I was looking for an economical way to speed up rendering(presently only having an i7-860) so I put together a barebones system with one of the new 8-core chips this weekend. Seems to work pretty well with Brazil, for the money I mean(the initial "updating database" and updating ray server, etc phases are no faster, maybe even slower.) The fact that there are only 4 actual floating-point units doesn't seem to hurt much. I notice that the speedup isn't proportionately as much with Neon, though the bargain video card might have a little to do with that even if it's not actually using the GPU.
If we just had network rendering I'd get another one or three and really see some productivity gains.
I've also been considering an 'upgrade', but sticking with AMD and it's new 8-core chips (currently running a Phenom quad-core). Are you actually seeing a linear (x2) performance in going to the new AMD 8-core system?
I don't know, I haven't had an Althlon since before the Phenom, these rendering benchmarks have the 8350 and some older ones... http://www.tomshardware.com/reviews/core-i7-3970x-sandy-bridge-e-be.... It would seem to be slower clock-for-clock than the Phenom IIs.
On a simpler part of the project I'm rendering out now, where the not-so-multithreaded startup steps dominate, it's actually slower than my i860. Which is three years old. I'd build a render farm out of them but I certainly won't be using one in my main desktop.