Did you use RT-PCR and gel-running to check your actin primers? After you have cDNA you could you your primers to do normal PCR and then do gel-running to se your PCR products. If all your samples show the same product (size and intensity) as you use the same concentration of cDNA for PCR, it mentions your primers are good. If not you need to re-design your primers. You should do some replicates to confirm your results.

Slight variation in endogenous control is acceptable however use of multiple endogenous controls will give you more accurate result. Find out endogenous genes from literature of ur interested organism and find out more stable control for your experiment. 

refer this article

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0078555

The most stable and highly housekeepers tend to be species and tissue specific but in the absence of literature for your model the above 3 are commonly used

Does the variation you see make sense in relation to the experimental set-up?  I mean do you see changes in beta actin in only your test samples and no changes in your negative control?  If you see changes in only your test samples, then your gene of interest may directly affect beta actin expression.  In that case, you'd want to pick a different housekeeping gene to run, such as those mentioned by Laurence.  If the changes you're seeing ARE NOT consistent within your experimental set-up, there are a number of things that could be occurring.  First of all, how many wells are you running per sample?  I typically recommend 3 wells per sample, so that any slight inconsistencies in pipetting or whatever can be accounted for.  Wells can be averaged before final analysis of CT, as long as the results are relatively similar.  Second, are your samples normalized to a set concentration before running?  I usually normalize my RNA to 1ug before converting to cDNA to ensure that when I load each sample, the results are immediately comparable since the concentration for each is the same.  Third, can you give an example of the variation you're seeing within or between your samples?  Some variation between samples is expected as no one, not even a robot, can pipette every sample perfectly. 

Debes revisar la literatura relacionada con el tema, y verificar que este gen si sea apropiado para este caso, los genes normalizados que utilices deben ser estables en el modelo que utilizas para que te sirvan como control. Si lo anterior es positivo, revisa el diseño de los cebadores y las condiciones de amplificación y finalmente revisa que estés utilizando una cantidad homogénea de mRNA par tu experimento.

That depends of your conditions, you have HSP90, 18S, HSF1 try to see what gene fits better to your experimental conditions