You can choose any gene that does not change between your samples with the treatment you are using. No gene is universally static though, so we make a guess as to which gene is most likely to remain stable during the course of our experiment. Actin and beta-tubulin and 'housekeeping' genes are very commonly used, but be aware that a good control gene in one experimental condition may not be the best in a different one. Be very careful to load the same amount of RNA/cDNA in each tube and go through a selection of genes until you find one that is reliable. You should also choose a gene that is expressed on a similar level as your target.
I agree with all others - it is important to normalize qPCR with a proper control gene(s) which do not change expression in your particular setting. Preferentially, one would test several genes and choose the best ones.
Here is also one useful link with description how to do qPCR experiments (based on Vandesompele et al article cited by Hew Torrance):
You must test the genes using your specific experimental conditions and their expression must be stable so that they can be used as control genes. If they are both ok, you can and should use both.
The choice of the control gene also depend of the level expresion of the gene of interest, so if this gene is highly expressed is preferred to use actina but if this gene is poorly expressed is better to use GDPH