Unknown Error Using gpuArray

When I try to create a gpuArray, I get the error:

Error using gpuArray

An unexpected error occurred on the device. The error code was: UNKNOWN_ERROR.

There is a previous question regarding this. However, the given solution didn't seem to work in my case. The main suggestion was it might be a kernel timout, but when I run gpuDevice, I get that KernelExecutionTimeout is set to zero. My understanding is that if this is set to zero you shouldn't have any timeout issues. The older post also suggests that it might be a memory problem. I'm using AWS EC2 and the instance I'm using has 732 GB of memory. Is this really not enough? I have a fairly large matrix, but this seems like a lot. Additionally, I've run gpuArray on a small array (3x1) and that worked fine, so it seems like it might be the size. Any suggestions on debugging this would be really helpful!

[Joss Knight]

Hi Melissa. There are no GPUs with 732 GB of memory so perhaps you are referring to the CPU memory? What does gpuDevice say about TotalMemory?

There are lots of reasons you might get UNKNOWN_ERROR so you will need to tell us your AWS instance type, your MATLAB version, and give us a snippet of code that reproduces the problem.