PHD papers in Computer Science
My understanding is that, similar to user3780968's answer, CS is a different animal than other areas. But even within CS it may differ. In some of the newer, more applied areas-particularly those in the AI-related fields-conferences are generally better, if only because those fields are evolving so rapidly. In more theoretical work, especially theoretical fields that have been well studied for decades (complexity theory comes to mind here), journals may be a better bet.
Part of the usefulness of conferences is that they generally accept fewer submissions and they allow you to present your work to your colleagues. This is like a PR campaign. In a fast innovating area, even the best ideas can be overshadowed by mediocre ones if they are not picked up on by the community at large (a la VHS and Betamax). In a journal, you may end up on page 945 and nobody will know unless they are looking for your specific topic.
The way I view it is: if your work is novel in a rapidly growing/changing area, conferences are your best bet to get your ideas out there and known now. If that's not the case, then there is nothing wrong with publishing in a prestigious journal. It really just comes down to the circumstances of the topic at hand.