BOINC 7.0.40 update heads up

[nanoprobe]

Just an FYI for something else for you guys to play with if interested. I've been experimenting with assigning CPU cores to a specific project. I started with an i7-860 running 4 cores, 3 on SN2S and 1 core on CEP2. Worked fine. Then I wondered if dropping an app_config file into other project folders (not WCG) would work. It does. The other 4 cores are now running 2 on Skynet POGS and 2 on SIMAP. So, if you wanted to run GPU tasks and a certain number of any CPU tasks it can be done. I have run into some cache problems running the different WCG and non WCG projects together but it's manageable and I'm still playing with it. Probably setting equal resource shares would be a good place to start but I'm still working on that.

[vitchilo]

Where do I put the app_config file again?

Because in BOINC, there's no project folders.

[OldChap]

<boinc>\data\projects\www. worldcommunitygrid. org

[Rob_B]

That could be a hidden folder.

[nanoprobe]

That could be a hidden folder.

Correct. Boinc folder which holds the project folders is hidden by default. You have to set your OS to show them. In Win7 click on start and the type hidden files and folders in the search pane at the bottom on the window. Hidden file and folders should now show at the top of the window, Click on it and select show hidden files and folders. While you're there scroll down a bit and remove the check mark from hide extensions of known file types. Hit apply, then OK and your done.
You'll now find the project folders at Computer>Local Disk C>ProgramData>BOINC>projects.

[OldChap]

When I install Boinc I find it much easier to put a folder called BOINC on c: then inside the BOINC folder I put another called BOINCDATA. If you point your install at these two folders all runs smoothly and of course nothing is hidden.

Easy to find everything and also makes it easier to copy say to a stick when re- installing os etc. Good thing for folks running for a short period like SC week too

[pirogue]

When I install Boinc I find it much easier to put a folder called BOINC on c: then inside the BOINC folder I put another called BOINCDATA. If you point your install at these two folders all runs smoothly and of course nothing is hidden.

Easy to find everything and also makes it easier to copy say to a stick when re- installing os etc. Good thing for folks running for a short period like SC week too

That's pretty much how I've always done it. It makes the whole Windows 7 program files+program data folder business much easier to deal with. If you have a D: or E: drive to install to, it also helps keep the churning off of the C: drive.

[werdwerdus]

I always make my data folder D:\BOINC\ (whatever drive letter). I don't change the program folder since we don't ever need to do anything in there and you can migrate by just copying the data folder to a new installation.

[pirogue]

Having everything under x:\BOINC makes the migration *slightly* easier because you only have to worry about copying 1 folder.

[my87csx481]

Currently testing 7.0.42 with the new app_config.xml. Works great and it's way easier to configure. You don't lose all your cache if you take it out of the project folder and you don't have to have tasks in cache to add it to the project folder. If you'd like to run GPU and CPU tasks you can assign 1 CPU to each project. IE you could run 1 CEP2, 1 GFAM, 1 DSFL etc, or any combination of GPU/CPU tasks you'd like to run.
WARNING: If you're currently using an app_info file I would suggest you run through all the tasks your in your cache before you make the switch because when you switch from app_info to app_config you will lose them all. The old problem of losing them when you remove the app_info file can't be stopped AFAIK.


Here's what running 10 GPU tasks looks like.
<app_config>
<app>
<name>hcc1</name>
<max_concurrent>10</max_concurrent>
<gpu_versions>
<gpu_usage>.10</gpu_usage>
<cpu_usage>.80</cpu_usage>
</gpu_versions>
</app>
</app_config>

The max_current is the amount of tasks you want to run
The gpu_usage replaces the <count> tag from the app_info
The cpu_usage replaces the <avg_ncpus> and the <max_ncpus> tags from the app_info

I haven't tried it but to run CPU tasks I think it would look like this for each project you want to run.
<app_config>
<app>
<name>name</name>
<max_concurrent>N</max_concurrent>
</app>
</app_config>

If you want to run both just add this to the GPU code above.
<app>
<name>name</name>
<max_concurrent>N</max_concurrent>
</app>

Make sure you name the file app_config.xml and it still goes into the WCG project folder.

Have fun-crunch on.

I may be able to make some time this weekend to play around with this. Where can I find the correct project names to be used in place of "name" in the cpu section of the config file (bolded in the quote)?

[OldChap]

Saw this but have not yet tried it:

<app_config>
<app>
<name>hcc1</name>
<gpu_versions>
<gpu_usage>.125</gpu_usage>
<cpu_usage>.25</cpu_usage>
</gpu_versions>
</app>
<app>
<name>dsfl</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>dddt2</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>sn2s</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>gfam</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>c4cw</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>faah</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>hpf2</name>
<max_concurrent>6</max_concurrent>
</app>
<app>
<name>beta</name>
<max_concurrent>8</max_concurrent>
<gpu_versions>
<gpu_usage>.125</gpu_usage>
<cpu_usage>.25</cpu_usage>
</gpu_versions>
</app>
</app_config>

[nanoprobe]

OC. Those are all the correct project names for WCG. One little quirk I have found is that it seems like you have to have downloaded at least one of each task before you try to use the config file to set the cores you want to run for a particular project. I didn't notice that before I set my config file to run 3 SN2S and 1 CEP2 task because I downloaded some tasks first. When I added C4CW to the config file I got a message in the event viewer that said World Community Grid | app c4cw not found in app_config.xml. It wouldn't download any C4CW tasks until I removed the code from the config file. After I removed it and restarted boinc it downloaded some c4cw tasks. I then shut down boinc and put the code back into the config file and all was fine. Maybe I botched something on this so if someone else tries it let me know your findings.

[my87csx481]

I will make some time to try using app_config....possibly even this evening.

The rig is a 990x with HT on (so 12 threads) with a 7970 currently running 12 threads. No cpu tasks except the cpu portion of the hcc gpu tasks. I would like to utilize the cpu's "idle" time with other projects. Do I need to tick those other projects in the WCG profile the machine is using, or not? Or should I ask: does app_config override the profile settings?

I have other questions but, they will probably be answered by the answers to the above.

[OldChap]

I wonder what will happen with Beta's then?

Those are the reason I need to have cpu running at least one core else I will never get my badges.

[nanoprobe]

I wonder what will happen with Beta's then?

Those are the reason I need to have cpu running at least one core else I will never get my badges.

Good question. I have beta already added to my config file and I get the same message I got for C4CW. World Community Grid | app beta not found in app_config.xml

[my87csx481]

Something tells me we're out of luck on the betas....and maybe dddt2 too....unless you can catch 'em when they're coming and disable app_config to get one, then re-enable.

I'm getting work units based on my profile's settings. Getting red warnings in event log:

1/4/2013 9:34:51 PM | World Community Grid | app dddt2 not found in app_config.xml
1/4/2013 9:34:51 PM | World Community Grid | app gfam not found in app_config.xml
1/4/2013 9:34:51 PM | World Community Grid | app beta not found in app_config.xml


I'm using the following (not yet correct) app_config:

<app_config>
<app>
<name>hcc1</name>
<max_concurrent>12</max_concurrent>
<gpu_versions>
<gpu_usage>.0833</gpu_usage>
<cpu_usage>.8</cpu_usage>
</gpu_versions>
</app>

<app>
<name>dddt2</name>
<max_concurrent>6</max_concurrent>
</app>

<app>
<name>gfam</name>
<max_concurrent>2</max_concurrent>
</app>

<app>
<name>beta</name>
<max_concurrent>4</max_concurrent>
<gpu_versions>
<gpu_usage>.083333</gpu_usage>
<cpu_usage>.8</cpu_usage>
</gpu_versions>
</app>

</app_config>

My desire is to run 12 hcc gpu threads with 8 cpu threads supporting them, and run a couple cpu threads of dddt2, gfam, and maybe some betas. I do not want hcc cpu work units but, I'm getting them sent to me. How do I stop them?

Edit: I didn't do anything but hit update and the gfam error went away and I'm getting gfam work units.

[OldChap]

I will add a couple tomorrow.

I mentioned this in the optimizing thread.... I was watching to see what the maximum cpu load was for around 10 mins tonight whilst my 7950 was running. It topped out at 82% for a second or two which seems to indicate to me that it may be possible for me to run a thread or two cpu and still not adversely effect GPU crunching

Gpu is getting higher priority than normal work units in the same way Beta's used to which leads me to wonder how beta priority is against GPU

[nanoprobe]

Something tells me we're out of luck on the betas....and maybe dddt2 too....unless you can catch 'em when they're coming and disable app_config to get one, then re-enable.

I'm getting work units based on my profile's settings. Getting red warnings in event log:

1/4/2013 9:34:51 PM | World Community Grid | app dddt2 not found in app_config.xml
1/4/2013 9:34:51 PM | World Community Grid | app gfam not found in app_config.xml
1/4/2013 9:34:51 PM | World Community Grid | app beta not found in app_config.xml



Edit: I didn't do anything but hit update and the gfam error went away and I'm getting gfam work units.

That's encouraging to hear. Maybe you don't have to have tasks in cache before you add the project to the app_config file.

[Bun-Bun]

What about discerning Nvidia from ATI?

Two scenario's.

1. Nvidia and ATI GPU's in the same rig crunching. You want multi GPU's for the ATI cards but maybe only single or dual threads for the Nvidia cards.

2. Nvidia and ATI GPU's in the same rig with only ATI crunching. I have a GTX670 in my media center that I use for gaming, watching movies, web browsing etc. I threw a 7850 in there to crunch 8 threads leave the GTX670 free to do everything else. This way the system doesn't lag at all and I can still crunch. Using app_info this is no problem. I don't see a way to do this with app_config though.

[OldChap]

1. is possible on the app_config. I have 16 and 3 running but I think the proviso may be which card Boinc/wcg sees as the primary card

2. may work too because you can state use one card in config. Not sure how to choose which though.

[Bun-Bun]

Further investigation requred then.

I updated my main rig (7950) to 7.0.42 just now. It is downloading CPU tasks with GPU tasks even though only GPU tasks are running. What are others doing about this?

[jeanguy2]

Welcome back Bun-Bun,

If you do not want CPU tasks just go to the WCG web device manager for that rig and select "No CPU Task"

[Bun-Bun]

When I do that I get no tasks.

[OldChap]

The key bits are highlighted:



If this does not get you going then we need to look deeper.

[nanoprobe]

What about discerning Nvidia from ATI?

Two scenario's.

1. Nvidia and ATI GPU's in the same rig crunching. You want multi GPU's for the ATI cards but maybe only single or dual threads for the Nvidia cards.

2. Nvidia and ATI GPU's in the same rig with only ATI crunching. I have a GTX670 in my media center that I use for gaming, watching movies, web browsing etc. I threw a 7850 in there to crunch 8 threads leave the GTX670 free to do everything else. This way the system doesn't lag at all and I can still crunch. Using app_info this is no problem. I don't see a way to do this with app_config though.

For scenario # 2: If you want to exclude running your GTX670 and let the 7850 crunch the WCG HCC GPU tasks (which I would highly recommend) then add this code to your cc_config.xml file which is in the BOINC folder, not the project folder.
<exclude_gpu>
<url>project_www.worldcommunitygrid.org</url>
<device_num>N</device_num>
<type>nvidia</type>
</exclude_gpu>

Device N is which ever the nvidia gpu is, 0 or 1.

If you don't have a cc_config.xml file already in the BOINC folder then just copy the code below into notepad, rename it cc_config.xml and add it to the BOINC folder. Just remember to give the N tag a number for your nvidia card. You'll have to shut down and restart BOINC for the changes to take effect.

<cc_config>
<options>
<exclude_gpu>
<url>project_www.worldcommunitygrid.org</url>
<device_num>N</device_num>
<type>nvidia</type>
</exclude_gpu>
</options>
</cc_config>