V10/11 of modified SETI MB CUDA + opt AP package for full multi-GPU+CPU use

[MaGNeT]

@ Raistmer

I got a 9800GTX+ today for experimenting purposes and found this topic.
Installed the V10 with the update and it looks like everything is working great.

Will let you know in a few days (or earlier) if I encounter any problems.

Great work!

[elec999]

I am a few computition errors once in a while, but overall is going strong.
Thank you

[k6xt]

I came to grips, internally, with AP vs SETI vs MB work. After reading in the SETI boards that someone had success manually managing the division of work I stopped AP WU downloads temporarily . Sorry I don't recall where, was reading a lot. Its working for me.

Why I post is for this interesting factoid which I have not seen elsewhere. Temporarily I halted inprogress AP until one core switched to SETI and began processing. It seems it was a CUDA unit. After a time I restarted AP processing..lo and behold! The CUDA unit continued processing, albeit very slowly such that for a time the BOINC tasks tab timer was increasing rather than decreasing.

This is probably not a good thing and I won't do it again. My result does not match wingman, and a second wingman is now processing. I'm sure I will lose credit. I believe this is the result:

http://setiathome.berkeley.edu/workunit.php?wuid=424562219

The good news viz AP is patience is a virtue. I've completed about 30 AP WU which have finally begun to verify. V10 is calculating these at ~12 hr per WU. After shutting off further AP download I finally got some SETI which will begin to process tomorrow as the AP cache empties.

If the gurus have a better way to manage work I am all ears. It could be, if I was patient enough, the work would manage itself as the SETI WU return deadline gets near such that BOINC would prioritize them. The trick seems, if some SETI WU are desired, to occasionally manually disable AP downloads, let the cache empty. OTOH I read a blog by Matt Lebofsky implying the reason we got lots of AP recently has to do with differences in AP vs SETI server queue distribution, such that it will eventually manage itself. I wish I'd recorded where I read that on the SETI boards.

Finally, rummaging through the several xml files I noticed that both SETI and BOINC have "Maintain this minimum work" and "Maintain that extra days work" with min set to 0.1 days (or maybe 0.01 I forget) and of course the max of 10 days. I have now set these both to 5 days, i.e. maintain 5 days min and 5 days additional is how the xml states it. I know not, TBA, if this is useful - just did it this AM. Could not find enlightenment on the boards. Some posters have expressed dismay at 10 days cached work. Out here in southwest Colorado USA flyover country we lose power, internet access, and both such that 10 days can be a minimum during stormy season which is fast approaching.

Searching the boards is one time consuming, albeit productive, activity for those of us who are not in the guru category, although fully capable of adding 2 and 2. Occasionally a succinct summary of current events, functionality, expectations etc would be very useful troubleshooting to compare what we SHOULD see vs. what we DO see.

V10 is working for me! Profuse thanks to those who built it.
Art

[petefan]

Hi guys....

I read page one and just installed V10 (the third one) for a P4 + one 8500GT (driver 182.08) to see the effect....
My BOINCMGR version is 6.6.11, OS XP SP3 x86.
The CPU is not crunching at all, the process AK_v8b_win_SSE3_GPU_CPU_team_V10 is idle.
BOINCMGR shows "No work available for the applications you have selected. Please check your settings on the web site."
This is my cc_config.xml:
<cc_config>
<options>
<ncpus>2</ncpus>
</options>
</cc_config>


How may I correct this problem?

This is a test (taste) for this new app I came across, I am so willing to use this app for all my computers with GTX295 to get higher RAC!!

So many thanks!!

Peter.

[Devaster]

it seems that team mod is not needed for SETI main - if u install 6.6.15 and above it will use CPU for SETIMB too

[elec999]

Is it worth running my AMD 3400+ and Intel P4 HT. Or are they just slowing down my other systems. Right now those two got work, and my other systems get no work.
Thank you

[Marius]

A strange problem which occurs here from time to time is that sometimes when boinc 6.4.7 want to do some high priority tasks it starts a few new tasks but it seems to ignore the max amount of cpu's mentioned in the cc_config.xml (or is counting them wrongly). At such moments my q6600 and nvidia 9800gt are then running 5 cpu tasks + a gpu task (so 6 tasks in total).

Is this problem a boinc related problem or has it todo with a changed cc_config.xml or optimized client?

Any ideas?
Greetings,
Marius

Processor: Intel(R) Core(TM)2 Quad CPU Q6600
GPU processor:   GeForce 9800 GT (driver 185.20)

cc_config.xml
<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

[Marius]

Another problem i noticed; Sometimes the MB_6.08_mod_CUDA_V10.exe stops while its at 1 second of cpu (MB shows no activity in task manager and its using only 32.412kb of memory). It stays this way until i restart boinc. Is this a known problem/bug or can i solve/finetune boinc or nvidia software to avoid this?

It could be related to high priority boinc units which seem to overrule running threads (but i just as well be plain wrong)

Greetings,
Marius

Processor: Intel(R) Core(TM)2 Quad CPU Q6600
GPU processor: GeForce 9800 GT (driver 185.20)

cc_config.xml
<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

[Raistmer]

Another problem i noticed; Sometimes the MB_6.08_mod_CUDA_V10.exe stops while its at 1 second of cpu (MB shows no activity in task manager and its using only 32.412kb of memory). It stays this way until i restart boinc. Is this a known problem/bug or can i solve/finetune boinc or nvidia software to avoid this?

It could be related to high priority boinc units which seem to overrule running threads (but i just as well be plain wrong)

Greetings,
Marius

Processor: Intel(R) Core(TM)2 Quad CPU Q6600
GPU processor: GeForce 9800 GT (driver 185.20)

cc_config.xml
<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Look this thread about CUDA app freezings: http://lunatics.kwsn.net/12-gpu-crunching/vbscript-fights-cuda.0.html

[RottenMutt]

I'm running "stock" Boinc 6.6.17-64bit on my lappy, no optimized application installed yet. I immediately noticed that it will keep the GPU fed with work units, it also labels the Enhanced 6.08 work units as "(cuda)". It appears to me the latest Boinc runs cuda applications and fetches cuda applications as you would expect, woot!

I noticed in the c:/program data/.../setiathome.berkeley.edu folder there is not an app_info.xml file. If a install Raistmer's Opt V10a Pack along with a appropriate cc_config.xml file (or none) I end up running one more AP work unit then the number logical CPU's, rather then placing a 6.08 cuda work unit on the GPU.

So how can one use Raistmer's Opt V10a Pack and get it to behave the same as stock???  I think what is in the app_info file needs to be updated for the latest boinc.

what would keep me from renaming Raistmer's executables the same a the stock executable and copy the dll files?  i really don't want to do that, they could easily be over written by Bionc.

edit: also posted http://setiathome.berkeley.edu/forum_thread.php?id=52628&nowrap=true#878280

[Raistmer]

I'm running "stock" Boinc 6.6.17-64bit on my lappy, no optimized application installed yet. I immediately noticed that it will keep the GPU fed with work units, it also labels the Enhanced 6.08 work units as "(cuda)". It appears to me the latest Boinc runs cuda applications and fetches cuda applications as you would expect, woot!

I noticed in the c:/program data/.../setiathome.berkeley.edu folder there is not an app_info.xml file. If a install Raistmer's Opt V10a Pack along with a appropriate cc_config.xml file (or none) I end up running one more AP work unit then the number logical CPU's, rather then placing a 6.08 cuda work unit on the GPU.

So how can one use Raistmer's Opt V10a Pack and get it to behave the same as stock???  I think what is in the app_info file needs to be updated for the latest boinc.

what would keep me from renaming Raistmer's executables the same a the stock executable and copy the dll files?  i really don't want to do that, they could easily be over written by Bionc.

AFAIK you need app_info in all cases when you want to use opt apps. No mater what BOINC version do you use or do you use CPU only or CPU+GPU.
Working w/o app_info will enable only stock apps on your host.

[RottenMutt]

what is the benefit of your optimized cuda over stock cuda???

I'm going to try MarkJ's app_info, just wondering if it is worth while to insert your cuda app over the stock.

this should finally get my gpu loaded.

[Raistmer]

what is the benefit of your optimized cuda over stock cuda???

1) VLAR-kill mod
2) correct priority settings
3) elapsed time information availability
[4) affinity binding and ignoring of BOINC suspend order]

Last position is relative advantage, not in all cases it's good.
In coming V11 (at least CUDA app if not team pack) will be reduction of initial CPU time.

But I spoke not only about CUDA app itself. If you will not use app_info.xml you will not be able to use opt AP (~5x performnce loss) and opt AK_v8 (~1,5-2 performance loss).
All opt apps need appropriate app_info to be used.

[RottenMutt]

ok, i plugged in your cuda app.  i'm running all your apps on Boinc 6.6.17 using MarkJ's app_info.xml, i replaced the stock executables with yours in the app_info file and everything works.  i have not had a chance to see if the gpu count option works yet at the bottom of the app_info file.

see the following thread for MarkJ's app_info instructions.
http://setiathome.berkeley.edu/forum_thread.php?id=52589

edit: i have not had to modify the gpu count at the bottom of MarkJ's app_info file on machines with more then one video card.

[chronek]

I wonder if could be 2 quenies , one for ap , sec for gpu - all mb go for gpu , and we could set cpu cores for quenie ap  , and units gpu with quenie mb , and if ap quenie would be empty they could fetch from sec quenie mb - that could solve problems with gpu idle - and cpu idle , and for cpu - eliminate situation when work at 3 wu on 2 cores