I've only noticed 1 problem in 6.3.8 for Windows32 relating to the GPU.

I will report this on the alpha mailing list, I jsut want to know if anyone else has seen this ?

Most of the time on my P4-HT which should always be running 2 tasks, either 2 CPU or 1 CPU and 1 GPU they will be. But several times I noticed they would only be running 1 GPU task. Being busy at work I could not fully investigate at the time, but then a check later in the day would show them running 2 again. One day I had time to keep an eye on the host and every time any task ended it would properly start a second. I came to realize that this only happens when BOINC restarts or after a reboot. After several days of testing this I figured this out. It happens on both computers and has happened every time. The problem I had before is I do not always reboot both each day and sometimes need to reboot one or the other in the middle of the day. It is not just during a reboot. I've shut down boinc and core client without rebooting and restarted both, it will do the same thing, only run 1 GPU task, and then after some odd number of minutes properly go back to running 2 tasks. I've observed this gap to be 2,11,14 or up to about 90 minutes. I think its when the scheduler makes a check for the next switch. Yes, my work units are set to switch every 90 minutes and not the default of 60.

Example log entries:
8/13/2008 8:53:56 AM||Starting BOINC client version 6.3.8 for windows_intelx86
8/13/2008 8:53:56 AM||This a development version of BOINC and may not function properly
...blah blah blah
8/13/2008 8:53:56 AM||Processor: 2 GenuineIntel Intel(R) Pentium(R) 4 CPU 3.00GHz [x86 Family 15 Model 4 Stepping 9]
8/13/2008 8:53:56 AM||Processor features: fpu tsc pae nx sse sse2 mmx
8/13/2008 8:53:56 AM||OS: Microsoft Windows XP: Professional Edition, Service Pack 2, (05.01.2600.00)
8/13/2008 8:53:56 AM||Memory: 2.00 GB physical, 3.85 GB virtual
8/13/2008 8:53:56 AM||Disk: 149.04 GB total, 79.62 GB free
8/13/2008 8:53:56 AM||Local time is UTC -4 hours
8/13/2008 8:53:56 AM||Not using a proxy
8/13/2008 8:53:56 AM||CUDA devices found
8/13/2008 8:53:56 AM||Coprocessor: GeForce 8800 GT (1)
...url's deleted to hide test projects, sorry no secrets revealed.
8/13/2008 8:53:56 AM|PS3GRID|URL: http://www.ps3grid.net/; Computer ID: 6576; location: work; project prefs: work
8/13/2008 8:53:56 AM||General prefs: from malariacontrol.net (last modified 12-Apr-2008 07:22:29)
8/13/2008 8:53:56 AM||Computer location: work
8/13/2008 8:53:56 AM||General prefs: using separate prefs for work
8/13/2008 8:53:56 AM||Reading preferences override file
8/13/2008 8:53:56 AM||Preferences limit memory usage when active to 1227.89MB
8/13/2008 8:53:56 AM||Preferences limit memory usage when idle to 1739.51MB
8/13/2008 8:53:56 AM||Preferences limit disk usage to 4.56GB
8/13/2008 8:53:57 AM|PS3GRID|Restarting task lt31469-TEST12-0-5-acemd_2 using acemd version 628
At this point log stops and only the 1 task will be running.

Next entry will be
8/13/2008 9:42:54 AM|Poem@Home|Restarting task data_370_1218579663_2008419030_0 using poem version 9

Notice the time difference before a second task is started.
____________
Alpha Tester ~~ BOINCin since 10-Apr-2004 (2.28) ~~~ Join team USA

I have seen the same behavior with 6.3.8 on Linux, some times it starts 4 tasks, some times it processes just the PS3GRID app, while the other processes are shown as "Waiting to run". After restarting BOINC or suspending and resuming the second app it works again without problems. I never tried to wait until the hourly switching comes.

I have seen the same behavior with 6.3.8 on Linux, some times it starts 4 tasks, some times it processes just the PS3GRID app, while the other processes are shown as "Waiting to run". After restarting BOINC or suspending and resuming the second app it works again without problems. I never tried to wait until the hourly switching comes.

I've also seen and reported it for 6.3.5/Linux64 and it actually got better with 6.3.8, but it still happens... Mostly if BOINC is started, or after the benchmarks ran.

Solution is to suspend a task in your queue which is not running, and then BOINC starts to use all CPUs. Or you wait until the next task switch...
____________

pixelicious.at - my little photoblog

OK,
I did some debugging for Dr.Anderson and hopefully he found a solution.

Thanks - this is exactly the info I needed.

I checked in what I hope is a fix.
We'll make a new release; I'll let you know when it's available.

OK,
I did some debugging for Dr.Anderson and hopefully he found a solution.

Thanks - this is exactly the info I needed.

I checked in what I hope is a fix.
We'll make a new release; I'll let you know when it's available.

Good!
D Anderson is also going to visit us in Barcelona on 5 September, so please write down any other issue.

gdf