Hi guys, I wonder if some one could give me a bit of a pointer?
I have both an Nvidia & an ATI garphic card in each of my box's, now what i'm trying to do is select only 1 x GPU (the ATI) to crunch Help Conquer Cancer GPU work units at WCG. The other card(Nvidia) will crunch something it's better suited for.
With every other project that supports both GPU types there's an option to select one or the other or both on the projects home page, but not at WCG.
You have to set up a custom file (cc_config.xml) to do the job. To try & cut a long story short i've set up the file & while it seems to work, in reality it does not. At least not for me..
So I was wondering if someone could see what i've done wrong? This is the cc_config I created..
<cc_config>
<log_flags>
</log_flags>
<options>
<exclude_gpu>
<url>http://www.worldcommunitygrid.org/</url>
<type>nvidia</type>
</exclude_gpu>
<skip_cpu_benchmarks>1</skip_cpu_benchmarks>
<report_results_immediately>1</report_results_immediately>
</options>
</cc_config>
It seems to work OK as the "report_results_immediately" flag works & shows in the message logs OK. The "exclude_gpu" flag comes up in the log as well an seems OK also. But it not working the way it should. These are the lines from the log.
" 20 16-12-2012 07:28 PM Config: report completed tasks immediately P6T-PC"
" 21 World Community Grid 16-12-2012 07:28 PM Config: excluded GPU. Type: nvidia. App: all. Device: all P6T-PC"
All seems sweet but yet I still recieve both ATI & Nvidia WU's. Please help as the only help I now find on WCG is a link to the cc_config.xml options set-up help page.
Thanks in advance JN
Your cc_config.xml looks alright.
You could try using an app_info.xml file with only the ati application details and no other WCG application or task types should be downloaded. If you wanted to crunch other WCG CPU projects you would need to add them to your app_info. You have to download and save the relevant application files first though because WCG has no publicly accessible ftp download site like many other projects.
This is just a possibility I thought of and something you could try, you don't have to run more than 1 concurrent GPU task with an app_info.xml file if you don't wish to. Hope you get it sorted, mixed GPU brand configurations have been a challenge on many GPU projects.
If it works it's only a temporary solution. They intend to stop the use of app_info.xml at WCG and have been working with BOINC developers to have a new type of file called an app_config.xml instead.
Didn't need to be Nostradamus to see they were heading for trouble at WCG ages ago trying to get their old BOINC version to work reliably for GPU processing and work fetch but when I suggested people would be better off using a current BOINC recommended BOINC version for GPU I was attacked, so I left them to it to make their own mistakes and they've been trying to get it to work reliably with decent CPU and GPU project choice user selections ever since. Looks very unlikely they'll ever have user configurable parameters for target frequency, chunk size, etc to suit the varying performance potential of different GPU models.
More efficient for cancer research overall, but similar to POEM GPU basically their GPU app is a real stinker in terms of using all your CPU cores. I mean doing more research overall is great but if GPU contributors use all their CPU resources to support GPU processing it means nothing left for other WCG or other project CPU applications. That just seems contrary to the spirit of BOINC in a way. I'm not having a go at anyone that chooses to do that, I did it myself for a while at POEM, it's only that "it's BOINC Jim, but not as we know it, not as we know it, not as we know it." ;D
A thought JN. Try your setup on another Nvidia/ATI project, maybe Milky Way and see if you can get it working there.
Thanks guy's..
@ Kashi: Thanks for giving it the once over I was hoping you'd give it look. Yea it looks like an app_info is the only way to go. I was trying to avoid using one cause they are more complex than other projects app_info's especially if you decide to crunch some CPU work as well for a while here & there, the constant changing is a pain & as you say I read there thinking of pulling the anonymous platform anyway. While I was over there looking around I also noticed info about the app_config.xml. Sound interesting but at a quick glance looks nowhere near as flexable as an app_info & only available in 7.0.40 + beta versions of boinc, but I'll have to look into more, it was the first i'd heard off it.
I've had a couple of problems running both ATI & Nvidia before but overall I find it much more flexible way to do multiple projects with limited resources. And so far the good outweighs the bad, but time will tell.
And yea WCG can be a bit "my way or the highway" which is why I posted here first.
@ Lawry: Thanks for the suggestion mate, I may have to crunch elsewhere until either I figure out an easy fix, or WCG gets there act together, but ATM i've got the new box crunching a few WU's for the time being minus the Nvidia card. I'll throw the nvidia card in a bit later. I may just grab a milestone & move on.
Thanks again JN
As long as you have a basic familiarity with app_info.xml files it's not too bad actually. Once you download the necessary application files first and then get your app_info.xml installed and working then it works much like other projects. Unless WCG change the applications that is. >:D
I ran both HCC GPU and CPU a while ago until I got my ruby HCC badge because I thought they may run out of work soon. Didn't like running more than 3 or 4 concurrent HCC tasks on my HD 5870, so ran some HCC CPU as well to get the days up to 180 a bit quicker. I found the HCC GPU app_info.xml somewhere and kludged up the HCC CPU part from the client_state.xml file and added it. Then someone put it up on xtremesystems, so I used their one instead. It was almost the same as mine just in a different order. The <executable/> they have put after <name>hcckernel.cl.7.05</name> in the AMD/ATI GPU section is unnecessary but it works either way.
I just had another look there and all the CPU projects are now included in a big app_info file. Just delete the green NVIDIA section and any CPU projects you don't want. They have set the <count> value to 0.125 so that will run 8 concurrent GPU tasks. If you want less or more than that you will need to adjust the value. There are also instructions there for how to run 10 concurrent GPU tasks on an 8 core computer if your card is up to it. Possibly not any more efficient than running 5 or 6 concurrent tasks but it does get your badges quicker. ;D
http://www.xtremesystems.org/forums/showthread.php?283509-Working-app_info-files
This message will self-destruct in five seconds. Good luck, Jim. :boom:
Is there some trick to getting GPU WU's from WCG ?
I tried last night but got nothing but CPU ones....
Only ticked for work from HCC... are they out of GPU work maybe...? When I attached to WCG on my smaller machine (with an ATI card), it downloaded the GPU app..but didnt get any work.
Would like to continue running POGS on my CPU's but get WCG moving again on a couple of vid cards..
Greetings,
@JN, sorry to hear about your card troubles, cannot really add anything, as I have only a single card setup.
On another note though, I went back to WCG to do some GPU crunching, tried the app_info.xml but it got knocked back, so I scratched around the WCG forums and found out that the latest build of BOINC allows you to use what is called a app_config.xml and that is what I am running
Boinc 7.0.42
app_config.xml
Now am successfully running WCG GPU x 2 units at the same time on run AMD HD 5850 1GB board
But alas, I did also pull down a lot of HCC CPU tasks as well, it has been noted on the WCG forums numerous times
Heres the link to the BOINC developers forum about app_config.xml
http://boinc.berkeley.edu/dev/forum_thread.php?id=8051
Regards
tazzduke
Veebee have you ticked the box at the bottom of the My Projects page where you request gpu work for projects when this is available.
@Tazzduke. Thanks for that info, most interesting, time to PLAY!!
Karma too.
Afternoon All
Apparently they are playing around with this over at MOO as well
Once I have emptied my cache of all GPU workunits, (I have MOO, WCG and Seti Beta), I am going to do more tinkering with it, you create this file and put it the respective project directory of the boinc folder.
Till then keep on crunching and anyone else has done some trials, feedback here would be helpful as well, especially any hiccups you have encountered
Regards
Tazzduke
P.S. dont forget to do a full shutdown of boinc client, make the file and then restart boinc
Quote from: tazzduke on December 18, 2012, 08:27:19 AM
Greetings,
@JN, sorry to hear about your card troubles, cannot really add anything, as I have only a single card setup.
On another note though, I went back to WCG to do some GPU crunching, tried the app_info.xml but it got knocked back, so I scratched around the WCG forums and found out that the latest build of BOINC allows you to use what is called a app_config.xml and that is what I am running
Boinc 7.0.42
app_config.xml
Now am successfully running WCG GPU x 2 units at the same time on run AMD HD 5850 1GB board
But alas, I did also pull down a lot of HCC CPU tasks as well, it has been noted on the WCG forums numerous times
Heres the link to the BOINC developers forum about app_config.xml
http://boinc.berkeley.edu/dev/forum_thread.php?id=8051
Regards
tazzduke
Hey tazzduke thanks for the pointer to app_config.xml - looks like t will be extremely useful and way less painful than app_info.xml
If anyone has already updated to BOINC 7.0.40 or higher, these newer BOINC versions use the new app_config.xml, so you can't use app_info.xml on WCG with them.
BTW BOINC 7.0.40 has a bug that overwrites the settings in the app_config.xml when new work is downloaded. If you want to use app_config.xml it is advisable to use BOINC 7.0.42.
My advice to JugNut to try an app_info.xml was specifically for his problem with receiving unwanted HCC NVIDIA GPU tasks. For anyone else who doesn't have that problem and who wants to try GPU processing at WCG the new app_config.xml is a much simpler and easier way to configure multiple concurrent WCG HCC GPU tasks.
Quote from: kashi on December 17, 2012, 05:31:39 AM
their GPU app is a real stinker in terms of using all your CPU cores
Amen to that. Works fine on my old 32bit clunker but only 1 WU at a time (No cores left).
However, on my quad 64 bit machine running Win 8 it is a real clusterf... Keeps jumping from WU to WU after a few percent of processing. So I changed the app_config to process 1 WU at a time. Restarted and rebooted and damn me if it still isn't doing 2 WUs at a time.
I may have issues with this machine and Win 8 and they are causing weirdo behaviour. I'll rever to Win 7 and see what happens.
But the most annoying thing is the CPU WUs that keep downloading. This is the app_config file I am using if anyone can see a glaring mistake please let me know
<app_config>
<app>
<name>hcc1</name>
<user_friendly_name>Help Conquer Cancer</user_friendly_name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>0.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
Using BOINC 7.0.42
:hbang: :hbang: :hbang:
EDIT: Added screen shot of Afterburner
[attachment deleted by admin]
Hi Lawry, that's a
similar one that i'm using at the moment except for theses lines...
<app_config>
<app>
<name>hcc1</name>
<user_friendly_name>Help Conquer Cancer</user_friendly_name>
<gpu_versions>
<gpu_usage>
0.125</gpu_usage>
<cpu_usage>
0.4</cpu_usage>
</gpu_versions>
</app>
</app_config>
The <gpu_usage> of 0.125 in mine = 8 wu's at once are crunching on my GPU. Just like app_info's "count". So 0.50 = 2, 0.33=3, 0.25=4, .2=5, 0.1=10 ect
You have your "gpu_usage" at 1.0 so, only one 1 WU per GPU.
The <cpu_usage> of 0.4 in my settings means it's using about 0.4 of a CPU core per WU, so 0.4 x 8 WU's=3.2 cores used(actually uses bout 4 cores).
So if you wanted to use say a quarter of a CPU core for each WU it's 0.25 CPU, likewise 0.5 = half core per WU ect.
Also you have yours using 1 GPU WU per "0" cores. I think it's kinda freaking it out a bit.
@ Veebee: You can turn off CPU crunching by going in "Device Profiles" on the website & make sure to make "home" profile or whatever you have is set as default & you have "Allow research to run on my CPU?" = NO. Also turn off "If there is no work available for my computer for the projects I have selected above, please send me work from another project." is unticked. This worked for me anyway to disable CPU work.
Hope this helps it's working on mine JN
PS sorry i'm not very good at explaining stuff.
@JN. Well you are you know. I understood all that. I think I'm having other issues, like a dying video card or arghhh mobo. Have tried everything short of going back to Win 7. Then again it could be the ATI drivers, so I'll play around with them a bit before I take an axe to the PC :furious:
Reverted to WIN 7 64bit and all is crunching very well indeed. Same with WIN 8 32bit.
Just loaded WCG HCC1 onto my WIN 8 64 bit machine. Same problems as the other machine yesterday. Video Driver keeps stopping and restarting.
So is anyone running Win 8 64bit with WCG HCC1 GPU WUs? Having problems? Which ATI driver are you using? Please!
@JugNut. If you think you are victim of a conspiracy - The above is a conspiracy perpertrated by Aliens!! Maybe I should chill as we are all supposed to perish tomorrow or Saturday.
AAAARGH!!! :sg1:
That's to bad Lawry, I thought windows 8 was giving you a good run? Can't help there though.(no win8)
Mmm Aliens hey? Works for me...
@JN. It IS the ATI drivers + Win 8 + WCG GPU WUs. They are the only ones that have problems, all the others are fine. Two entirely different PCs with the same results sort of prove (IMHO) THAT boinc and AMD haven't kept up.
I think I found the answer to JN's original problem in the first post - see http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,34749 (http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,34749)
Looks like the BOINC documentation was wrong and the ati/nvidia type string must be in upper case.
Anyway I just tried the following cc_config and so far it has not downloaded any more NVIDIA WUs:
<cc_config>
<log_flags>
</log_flags>
<options>
<exclude_gpu>
<url>http://www.worldcommunitygrid.org/</url>
<type>NVIDIA</type>
</exclude_gpu>
</options>
</cc_config>
BOINC Event Log says:
21/02/2013 7:10:39 PM | World Community Grid | Config: excluded GPU. Type: NVIDIA. App: all. Device: all
Thanks Dyeman, that might be just what the doctor ordered. I'll give it a crack a bit later when I get a chance and let you know what happened..
Sorry it took me so long to get back Dyeman, some real life drama's that couldn't be avoided.
So far with a quick trial it seems to working & only asking for ATI tasks. :cheers:
I'll give it a full trail when I switch over to WCG full time in a day or two for the AA
Thanks for the help mate + Karma for your effort
The following is copied over from anotgher thread and is by Kashi.
Quote
Multiple GPU tasks for World Community Grid Help Conquer Cancer
Update BOINC version to at least 7.0.42.
Adjust World Community Grid Device Profile to allow GPU work.
Download at least one GPU task.
Copy and paste the following into Notepad:
<app_config>
<app>
<name>hcc1</name>
<gpu_versions>
<gpu_usage>.25</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>
File>Save As..
Filename:app_config.xml
Save as type: All Files (*.*)
Place completed app_config.xml file in \www.worldcommunitygrid.org folder
Folder path in Windows 7 is C:\ProgramData\BOINC\projects\www.worldcommunitygrid.org
Stop and exit BOINC and then restart it to initialise the app_config.xml file.
This app_config.xml file will run 4 concurrent Help Conquer Cancer GPU tasks using 1 CPU core per GPU task.
If you have a faster or more recent architecture GPU than my HD 5870 you may wish to run 8 or more GPU tasks concurrently. To do this, open the app_config.xml file in Notepad and edit the <gpu_usage> value and then save the change. It works in fractions, so:
<gpu_usage>.5</gpu_usage> runs 2 concurrent GPU tasks
<gpu_usage>.33</gpu_usage> runs 3 concurrent GPU tasks
<gpu_usage>.25</gpu_usage> runs 4 concurrent GPU tasks
<gpu_usage>.2</gpu_usage> runs 5 concurrent GPU tasks
<gpu_usage>.167</gpu_usage> runs 6 concurrent GPU tasks
<gpu_usage>.143</gpu_usage> runs 7 concurrent GPU tasks
<gpu_usage>.125</gpu_usage> runs 8 concurrent GPU tasks
und so weiter
If you want to run more concurrent tasks than the number of CPU cores you have or you wish to leave some CPU resources free for a CPU project you can adjust the amount of CPU resources allocated for each GPU task by adjusting the <cpu_usage> value. For example:
<gpu_usage>.125</gpu_usage>
<cpu_usage>.5</cpu_usage>
would run 8 concurrent GPU tasks using half a CPU core each. This would use 4 CPU cores in total so all CPU resources of a quad core computer or half of the CPU resources of an 8 core computer.
The Help Conquer Cancer GPU application uses a great deal of CPU resources, so the more CPU you allocate to support GPU processing the more efficient the GPU application is. How you decide to set these variables is up to you depending on your hardware potential and how you wish to contribute to WCG. A bit of trial and error may be necessary to find the best values to suit your individual preferences. Make sure to monitor the GPU temp to ensure your card does not overheat especially on hot days and/or if you have a closed case. Some cards notably NVIDIA are unable to run WCG GPU tasks without suffering excessive screen lag. If this is the case you can adjust BOINC Manager preferences and untick "Use GPU while computer is in use" which will stop GPU processing while you are using the computer.
Remember to join BOINC@AUSTRALIA team because
If you have trouble following this a comprehensive guide at tech Power Up! forums Guide to running multiple GPU work units includes a link to m&m's Auto-setup utility for app_config.xml.
If you do not have a usable GPU or prefer to concentrate on WCG CPU applications it's all good, the other WCG projects are all worthy and cannot be done on GPU so are equally valuable, they just complete a bit slower.
I am doing some other CPU projects until the AA33 start date but for those who are starting early good luck and good crunching.