Many Errors

Hi Bob, Welcome back to Einstein!

The easiest way to get a better idea of the cause of a compute error is to pick a failed task in the tasks list for the host in question on the website.  If  you click the Task ID link for the failed task, it will take you to the full stderr.txt output that was sent back to the project for that task.

I've looked at some example errors from all 3 hosts and something like the following is pretty common for the error condition.  This one is from the 570 but I saw similar memory allocation failures on the other two as well:-

% Starting semicoherent search over f0 and f1.
% nf1dots: 41 df1dot: 2.512676418e-015 f1dot_start: -1e-013 f1dot_band: 1e-013
% Filling array of photon pairs
Error in OpenCL context: CL_MEM_OBJECT_ALLOCATION_FAILURE error executing CL_COMMAND_NDRANGE_KERNEL on GeForce GTX 570 (Device 0).

The 570 may be too old and have not enough available memory, 'available' being the key word.  If you happen to be tying up memory elsewhere or attempting to run multiple concurrent GPU tasks, you should adjust things to see if more available memory will solve the problem, particularly for the two more recent GPUs which do each have 2GB in total and really that should be quite sufficient to at least run tasks singly.

I use Linux and only AMD GPUs so don't have relevant experience with either Windows or your type of GPU.  It seems pretty clear though that you somehow need to make more memory available to the app in order to allow all tasks to complete.  As well as the huge number of failures, you do have validated tasks for the gamma-ray pulsar app so stable operation should be possible.  To test that, you should change prefs to allow that app only with single tasks until you get something that works reliably.  That app will probably have lower memory requirements than the GW app.

Hope this helps!