Printable View
I just got done installing a new ATI video card so I can crunch with GPU also. Computer is crunching 3 work units on the CPU and one on GPU. Is this the way it is supposed to work? I thought it would be 4 on CPU (quad core) and ??? on GPU...
Q6700
7950
The GPU still requires some CPU resources to handle moving data around and general management type stuff as well as sending and receiving new work. There's various app_info (or something like that) files people are using to maximise their GPU production and leaving the CPU to handle all the general purpose stuff. In other words, they're doing most if not all their crunching on the GPU and just using the CPU for support tasks.
That sounds like I'd have to go to strictly HCC to do it that way. I run a mixture of projects (almost everything!). Do I just leave my computer running as it is?Quote:
Originally Posted by D_A
David the app_config can be set to run 2, 3 or 4 GPU WUs. The numbers you input in app_config will regulate whether you run all GPU WUs or a mixture of GPU and CPU WUs. The choice is yours to make. GPU crunching is much more efficient and a better use of your available resources. There is only ~ 80 days or less worth of ati_hcc1 WUs left. You should consider maximizing your results inputs by going all out for GPU WUs. You can always return to the CPU projects after the GPU WU bonanza is exhausted.Quote:
That sounds like I'd have to go to strictly HCC to do it that way. I run a mixture of projects (almost everything!). Do I just leave my computer running as it is?
In summary: If you wish to continue with crunching 3 CPU WUs and 1 GPU WUs do nothing.
If you wish to go all GPU or do a 2x2 or 3x1 GPU to CPU split then post what you desire.
Folks will then generate a pertinent app_config file and guide you on how to insert it into the appropriate BOINC folder.
Have fun with your new 7950... It is a very good crunching video card.
OK. Thank you for clarifying that for me. If someone can tell me how to do all GPU work units while there are still some left to crunch, that's what I'll do.Quote:
Originally Posted by jeanguy2
David, it is fairly straightforward, what you need is an app_config file. EDIT: FOR APP_CONFIG YOU NEED TO RUN BOINC 7.0.42 OR LATER !! Select the one for your os on this page: http://boinc.berkeley.edu/download_all.php There are advantages to using 7.0.54 or later.
Users should be aware that due to the lack of driver support, namely open cl support, AMD 7xxx cards will not crunch using Windows XP.
Here is one that I use on a q6600. I have also used the same on 3770K
<app_config>
<app>
<name>hcc1</name>
<max_concurrent>16</max_concurrent>
<gpu_versions>
<gpu_usage>0.0625</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
<app>
<name>beta</name>
<max_concurrent>16</max_concurrent>
<gpu_versions>
<gpu_usage>0.0625</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
</app_config>
copy this into notepad then save it in the following folder: program data> Boinc> projects> www.worldcommunitygrid.org ensuring you select the "all files" option in the "save as type" box. Name the file app_config.xml
Unless you changed it already files and folders like this are hidden. Unhide them to get this done: (in search box type folder options and select that from results> you will see this
http://www.lakecityquietpills.com/ph...9759126224.jpg
As in the highlighted area select show hidden files and folders)
When you stop and re-start boinc this should then start working on 16 wu's running concurrently on your graphics card but your rig may have a number of other projects waiting to run. Go to your my grid page and change your profile to stop cpu work then abort any cpu tasks you may have in cache.
To get the best out of your rig it is necessary to spend some time separating in time the concurrent wu's so that they are all at a different stage of completion at any one time. May I suggest a figure of around 50 seconds to start with until you get a feel for how your rig is performing. It is the figure I use on a CPU @ 3.3, Gpu @ 1300 Rig. (If you run an overclocked 3770K or 2600K make this 40 seconds) Unfortunately this is not an exact science and you may find that, over time, the wu's tend to group together. If you are one that wants absolute best performance it is necessary to re-allign regularly (daily) or after you use graphics for anything else like watching vids
You may want to experiment with how many wu's to run concurrently so here is a little insight as to what each figure in the config file is doing
<app>
<name>hcc1</name>
<max_concurrent>16</max_concurrent>This is the total number of hcc1 wu's running at any one time on your RIG
<gpu_versions>
<gpu_usage>0.0625</gpu_usage>This is 1 divided by the number of wu's you wish to run on a CARD
<cpu_usage>0.25</cpu_usage>This is the number of cores/threads on your cpu that you wish to give to the GPU work divided by the number of GPU wu's you wish to run concurrently on the RIG
</gpu_versions>
</app>
Feel free to ask any question if I have been unclear on something
Excellent post OC... This should be made into a sticky for easy reference to by all Xtremers wanting to get into GPU crunching with the app_config method :up:
OldChap - where would I save the app_config.xml file if I am using Linux? Thanks.
Same place, in the projects folder.
I don't have one running just now but I seem to recall that Boinc in Linux is not dissimilar to windows and finding it was quite intuitive.
Thanks OC! I'm going to see if I can get this figured out tonight.
Quote:
Originally Posted by OldChap
Evidently I'm not smart enough to do this. I had to Google how to find the BOINC versions and the program data stuff. Managed to get 7.4.52 installed IN ADDITION TO 6.10.58. Don't think I got all that stuff saved into the program data file either. I put this video card in a computer with Vista Ultimate 64 that I use for nothing but crunching. I'm not as familiar with Vista as I am XP. XP is on the computer I actually use for something other than crunching. I'll try again someday when I have more time.
I had a helluva time trying to get AMD drivers for the 7770 installed on vista. Ultimately failed and used linux instead.
It seems I presume too much..... I have added some more to that how-to so that maybe more folks can use it.
I'm afraid I cannot help with drivers. Having never had an issue installing or updating I have not had to resolve any problems...thus I have no experience to add.
Your edit did help but still no joy. I'm off tomorrow so I'll have more time. I'll try again then. Thanks for the help.Quote:
Originally Posted by OldChap
I finally got it OC! Currently have 16 GPU work units running. It does look strange with only one project showing up though. I normally run several projects.
Thanks for the help.
That's what I like to hear :up:
Sooo, what are the odds that WCG will use GPUs again in the near future? I just got my old (ish) dually back up, running and crunching and thinking that maybe it is time for a semi-decent GPU, especially if it can help knock out some WUs. Was initially thinking a 7770, but am now leaning towards a 7850. The thing just seems like it would do a lot more WUs with 1024 stream processors vs 640.
Next it will be a 7950 lol.Quote:
Originally Posted by [XC] serlv
Heh. Close to $200 is a stretch for me these days. I used to sink a lot of $$ into crunchers, was running 25 of them at one time. But I don't seem to have the cash for that kind of crunching anymore. Not even close.
Haven't crunched in a while. I was number 2 on the team for years, but have barely been doing anything for a while. Was shocked to see I'd dropped to #16. Jeez! LOL
So now I'm just thinking of maximizing this rig I just got back up and running. Would a 7850 do any damage with a dual Harperton?
Scott, I think you will be on a scale somewhere between "pleasantly surprised" and "outright shocked" at what a half decent GPU will do compared to what was not so long ago top tier CPU hardware.
Hi guys, can someone look at my config files and tell me, why always computing with 1 Wu per 1 gpu? thank you
http://extreme.overclocker.sweb.cz/wcg/App.gif
http://extreme.overclocker.sweb.cz/wcg/cc%20conig.gif
http://extreme.overclocker.sweb.cz/wcg/pojects.gif
http://extreme.overclocker.sweb.cz/wcg/snap.gif
My system- Alienware m17xR3 with 2670qm, Ati 6970m gpu and 8Gb memory.
Your app_config.xml seems to lack the closing tag:
</app_config>
Also post which BOINC version are you using.
Quote:
Originally Posted by ex.treme
Pic shows 7.0.28.
yep the app_config is incomplete.
In order to run app config you need to update to the latest Boinc beta to get it running with a few other benefits too.
Take a look at http://www.xtremesystems.org/forums/...estions-thread and post further questions there please.
Ouch my fault, i was copy and bad paste from your config up there. Why is there the beta name?
<app_config>
<app>
<name>hcc1</name>
<max_concurrent>16</max_concurrent>
<gpu_versions>
<gpu_usage>0.0625</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
<app>
<name>beta</name>
<max_concurrent>16</max_concurrent>
<gpu_versions>
<gpu_usage>0.0625</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
</app_config>
Last Boinc 7.0.59 i give it try. With Boinc 7.0.28 i can run only 1Wu/ 1 gpu?
One more question:
Do you think guys i can improve points with settings like Oldchap has- more wu's per gpu?
My main system:
Gigabyte Z68XP-UD4+ i2500k, 2xHD 7750, 2x4GB Corsair DDR3 1600. Cpu has only 4 cores, no HT.
One HD 7750 make one Wu about 5 minutes.
Thank you.
The Beta app was included for folks that are chasing beat badges only.
The only way to find out how many concurrent wu's to run is trial end error. I would try 4 on your card and then try 8 then 12 etc. each time let it run for a while to observe both points and the separation behaviour.
Your rig can handle it. I have a 7950 on a q6600. Not sure what to expect from that card.
More wu's does not mean more points but for me it does mean that in general the separation lasts longer.
OC - FYI - My 7770s erred at your 0.25C + 0.0625ATI setting. I backed it up to 0.125ATI and so far so good.
Oldchap:
sorry for noob question. If i understand it, why separate this wu's and cpu time is effective use cpu and gpu computing time.
If i running 2x gpu / 2x wu's and start it one time is not so effective like when i start second after 1 minutes. Is that point?
I saw every time at 50% and 100% works time waiting system one minutes...
So when i switch i2500k for i 2600k save some seconds with your solutions...
Thank you.
Offsetting in time is done because during the course of completing a wu there are GPU only (+ a very small amount of cpu) and CPU only (+ a small amount of GPU) phases to the work. starting one then a while later starting another ensures that at any one time a work unit is in GPU work while the other is in CPU work
and adding more concurrent wu's is done in an attempt to get the GPU load up in excess of 97% because GPU is so much more efficient than cpu we try to load it as much as possible without errors. The constant overlap of various wu's helps us get as much out of the rig as possible.
I am not sure you will gain a lot with your GPU when you swap cpu's
Ok, i will test in few days.
Interesting thing:
AMD HD7750 - time 1 Wu about 5 min
AMD 6970m - time 1 Wu about 10 min
even if 6970m is more powerfull in 3d mark Vantage - 10700 vs 11300points
Guess more powerful opencl with 77xx cards.