Hey, not to worry, Poem is just spotty with GPU work availability. It does have work for NVidia and ATI/AMD via OpenCL. I just don't think their servers can keep up with demand. Check out this thread
for more complaints.
I usually set Poem as a backup to WCG. Every once in a while I'll get a big batch of work, but for the most part, the well is dry. At some point, they may be able to satisfy demand, but I'm not holding my breath. This spreadsheet
is relatively updated with ways you can use your GPUs. Anything CUDA is NVidia only. Anything CAL is ATI/AMD only. And OpenCL is for both, but it may favor one or the other depending on the project.