Why is Spark's show() function very slow?

I have

df.select("*").filter(df.itemid==itemid).show()

and that never terminated, however if I do

print df.select("*").filter(df.itemid==itemid)

It prints in less than a second. Why is this?