Task runs for 3-4 seconds, the GPU crashes and restarts, but the task does not resume. Will only restart if I shut BOINC down, shut down the V8 task manually in Task Manager and then restart BOINC manager. Then get the same result.

---

You have only 3 tasks "Error while computing" and they are not 3-4 seconds:
http://setiathome.berkeley.edu/results.php?hostid=7904629&offset=0&show_names=0&state=6&appid=29

You have 11 Valid tasks (and 4 that finished OK and are "pending")
http://setiathome.berkeley.edu/results.php?hostid=7904629&offset=0&show_names=0&state=4&appid=29

(All of them are GPU tasks: SETI@home v8 v8.00 (opencl_ati5_cat132))


If you now see what you describe (with currently running tasks) - first try to Reboot the computer

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Rebooted. Tasks start to compute and after some time, I suddenly get a message that the GPU driver crashed but successfully recovered. This appears to be random, as some GPU SETI tasks do run to completion. The SETI V8 GPU tasks stop when the driver crashes. If I'm not running BOINC, the display driver for the GPU is stable, so I'm assuming the problem with with BOINC and the SETI V8 GPU drivers.

I'm running 64 bit Windows 10 Pro. The display card is an AMD Radeon 5700 HD running Catalyst 15.7.

I have no other applications running in background.

I may have to just stop accepting GPU tasks for SETI.

---

so I'm assuming the problem with with BOINC and the SETI V8 GPU drivers.

There are no Seti v8 GPU drivers.
There's BOINC as the managing program and Seti v8 for the science program that does calculations on your GPU.

For this you do need sturdy drivers, but those drivers come solely from the GPU manufacturer, in this case AMD. You shouldn't use the drivers provided by Windows, but always only the drivers from the GPU manufacturer.

Problems with the graphics card drivers restarting are usually attributed to trouble with the videocard/GPU, be it heat (cleaned out the dust lately?) or damaged hardware. Running OpenCL calculations on a GPU stresses the hardware more than just sitting in Windows does. Gaming will stress the hardware harder as well, depending on the game.

So essentially, good hardware, clean hardware, sufficiently cooled, and with good sturdy drivers are a requirement here. What you can try still is to uninstall the drivers completely, then run Driver Sweeper in Safe Mode to clear all remnants of the drivers, reboot, install Catalyst 15.7.1 and see how that fares.

If still problems after that, try older drivers. 14.4 or 14.9 are a good bet.
If still problems after that, perhaps you need a better power supply unit.
If still problems after that, you really ought to ogle your videocard and check it doesn't need replacing with something newer.

Plenty of us who run work on Windows on our ATI/AMD GPUs without having the display driver reset every hour, so if we can do it, so can you.

---

I may have found what is triggering my problem. For Christmas, my wife gave me a new monitor. I'm a writer and kept the old monitor and started running both using an extended desktop. I'm only getting GPU resets if I'm in extended desktop mode. If I turn it off, the SETI task recovers and starts running again.

At least at night I can shut off the extended desktop feature so SETI can work in peace....

Thanks for all the advice.

---

 
You can try a Setting to reduce lag and driver restarts

Go to SETI@home directory (<BOINC_Data>\projects\setiathome.berkeley.edu\)
If you don't know where is <BOINC_Data> directory - look the first lines in your Event Log (Ctrl+Shift+E) - it is shown as "Data directory:"

In ...\projects\setiathome.berkeley.edu\ you will see some .txt files with zero length (0 bytes, empty)
Find the one with similar name (your may have different name but will start with mb_cmdline):
mb_cmdline_win_x86_SSE2_OpenCL_ATi_HD5.txt

Edit the file with Notepad and put the following line in it:
-period_iterations_num 200

'Snooze GPU' for a few seconds to enable the new Settings

P.S.
The default value is 50 as you can see in stderr.txt
http://setiathome.berkeley.edu/result.php?resultid=4671593243

Used GPU device parameters are:
Number of compute units: 10
Single buffer allocation size: 128MB
Total device global memory: 1024MB
max WG size: 256
local mem type: Real
LotOfMem path: no
period_iterations_num=50


Lower values make the GPU work harder
Higher values like 200 will Reduce the load (reduce lag and possibility of driver restarts)
You may try even 300 or 400


! If you want to return to the default - make the file empty again, do Not delete the file!
 

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I have similar problems.

Win 7 64bit, HD5450 Graphics, Catalyst 15.7.1 installed.
A "Seti@home v8 8.00 (opencl_ati_cat132)"-task crashed the display driver.
I let it run about 20 times, the driver always crashed at 3,529%.

Downgrading display driver to Catalyst 13.12 solved THIS problem.

Task is now at about 36%, but the computer is lagging a lot while
the GPU-task is running.
This seems to be worse when 2nd display is activated.

A view on my account shows, all GPU-tasks (done with 15.7.1) I can see seem to have calculation errors.

---