Hi! The app "Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati)" crash on my laptop every time it started. I don't know how to get further detail (e.g. the log file of the app) in order to get this fixed. The OS is Windows 7 Home Basic, ATI driver version: 8.951.0.0 on ATI HD 7730M. And it seems that this laptop can choose which video card to use. There is another integrated Intel HD Graphics 4000. I'm to quite sure about that. What should I do?
Hi!
Is that the most recent driver available for that card? Updating the driver (usually) can't hurt.
Unfortunately the app isn't producing any meaningful diagnostic messages before it terminates.
Are you subscribed to other BOINC projects which offer OpenCL apps for ATI/AMD cards? I wonder whether this is a general problem with the OpenCL support on this particular host or it is specific to teh E@H app.
Cheers
HB
Hi, Bikeman
I don't think it would be cause by an old version of ATI driver, because I just bought it no more than 3 month ago. It is a brand new 3rd gen i7 core laptop. The driver should not be too old to make such a crash. I subscribed SETI@Home as well. And there are tasks using ATI opencl app ( AstroPulse v6.04 opencl_ati_100 ). It works fine so far. Not sure if it will crash as well as it is still running.
So sorry to hear that no log could be find. Thank you all the same.
Especially for notebooks, the GPU load can be quite nagging: slow down of the user interface, heat and fan noise. I really like that BOINC now has the option to selectively suspend the GPU in the "Activity" menu. So you can, for example, leave GPU suspended except when you really don't care (say at night). You will still get a lot more science done when using the GPU app during the night only compared to all-day CPU-only operation.
Cheers
HB
In the past when I tested older laptops by running Einsteins or LHC/T4T they did have those problems but the quad-core I am on right now never has any problem and it runs 2-core T4T,LHC,and 2 Einstein BRP4 Cuda's 24/7 with no heat problems and of course no fan problem since June 15th
Now this GeForce and processor isn't quite the same as my desktops with nVidia 550ti overclocked or 660Ti superclocked.
To be complete here, Sumtec has since his post posted at Seti that his system BSODs when running Astropulse tasks on his AMD. I'm asking him to uninstall his Intel HD Graphics 4000 drivers, as I fear that these interfere with the OpenCL drivers on his system (I am investigating multiple similar situations for the BOINC developers).
There are numerous notes (mostly for Windows 8) like:
[Optimus notebook]: When attempting to install the NVIDIA driver over a prior driver version, the installation may fail with a blue-screen crash pointing to dxgkrnl.sys. [1000723]
This intermittent issue occurs on only a small number of systems.
• To avoid this issue, first remove the older NVIDIA driver using Add/Remove Programs in the Windows Control Panel, then install the new driver using NVIDIA setup.exe.
• If you encounter this issue, boot to Safe Mode, then remove the older NVIDIA driver using Add/Remove Programs in the Windows Control Panel. Install the new NVIDIA driver using NVIDIA setup.exe.
[RH - the 'Optimus' trademark refers to the dual technology where an NVidia GPU is teamed with an Intel HD graphics chipset]
Prior to this new app i was having problems getting anymore than 75% gpu load, this includes crunching simultaneous gpu wu's. I tried leaving a cpu core dedicated to the gpu units etc but again failed to get beyond 75% gpu usage.
It seems 1.28 has totally resolved the above issue & im now at 90% of gpu load with massive reductions in gpu crunch time.
The only issue with 1.28 i encountered was comodo firewall blocking it, a number of wu's exited with the error couldn't start CreateProcess() failed - Access is denied. (0x5): -148, this was easily corrected by creating a rule to allow the new app in the firewall.
I've just reached the CUDA v1.28 batch on my new GTX 670 (Kepler) - host 5744895.
Seamless transition, with runtime dropping 25% (from ~2400 seconds to ~1800 seconds) and CPU usage dropping 50% (~700 seconds to ~350 seconds). I'm running two tasks at a time, so that's closing on 100 tasks per day - what was the speed limit again for this project?
Seriously - great job, no problems at all. You won't see quite that throughput, because I run other projects on the card too, but maybe I'll get a finder's certificate one day...
... but maybe I'll get a finder's certificate one day...
You're more likely to get one than me since I have just one lowly 550Ti doing BRP4 tasks :-). It has a cache of 1.25 branded tasks (running 2 at a time) so I suspended them all to allow a couple of 1.28s to start. The first pair have just now completed and the improvement is very pleasing.
Run time has gone from approximately 5600s to 4100s and the CPU time from approximately 1500s to 1100s. The card is in a host with a Q9400 quad core with one core free.
I'm very tempted to set up AP to run the remaining 1.25 tasks with the new app but it's getting late here and I need to go home. Maybe in the morning if I get really keen ...
... but maybe I'll get a finder's certificate one day...
You're more likely to get one than me since I have just one lowly 550Ti doing BRP4 tasks :-). It has a cache of 1.25 branded tasks (running 2 at a time) so I suspended them all to allow a couple of 1.28s to start. The first pair have just now completed and the improvement is very pleasing.
Run time has gone from approximately 5600s to 4100s and the CPU time from approximately 1500s to 1100s. The card is in a host with a Q9400 quad core with one core free.
I'm very tempted to set up AP to run the remaining 1.25 tasks with the new app but it's getting late here and I need to go home. Maybe in the morning if I get really keen ...
Don't need AP for that - provided you have the new app's and associated files already present, as you do, you can simply change the version numbers in client_state.
I don't recommend that you do this unless you really know what you're doing (as Gary does), but...
Compare the blocks for your v1.25 and v1.28 tasks, paying particular attention to , , and . If they match, a simple replace from 125 to 128 will do the trick.
You need to replace that string in both the and sections for cached tasks, but avoid changing it in the description - you can't do a complete global replace. It's probably best to report completed tasks first, and avoid 'rebranding' (as we call it at SETI) tasks in progress. Following Gary's lead, by suspending v1.25 tasks first and starting the v1.28 tasks, would ensure that.
I'm very tempted to set up AP to run the remaining 1.25 tasks with the new app but it's getting late here and I need to go home. Maybe in the morning if I get really keen ...
I had a similar temptation so I just wait until running tasks have finished and then reset the project. After resetting boinc resent all remain tasks with 1.28 version marks.
RE: RE: Hi! The app
)
Hi, Bikeman
I don't think it would be cause by an old version of ATI driver, because I just bought it no more than 3 month ago. It is a brand new 3rd gen i7 core laptop. The driver should not be too old to make such a crash. I subscribed SETI@Home as well. And there are tasks using ATI opencl app ( AstroPulse v6.04 opencl_ati_100 ). It works fine so far. Not sure if it will crash as well as it is still running.
So sorry to hear that no log could be find. Thank you all the same.
RE: Hi! Especially for
)
In the past when I tested older laptops by running Einsteins or LHC/T4T they did have those problems but the quad-core I am on right now never has any problem and it runs 2-core T4T,LHC,and 2 Einstein BRP4 Cuda's 24/7 with no heat problems and of course no fan problem since June 15th
Now this GeForce and processor isn't quite the same as my desktops with nVidia 550ti overclocked or 660Ti superclocked.
http://einsteinathome.org/host/5448315
So far this is a nice laptop to have.
To be complete here, Sumtec
)
To be complete here, Sumtec has since his post posted at Seti that his system BSODs when running Astropulse tasks on his AMD. I'm asking him to uninstall his Intel HD Graphics 4000 drivers, as I fear that these interfere with the OpenCL drivers on his system (I am investigating multiple similar situations for the BOINC developers).
Thanks Jord, that was
)
Thanks Jord, that was valuable information. Let's hope this will be fixed eventually when BOINC supports Intel OpenCL.
Cheers
HB
Reading NVidia's
)
Reading NVidia's 306.02-win8-win7-winvista-desktop-release-notes.pdf, released 28 August 2012, it looks as if NVidia is not immune to this problem.
There are numerous notes (mostly for Windows 8) like:
[Optimus notebook]: When attempting to install the NVIDIA driver over a prior driver version, the installation may fail with a blue-screen crash pointing to dxgkrnl.sys. [1000723]
This intermittent issue occurs on only a small number of systems.
• To avoid this issue, first remove the older NVIDIA driver using Add/Remove Programs in the Windows Control Panel, then install the new driver using NVIDIA setup.exe.
• If you encounter this issue, boot to Safe Mode, then remove the older NVIDIA driver using Add/Remove Programs in the Windows Control Panel. Install the new NVIDIA driver using NVIDIA setup.exe.
[RH - the 'Optimus' trademark refers to the dual technology where an NVidia GPU is teamed with an Intel HD graphics chipset]
Hi guys, Prior to this new
)
Hi guys,
Prior to this new app i was having problems getting anymore than 75% gpu load, this includes crunching simultaneous gpu wu's. I tried leaving a cpu core dedicated to the gpu units etc but again failed to get beyond 75% gpu usage.
It seems 1.28 has totally resolved the above issue & im now at 90% of gpu load with massive reductions in gpu crunch time.
The only issue with 1.28 i encountered was comodo firewall blocking it, a number of wu's exited with the error couldn't start CreateProcess() failed - Access is denied. (0x5): -148, this was easily corrected by creating a rule to allow the new app in the firewall.
Nice work ppl,
Jason
I've just reached the CUDA
)
I've just reached the CUDA v1.28 batch on my new GTX 670 (Kepler) - host 5744895.
Seamless transition, with runtime dropping 25% (from ~2400 seconds to ~1800 seconds) and CPU usage dropping 50% (~700 seconds to ~350 seconds). I'm running two tasks at a time, so that's closing on 100 tasks per day - what was the speed limit again for this project?
Seriously - great job, no problems at all. You won't see quite that throughput, because I run other projects on the card too, but maybe I'll get a finder's certificate one day...
RE: ... but maybe I'll get
)
You're more likely to get one than me since I have just one lowly 550Ti doing BRP4 tasks :-). It has a cache of 1.25 branded tasks (running 2 at a time) so I suspended them all to allow a couple of 1.28s to start. The first pair have just now completed and the improvement is very pleasing.
Run time has gone from approximately 5600s to 4100s and the CPU time from approximately 1500s to 1100s. The card is in a host with a Q9400 quad core with one core free.
I'm very tempted to set up AP to run the remaining 1.25 tasks with the new app but it's getting late here and I need to go home. Maybe in the morning if I get really keen ...
Cheers,
Gary.
RE: RE: ... but maybe
)
Don't need AP for that - provided you have the new app's and associated files already present, as you do, you can simply change the version numbers in client_state.
I don't recommend that you do this unless you really know what you're doing (as Gary does), but...
Compare the blocks for your v1.25 and v1.28 tasks, paying particular attention to , , and . If they match, a simple replace from 125 to 128 will do the trick.
You need to replace that string in both the and sections for cached tasks, but avoid changing it in the description - you can't do a complete global replace. It's probably best to report completed tasks first, and avoid 'rebranding' (as we call it at SETI) tasks in progress. Following Gary's lead, by suspending v1.25 tasks first and starting the v1.28 tasks, would ensure that.
RE: I'm very tempted to
)
I had a similar temptation so I just wait until running tasks have finished and then reset the project. After resetting boinc resent all remain tasks with 1.28 version marks.