Hi here are 2 tasks that generate errors but don't crash (anymore?)

http://setiathome.berkeley.edu/result.php?resultid=1108406670
http://setiathome.berkeley.edu/result.php?resultid=1108406664

http://www.efmer.eu/seti/23no08ad.15915.22976.9.8.127
http://www.efmer.eu/seti/23no08ad.15915.22976.9.8.131

Hi here are 2 tasks that generate errors but don't crash (anymore?)

Another thing I noticed that at that moment I get 3+1 tasks running instead of the normal 2+1 Cuda. Cuda task still runs about as fast as always and uses about 5% of the load.

Hi here are 2 tasks that generate errors but don't crash (anymore?)

http://setiathome.berkeley.edu/result.php?resultid=1108406670
http://setiathome.berkeley.edu/result.php?resultid=1108406664

http://www.efmer.eu/seti/23no08ad.15915.22976.9.8.127
http://www.efmer.eu/seti/23no08ad.15915.22976.9.8.131

Ok, it's AR~0.15 VLARs. Will run them standalone.

[quote]I have heard, but not substantiated, that 6.5.0 handles things correctly.

Better read this first
Problems with 6.5.0

[quote]I have heard, but not substantiated, that 6.5.0 handles things correctly.

Better read this first
Problems with 6.5.0

I tired 6.5 too.
Whe I close BOINC manager it coses its window, but leaves its process in memory (boincmgr.exe) leaves all other processes in memory (boinc.exe, all science apps). But boincmgr starts to consume whole CPU core (25% CPU in my quad).
I saw this behavior each time when I attemped to close BOINC manager. So I reverted to 6.4.5.

Not your gpu wont match 4 cores... Just for reference my gtx 260 core 216 does tasks in max 10 min or in other words the equivalent of 6 cores...

What size tasks does it complete in 10 min max. A 14 credit task or an 80 credit task? Can't a quad put out 4 14 credit tasks in under 15 minutes?

Tasks 60+ credits all take 10 min max
The 14 credit ones take around 3:30 (14 min for 4)

---

Do you Good Search for Seti@Home? http://www.goodsearch.com/?charityid=888957
Or Good Shop? http://www.goodshop.com/?charityid=888957

Tasks 60+ credits all take 10 min max
The 14 credit ones take around 3:30 (14 min for 4)

Oh, that's too cool, I like that.

Thanks

---

2. The Seti MB work units you download will all get branded as cuda and then they only run on the gpu, one at a time. The cpu will happily crunch away at AP and/or other projects and ignore the cuda work.

Alexander Klietz wrote:

Is this really the case? Don't get me wrong, I'm asking. I guess you mean the task is branded as cuda and not the whole workunit, because workunits are send out to CPU+GPU clients as well as to only CPU+CPU clients and so on. But nevertheless, I see no indication that the cuda tasks are different from others. How are they branded?

My above example shows that it is possible to crunch a workunit (task) received for cuda with the CPU *but* that was using an anonymous platform (app_info.xml) and tweaking it, so it might not tell much about the standard mechanism.
...
Alex

This project is producing work for two applications, named "astropulse" and "setiathome_enhanced". With an app_info.xml, the BOINC core client simply assigns new work to the app with the right name and highest version it knows about. Without an app_info.xml, the Scheduler gives the full list of files to be used for a task including the executable, DLLs, images, etc. and the core client downloads any which aren't already present. That's the "branding", highest known version in either case.

With a modified app_info.xml, it would certainly be possible to have an AK_v8 as version 6.03 plus the CUDA capable 6.05. Then perhaps a script could be set up to rewrite client_state.xml to have 6.03 used for VLAR work. The app_info.xml would probably need to cross-reference all files between the versions to keep the core client from deleting files when it thinks they will no longer be used, and the script would probably have to run with BOINC shut down.

You asked earlier about the CPU fallback capability, I don't think it's usable because it is just a check that there's a CUDA 1.0 or better device with at least 128 MB of global RAM. BOINC will not have started the task unless there's a CUDA 1.1 device with at least 254 MB. It would be slow anyhow:

Optimal function choices:
-----------------------------------------------------
name                
-----------------------------------------------------
              v_BaseLineSmooth (no other)
            v_GetPowerSpectrum 0.00041 0.00000 
                   v_ChirpData 0.01534 0.00000 
                  v_Transpose4 0.00997 0.00000 
               FPU opt folding 0.00342 0.00000 

That's the basic set of CPU functions from 5.15, the vectorized variants are apparently not compiled in.

                                                                Joe

Not your gpu wont match 4 cores... Just for reference my gtx 260 core 216 does tasks in max 10 min or in other words the equivalent of 6 cores...

What size tasks does it complete in 10 min max. A 14 credit task or an 80 credit task? Can't a quad put out 4 14 credit tasks in under 15 minutes?

Tasks 60+ credits all take 10 min max
The 14 credit ones take around 3:30 (14 min for 4)

Hmmmm... Not sounding promising - I've been trying to persuade myself that my passive-cooled GeF 7600GS needs replacing but my o/c'd Q9450 knocks out 4 x 14 credit tasks in under 10 minutes so even the GTX 260 would be a backward step there. OTOH, it would cut the time for 60+ credit tasks by 75% so... mumble - mumble -

F.

---

What size tasks does it complete in 10 min max. A 14 credit task or an 80 credit task? Can't a quad put out 4 14 credit tasks in under 15 minutes?

Yes, I do 4 in 14 min. , how much more could I do with a GPU ?

[quote]I have heard, but not substantiated, that 6.5.0 handles things correctly.

Better read this first
Problems with 6.5.0

I tired 6.5 too.
Whe I close BOINC manager it coses its window, but leaves its process in memory (boincmgr.exe) leaves all other processes in memory (boinc.exe, all science apps). But boincmgr starts to consume whole CPU core (25% CPU in my quad).
I saw this behavior each time when I attemped to close BOINC manager. So I reverted to 6.4.5.

Trak #817 raised for this problem (apparently there was a trak 814 but it referred to a service install, so it was rejected).

It will shut down the science apps if you tell it via the menu option. Advanced -> Shutdown connected client. It just doesn't do it when you tell it to exit (regardless of the exit dialog box).

I still haven't see what it will do when it finishes the last cuda task. My last cuda task should be done in 30 mins or so...

---

BOINC blog

when are the graphics coming back for the screensaver ? I miss the scree saver graphics.

---

Jason Tobin
Alien Hunting Specialist

when are the graphics coming back for the screensaver ? I miss the scree saver graphics.

App performance optimization and app graphics just two opposite directions of app improving :)

I have Intel Quad 6600 computers and have serveral questions. My computers are currently running the optimized app from Lunatics with very good results. My GPU is a GeForce 9800 GT.

First. Can CUDA crunch fast enough to achieve the current throughput of 4 CPU's? In other words will my current RAC remain the same?

I am running a standard Q6600 with a 8800GTX. Raistmer's latest gives me about 10% more credit (when I can get it to run :) can't get any MB at the moment :( not sure why) on the GPU than my stock Q6600 on its own.

Of course I've still got 99.96% of my 4 cores to crunch AP units or even other apps. Lovely.

I'm not getting any MB for my poor cool under-used GPU! I had a good old overflow & trashed umpteen lovely CUDA ready units approx 36 hours ago - got to sleep sometimes! I didn't get any uits then but just thought I had used my quota for the day. Since then I've grabbed Raistmer's latest offering with attempts at coping with memory leaks; 606r380.

I am not getting any scheduler requests for MB/CUDA. I'm missing something really obvious & it is bugging me beyond belief. Any ideas will be most appreciated. Below is my start up message set.


12/31/08 10:36:12||Starting BOINC client version 6.5.0 for windows_intelx86
12/31/08 10:36:12||This a development version of BOINC and may not function properly
12/31/08 10:36:12||log flags: task, file_xfer, sched_ops, coproc_debug
12/31/08 10:36:13||Libraries: libcurl/7.19.2 OpenSSL/0.9.8i zlib/1.2.3
12/31/08 10:36:13||Data directory: C:\ProgramData\BOINC
12/31/08 10:36:13||Running under account Dad
12/31/08 10:36:14|SETI@home|Found app_info.xml; using anonymous platform
12/31/08 10:36:14||Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 11]
12/31/08 10:36:14||Processor features: fpu tsc pae nx sse sse2 pni mmx
12/31/08 10:36:14||OS: Microsoft Windows Vista: Home Premium x86 Editon, (06.00.6000.00)
12/31/08 10:36:15||Memory: 3.00 GB physical, 6.17 GB virtual
12/31/08 10:36:15||Disk: 916.46 GB total, 535.69 GB free
12/31/08 10:36:15||Local time is UTC +0 hours
12/31/08 10:36:15||Not using a proxy
12/31/08 10:36:16||CUDA device: GeForce 8800 GTX
12/31/08 10:36:34|rosetta@home|URL: http://boinc.bakerlab.org/rosetta/; Computer ID: 971969; location: home; project prefs: default
12/31/08 10:36:34|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 4195698; location: home; project prefs: default
12/31/08 10:36:34||General prefs: from http://bam.boincstats.com/ (last modified 12-Sep-2007 21:07:15)
12/31/08 10:36:34||Host location: none
12/31/08 10:36:34||General prefs: using your defaults
12/31/08 10:36:34||Reading preferences override file
12/31/08 10:36:34||Preferences limit memory usage when active to 3069.32MB
12/31/08 10:36:34||Preferences limit memory usage when idle to 3069.32MB
12/31/08 10:36:34||Preferences limit disk usage to 100.00GB
12/31/08 10:36:35||Contacting account manager at http://bam.boincstats.com/
12/31/08 10:36:35|SETI@home|Restarting task ap_26au08aa_B3_P1_00120_20081108_20808.wu_3 using astropulse version 500
12/31/08 10:36:36|SETI@home|Restarting task ap_15no08ac_B2_P0_00223_20081227_26819.wu_0 using astropulse version 500
12/31/08 10:36:37|SETI@home|Restarting task ap_18oc08aa_B1_P0_00162_20081127_24690.wu_2 using astropulse version 500
12/31/08 10:36:38|SETI@home|Restarting task ap_15no08ac_B2_P1_00228_20081227_26849.wu_1 using astropulse version 500
12/31/08 10:36:41||Account manager: BAM! User-ID: 930
12/31/08 10:36:41||Account manager: BAM! Host-ID: 90795
12/31/08 10:36:41||Account manager: This host connected 5363 times to BAM!
12/31/08 10:36:41||Account manager contact succeeded
12/31/08 10:36:41||General prefs: from http://bam.boincstats.com/ (last modified 31-Dec-2008 10:36:41)
12/31/08 10:36:41||Host location: none
12/31/08 10:36:41||General prefs: using your defaults
12/31/08 10:36:41||Reading preferences override file
12/31/08 10:36:41||Preferences limit memory usage when active to 3069.32MB
12/31/08 10:36:41||Preferences limit memory usage when idle to 3069.32MB
12/31/08 10:36:41||Preferences limit disk usage to 100.00GB
12/31/08 10:40:50|SETI@home|Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks
12/31/08 10:40:56|SETI@home|Scheduler request completed: got 0 new tasks
12/31/08 10:40:56||General prefs: from http://bam.boincstats.com/ (last modified 12-Sep-2007 21:07:15)
12/31/08 10:40:56||Host location: none
12/31/08 10:40:56||General prefs: using your defaults
12/31/08 10:40:56||Reading preferences override file
12/31/08 10:40:56||Preferences limit memory usage when active to 3069.32MB
12/31/08 10:40:56||Preferences limit memory usage when idle to 3069.32MB
12/31/08 10:40:56||Preferences limit disk usage to 100.00GB

Again, thanks for any suggestions or "silly boy it's this" answers.

12/31/08 10:40:50|SETI@home|Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks
12/31/08 10:40:56|SETI@home|Scheduler request completed: got 0 new tasks
Again, thanks for any suggestions or "silly boy it's this" answers.

Your host recives just what it requests - that is - nothing :)
You got too many AP probably and now BOINC thinks you not need any more work.
Try to increase cache size or just wait until few APs will be crunched.

(And yes, I know AP are doing on CPU and your GPU free and idle - but current BOINC version too silly to understand this fact ;) )

Had task 1109002462 get stuck at the typical cpu time of 39 seconds like most VLAR task I've had. It also caused a sluggish/freezing effect on my pc.

I've been able to do this approximate AR before without any problems, but this task almost crashed Boinc. It appeared at the time I would not have been able to even go the the slots folder and check on the AR, (was before I had my batch file worked out), before Boinc would have crashed. I was surprised the abort button even worked.

Not sure why it got stuck, or if the "Breakpoint Encountered (0x80000003) at address 0x7C901230" error in the stderr is from aborting the task, or what caused the task to get stuck in the first place.

Did not see any AV interference, or other progs running which would have affected it AFAICT.

8500 GT
180.60
XP Home SP2

---

If you have CUDA app freezing look in this thread http://setiathome.berkeley.edu/forum_thread.php?id=51051 There is some VB-script for terminating freezed app.

Had task 1109002462 get stuck at the typical cpu time of 39 seconds like most VLAR task I've had. It also caused a sluggish/freezing effect on my pc.

I've been able to do this approximate AR before without any problems, but this task almost crashed Boinc. It appeared at the time I would not have been able to even go the the slots folder and check on the AR, (was before I had my batch file worked out), before Boinc would have crashed. I was surprised the abort button even worked.

Not sure why it got stuck, or if the "Breakpoint Encountered (0x80000003) at address 0x7C901230" error in the stderr is from aborting the task, or what caused the task to get stuck in the first place.

Did not see any AV interference, or other progs running which would have affected it AFAICT.

8500 GT
180.60
XP Home SP2

May not be related, but you do realise that for the time that your batch file file is scanning a task file, the task file will be locked for access just the same way as with AV progs?

F.

---

Yes, I figured it would be something along those lines, though I haven't had any problems with any tasks the times I've run it since having it available.

I do however know I wasn't using the batch file in this case, because it was this task that made me decide to work on the batch file and get one going. Up till then I didn't have enough of them to motivate me to do it, but I didn't like the feel of that WU :) and I thought at the time it was a VLAR affecting me more than they had in the past.

I've also had several tasks that had a sluggish/freezing effect, but that usually only lasts' about 10-15 seconds and then the tasks continue without error/

---