"While Microsoft was still involved with OS/2, the OS/2 development group requested that the company install a washer and dryer in their building so that they wouldn't have to go home to do their laundry. Although the group never got its washer and dryer, the message was clear: this team wanted to work. It didn't ask for promotions, more money, bigger offices, or fancy carpet; it asked for management to remove every conceivable roadblock so that it could concentrate on shipping a product."
Source: http://www.stevemcconnell.com/articles/art05.htm
Perks are an effective retention strategy only if they enable an already motivated employee to get more work done. And for the employee to be motivated, the work has to be rewarding.
Free food, on site gym, nice offices, etc I'd trade at a discount to salary, like$1 in salary for $2-3 in perks.
I also hope that a good PC with good monitors and a decent desk and chair should not be considered part of the perks.
With my friends I can't talk about the details of the awesome technical projects I'm doing at work-they won't find it interesting. But I can talk about how fun it is that we have tons of autonomy, about 30 different board games in the office that we play at lunch every day, a beanbag chair we can use for naps, and have a foosball table.
The lack of money and success create anxiety, but an abundance of these in terms of what an employee can hope to achieve do not influence my decision about where I work.
As for perks, I don't care about trivial asides that people think up in order to retain talent. This is an artificial device that will not stop the bleeding. Don't waste your time doing this if it's something you have to think up in order to achieve yet another metric. If you're going to do something fun do it because you want to do it because you yourself would enjoy it and know that at least a few of the others on your team would too. But then you're not doing it for retention, you're doing it because you want to.
Bottom line is if the factors don't add up, an employee will leave, but things like people, technology and product are what candidates should consider before they join, and they're not likely to change much over the duration of an employment, so if someone is leaving, they probably didn't know what they were getting themselves into, something big changed, or they think they are really failing.
Or maybe they just found a new thing that interests them, in which case let them go.