GPU Compute error 65 (0x00000041) Unknown error code

Hi and welcome to Einstein@home!

The message about not being able to read the checkpoint file is normal when a new task starts as there can't be any checkpoint written until the task has run for a while.

In the log (stderr output) that you linked to the following lines stand out:

error in opencl_qsort
13:20:45 (3846): [CRITICAL]: ERROR: MAIN() returned with error '1'

Are you sure the graphics driver and especially OpenCL support is installed correctly and working?

Hi Chris,
Welcome to Einstein@Home from me too!

I agree with Holmis that the problem is the later message and this seems to be with allocating sufficient memory.  Your GPU shows as having 2GB and this is plenty so I decided to do a direct comparison of the relevant part of your stderr.txt output with the same part of the output from one of my AMD 2GB cards.  The first block below is yours, the 2nd is mine.

boinc_get_opencl_ids returned [0x25ba260 , 0x25c3fd0]
Using OpenCL platform provided by: NVIDIA Corporation
Using OpenCL device "GeForce GT 640" by: NVIDIA Corporation
Max allocation limit: 523927552
Global mem size: 2095710208
OpenCL device has FP64 support
% Opening inputfile: ../../projects/einstein.phys.uwm.edu/LATeah1061L13.dat
% Total amount of photon times: 8950
% Preparing toplist of length: 10
% Read 1631 binary points
read_checkpoint(): Couldn't open file 'LATeah1061L13_292.0_0_0.0_22111467_1_0.out.cpt': No such file or directory (2)
% fft_size: 16777216 (0x1000000); alloc: 67108872
% Sky point 1/1
% Binary point 1/1631
% Creating FFT plan.
% fft length: 16777216 (0x1000000)
% Scratch buffer size: 136314880
% Starting semicoherent search over f0 and f1.
% nf1dots: 41 df1dot: 2.512676418e-15 f1dot_start: -1e-13 f1dot_band: 1e-13
% Filling array of photon pairs

- - - - - - - - - - - -

boinc_get_opencl_ids returned [0x15207d0 , 0x7f35a3149430]
Using OpenCL platform provided by: Advanced Micro Devices, Inc.
Using OpenCL device "Pitcairn" by: Advanced Micro Devices, Inc.
Max allocation limit: 1399062528
Global mem size: 2071986176
OpenCL device has FP64 support
% Opening inputfile: ../../projects/einstein.phys.uwm.edu/LATeah1061L12.dat
% Total amount of photon times: 8950
% Preparing toplist of length: 10
% Read 1631 binary points
read_checkpoint(): Couldn't open file 'LATeah1061L12_292.0_0_0.0_9753380_1_0.out.cpt': No such file or directory (2)
% fft_size: 16777216 (0x1000000); alloc: 67108872
% Sky point 1/1
% Binary point 1/1631
% Creating FFT plan.
% fft length: 16777216 (0x1000000)
% Scratch buffer size: 136314880
% Starting semicoherent search over f0 and f1.
% nf1dots: 41 df1dot: 2.512676418e-15 f1dot_start: -1e-13 f1dot_band: 1e-13
% Filling array of photon pairs

As you can see, apart from differences that are due to different tasks and GPU makes/models, there is just one thing that stands out as a significant difference.  That thing is the Max allocation limit.  Yours is ~0.5GB and mine is well over 1GB.

This is just speculation on my part but since these tasks require a *lot* of memory to do the FFT (Fast Fourier Transform) stuff, (and this is exactly where the thing fails) perhaps your much smaller allocation limit is the actual problem.  Since you have 2GB, there should be a way of convincing the driver to allow a greater limit than the ~0.5GB that shows.  Maybe it's as simple as setting an environment variable.  A question to nvidia or perhaps even a google search might yield something.

Have you tried running clinfo to see what that utility thinks about the OpenCL parameters/capabilities of your GPU?