I am getting a many GPU task starting to error out with the following

<core_client_version>6.12.33</core_client_version>
<![CDATA[
<message>
Incorrect function. (0x1) - exit code 1 (0x1)
</message>
<stderr_txt>
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
Using SSE2 path
Found 2 CAL devices
Chose device 1

Device target: CAL_TARGET_770
Revision: 2
CAL Version: 1.4.1457
Engine clock: 625 Mhz
Memory clock: 993 Mhz
GPU RAM: 1024
Wavefront size: 64
Double precision: CAL_TRUE
Compute shader: CAL_TRUE
Number SIMD: 10
Number shader engines: 1
Pitch alignment: 256
Surface alignment: 256
Max size 2D: { 8192, 8192 }

Failed to map resource: Operational error (CAL_RESULT_ERROR)
Failed to zero output buffer: No error (CAL_RESULT_ERROR)
Failed to create output buffer: No error (CAL_RESULT_ERROR)
Failed to map resource: Operational error (CAL_RESULT_ERROR)
Failed to zero output buffer: No error (CAL_RESULT_ERROR)
Failed to create out streams buffer: No error (CAL_RESULT_ERROR)
Failed to create buffers: No error (CAL_RESULT_ERROR)
Failed to release CAL resource: A handle parameter is invalid (CAL_RESULT_BAD_HANDLE)
Failed to release buffers: No error (CAL_RESULT_BAD_HANDLE)
Integral 0 time = 1.906686 s
Failed to calculate integral 0
03:32:58 (27728): called boinc_finish

</stderr_txt>
]]>


Any Idea of what the fix is. Just started.

Have you reboot the system?

---

One of the first things I have done is a complete shut down and a cold restart. The problem still was present. So I deleted the ATI drivers and download new drivers from ATI and did a fresh install. I deleted then downloaded a new copy of ATI SDK and installed ATI SDK. I then did another shut down and cold restart. The system will process both Milkyway and Prime tasks. But every now and then it will just stop processing the Milkyway task, the Prime Tasks continue to function. When this happens I will shut down BOINC. I sometimes will see an orphaned Milkyway task in memory. I will have to manually kill this task. I will then restart BOINC and everything starts to work.

One of the first things I have done is a complete shut down and a cold restart. The problem still was present. So I deleted the ATI drivers and download new drivers from ATI and did a fresh install. I deleted then downloaded a new copy of ATI SDK and installed ATI SDK.

Which version of the drivers are you using? With the more recent ones you should not need the SDK, I think... at least I didn't need it IIRC, I'm on v11.2.

The system will process both Milkyway and Prime tasks. But every now and then it will just stop processing the Milkyway task, the Prime Tasks continue to function. When this happens I will shut down BOINC. I sometimes will see an orphaned Milkyway task in memory. I will have to manually kill this task. I will then restart BOINC and everything starts to work.

That can be actually anything...

How many tasks do you run per GPU? Don't know how much RAM Prime needs, maybe the GPU runs out of it? You can check that with ATI Memory Viewer, google for it.

Also check the temperature of the GPU... and whatever else has a sensor on your graphics card.

---

Each of the Card has 1GB of memory. At most I am using 75 percent of the Card memory. Each of the Cards is only running one GPU task. The ambient temp is approx 20 C, and the cards are between 60 and 70 C.

It seems to get worse if an when I suspend GPU operations from inside Boinc.

I have started to just shut down BOINC, when I need to perform some other task that are graphic intensive, and the problem does not happen.

It is looking more and more like a software issue.

Perhaps not GPU memory but possibly running too low on system memory. With two 1GB video cards, 3GB of available memory and 6 CPU cores crunching on 4 CPU projects you may get buffer creation errors on GPU processing. You may need to do one lot of 6 CPU tasks at a time and not let BOINC switch between different CPU projects/tasks and/or only choose CPU projects that have a lower memory requirement. Keeping "Waiting to run" CPDN, Einstein, Malaria and Rosetta CPU tasks in memory while processing MilkyWay and PrimeGrid GPU tasks may be the problem.

Alternatively you could add more memory to your computer and switch to a 64-bit operating system.

You can have insufficient system memory for BOINC GPU processing even when Task Manager shows there is free memory available.

This is just a possibility, it may be another problem associated with running intensive video applications at the same time as BOINC GPU processing.

I think i have kindly the same problem:

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2548#50574

"trottled" with app_info and since calculation errors every time.

First thanks to everyone who has put forth a response to this issue. At this time my plan of action is to reduce the number of process that stay resident in memory instead of 80 percent of memory be used the system now sits between 55 and 62 percent of memory being used. All Boinc task are configured to only be in memory while they are running. Additionally the GPU card are configured to their factory defaults, with one exception the fans are locked at 95 percent of full speed. Where the system is noise is not an issue. I plan to run the system and see what if any issue pop up. Kashi comment about video buffer memory issue at this time makes the most sense.

I will keep this thread informed of the results.