Scintilla is correct. It's the fabrication process used. Specifically in reference to transistor width (I believe the gate width?).
Generally the lower the nm, the less power it *should* dissipate for a given frequency. Also transistors *should* switch faster.
If possible, get a 90 nm since it's obviously a newer revision of the silicon and should run a little cooler.
At the same time, I realize a 130 nm is probably comparatively less in price so I'd certainly understand. Also it's not like cooling these days is *that bad* and you can't keep a 130 cool anyway.
Also don't take this as me saying 130 nm is bad or anything. I like it a lot from a work standpoint since it wasn't as difficult to debug

.