I'm not convinced that giving already released hardware as the main prizes is a good idea. From my point of view, people don't choose to compete for the main prizes like they do for sponsor ones, and forcing them to take an Oculus Rift or something of the like that they could just buy if they wanted seems a little suboptimal. However, getting unreleased hardware is a value add in my book, and was something we considered for the latest PennApps with Myo, although they couldn't make the delivery date in the end.
Sponsor prizes are tricky. One thing we've done at PennApps is encourage more vaguely themed prizes rather than ones that reward the use of a specific company's API. OTOH, we're not ready to entirely restrict API-centric prizes since I do believe that for less "sexy" companies it is a value add (the writers behind this article both came from arguably "sexy" companies), and the competition for sponsor dollars is such that the negatives mentioned in the article about sponsor prizes do not outweigh the cost of removing them for us.
A well-organized hackathon is fairly bifurcated. You have new hackers who are there to learn and experienced hackers who want to tackle an ambitious project. For both groups, I think large cash prizes are useful. For the newbies, even if they're not planning to win, the prizes reinforce the message that hacking is something useful and worth pursuing (helping to counteract the societal narrative of geeks being lame). For experienced hackers, prizes are a way of affirming that you've accomplished something amazing, particularly since most of the coolest hacks will never generate other compensation for the creators.
My personal reason for not liking product prizes is an objection to the accumulation of stuff. I really don't need yet another Arduino (won too many already) and would much rather use the money to take my family out for a nice dinner/etc. Most people don't need more things.
The idea of experience prizes (a trip somewhere) is a good one though.
I actually prefer doing something fun at hackathons - case in point, the recent create a music app for Slash hackathon during SXSW, where I ditched my original idea to play with Qualcomm's beacon technology instead.
My biggest complaint about hackathons is if the hackathon is actually judged as a business pitch-a-thon, and that part wasn't clear at the beginning.