If you want to design your own, I recommend Primer-BLAST from Pubmed. You can also find sequences with some QC at PrimerBank: pga.mgh.harvard.edu/primerbank/ 3rd option: buy predesigned assays from commercial suppliers. I have tried them all and none is 100% guaranteed. As Junaid said, you may find your primers in the literature, but BLAST them anyway, because typing errors in published sequences may occur.
Below are not the reference genes you asked for, but I have tested many. GAPDH can be quite variable. Ideally you also wish to have reference genes involved in different process. For example Tubulin and Actin are both cytoskeletal. As a rule I would normalize to at least 2-3 ref genes if you can as you suggest.
I would try the following as a start (all human sequences and validated efficiencies across high dilution range) using qBasePlus. We use these for our leukemia work and have used in the past these in neuronal differentiation expts, breast cancer and melanoma work . All work in the different cellular models, but best to test in your differentiation model to ensure their levels stay constant. The sequences are as follows:
1) TBP:
F:CGGCTGTTTAACTTCGCTTC
R:CACACGCCAAGAAACAGTGA
2) GusB
F:GCC AAT GAA ACC AGG TAT CCC
R:GCT CAA GTA AAC AGG CTG TTT TCC
3) HMBS
F: gagagtgattcgcgtgggta
R:cagggtacgaggctttcaat
4) HPRT
F: TGACACTGGCAAAACAATGCA
R:GGTCCTTTTCACCAGCAAGCT
5) RPS18
F:TAG CCT TTG CCA TCA CTG CC
R:CAT GAG CAT ATC TTC GGC CC
Please blast these so you are sure they meet your requirements in terms of Tm, Size of product, span intronic regions etc. I would plug them in here so you can check your self:
http://www.ncbi.nlm.nih.gov/tools/primer-blast/
Just paste in the Fwd and Rev sequences in the "use my own Forward..." and "use my own reverse primer..." Once both are pasted, hit blast.
One other consideration - if you are using the Roche primer/probe system. You may want to try all five and look for the one that is most consistent for YOUR sample. Our local Molecular Resource Center has primer/probe mixes available for researchers to optimize this for their experiment. Remember, different tissue types or time points in the same tissue type could alter the expression levels of a given housekeeping gene making it less useful as a comparison in that particular case. That being said - TBP has worked well for us in many tissues.