrid benefits from a full CPU core I believe along with setting an environment variable swan_sync = 0. For myself, I run as many tasks as I can w/o sacrificing GPU utilization. GPU Projects can set it so that BM will automatically limit the CPU threads to limit for GPU work. In your screenshot I can see 0.989 GPU and I bet the rest says 1 NV GPU after that. Some will say 1 CPU + 1 GPU which results in 1 less CPU thread available for CPU work. This can be adjusted with an app_config.xml file along with how many GPU tasks per GPU. One task per GPU is preferred with GPUGrid due to the time bonus. It has has a very small queue, like 3 (?) per system even with 2 GPUs.
http://www.gpugrid.net/forum_thread.php?id=2123

Most people use a queue. No wasted time waiting for a download. Even if the download is quick its wasted time. Not all BOINC projects have the best uptime and there is always server maintenance. ATM I run WCG on a 3570k but I am trying to get my last badge, Help Stop TB. Work is hard to get without manual intervention while keeping CPUs busy. I get a queue of work in another WCG app (like 0.5 day), limit it to HSTB only and increase the queue so WCG thinks I need work. I am still crunching and get the work I want. I've gotten 1 HSTB WU over the last week w/o doing this and 18 since last night.

If you're wanting to monitor the BOINC Manager remotely, I use a Program called BOINCTasks. It's like HFM (multiple computer monitoring) with with all of the control of the actual BOINC manager. I actually don't have BM open on my main computer. I just use BOINC Tasks.

Theres a lot of info here about app_config, cc_config files and how you can interact with boinc via command line.
https://boinc.berkeley.edu/wiki/Client_configuration

The options here can make BOINC very configurable. There is no slot options for a piece of hardware like FAH but you can make it come close.

BOINC is available other types of CPUs. No PS3 but Android devices, Raspberry Pi's, miners and Odroids are some examples. I've ran WCG on 2 tablets if you have some that can be left plugged into a power source.
 
I think Yoyo@home still supports PS3 if you still have an old non-updated one with Linux on it. Limited to Ogr work I believe. And WUProp has an app but would need another project in order to be of any benefit.

Some helpful links for BOINC here in our forums are:
https://hardforum.com/threads/boinc-useful-tools-and-download-locations.1768577/
https://hardforum.com/threads/boinc-configurations-and-how-tos.1729016/
https://hardforum.com/threads/boinctasks-utility-use-and-setup.1852695/
https://hardforum.com/threads/all-inclusive-dc-project-list-v-2.1895029/

And depending on the topic, maybe a few others.

As far as cpu threads to feed the GPU's, like mmonnin said it will depend on project if you are maximizing things. A general rule across the board would be to leave a full thread per GPU and then you wont have GPU's waiting to be fed.
 
Last edited:
I do not believe so. Newer PS3's cannot run because SONY pulled the alternative OS option during a system update and newer models never had the option. Not to mention the power consumption just wouldn't be worth it.
 
I didn't even know it could run BOINC at all. I only knew of the FAHClient many years ago. And it was pretty good for awhile. The latest gen game consoles are the most they've been like regular PCs.
 
what is a bit disappointing in BOINC/WCG is the (un)speed of update stats. Make lots of small WU on SCC, even with quorum 1 and they get not really timely updated. Somehow on FAH at least once every hours (or whenever done) I get updated stats files. or 3 hours on the regular stat sites. need to slow down on that one ...

plus the status sites for BOINC are quite overloaded with information not really sure is useful. Still some time to adopt needed.
 
Seems I have two CPID

One for WCG, and for GPUGrid/SETI/Einstein etc

Bad luck or not hopeless ? Is there a way to link them together ? Or can I abandon my WCG account and make a new one with same name and different CPID ?

It might happen because I used different emails when I signed in few years back
 
Did some reading and want to confirm: if I align my email Adresses between the projects it will resync and merge under one CPID ?
 
ChristianVirtual said:
Did some reading and want to confirm: if I align my email Adresses between the projects it will resync and merge under one CPID ?
Click to expand...
I've heard its best to do crunch projects with both IDs on the same computer so that they merge but it may take awhile.
 
Best practice is to use the same email on all accounts AND to run multiple projects on at least one system so that it learns to associate them all to the same user and thus merging to one CPID. It is easy to do, but just takes a little bit of time for things to align. Not sure if changing the email address will cause any hiccups but should be done.
 
Ok, thanks guys. As I'm as the start of a BOINC carrier (partial) not much "damage" can be done. It not yet like with my two folding entities. Something for Saturday ...
 
Is there a way to remove old and invalid IP addresses from the BOINCManager dropdown box. I had some EC2 spot instances with temporary IPs. Now that those systems are down again the IPs are not longer needed. Is there a list I can remove those from ?
 
I have the same issue. Don't know how to get rid of them, but I hardly use that manager now. I use BoincTasks instead. You can control and view all the boxen at once.
 
If you highlight them and hit the delete key, does it not remove them like in a web browser?
 
Thought I tried, but my mac is now sleeping, don't want to wake him up. Takes always so much time to make him sleep again
Will try tomorrow once more.
 
Nice... I had never attempted to remove them before...good to know.
 
Next question I need some hints:

when I run one GPUGrid WU and added collatz another WU got assigned on the same GPU card. How can I prevent such "overload" and force a more sequential processing. I understand that I could configure a GPU to accept to assignments and force to higher utilization. But how can I prevent that ?
Screen Shot 2017-06-29 at 21.59.49.png
 
What you are saying is that it paused your GPUGrid work unit which you did not want to happen due to the QRB...right? That is the problem with QRB projects and supporting multiple projects. BOINC will try to balance resource shares. Since you have been running GPUGrid for a while, BOINC Manager will try to run other GPU work that you have allowed to balance that debt. It would be better to set Collatz to resource share 0 and then set GPUGrid to No new tasks so that it will run out of GPUGrid work. What will in essence happen is that since Collatz is priority zero, it will not ask for work unless no work is sent from other GPU projects. Then it will send enough work to keep the GPU busy as it continues to try and pull work from the other projects. But since you have set the project to No New Tasks, it will not pull any and just continue to pull from Collatz.
 
You must log in or register to reply here.
Back
Top