Discussion of ideas
parts analysis of rbpi idea: PoE idea: good one. not using PoP RAM: good one. Gigabit ethernet, not practical. requires CPU that has Gigabit Ethernet. that means almost exclusively Marvell CPUs, available only under NDA, and expensive. on-board WIFI: requires RF expertise, extra cost (NREs as well as product cost). jeapordises project as a mass-volume product. overall: product idea would be expensive and costly to develop. whereas if done as a Motherboard instead, there is less risk. PoE could be done on Motherboard. RF/WIFI: again on motherboard.
regarding LVDS idea: this can be done using a TI LVDS83b, it's a very common IC. we thought about having on-board LVDS (instead of 24-pin RGB/TTL) but: a) if you have only a 320x240 LCD you cannot get low-cost 320x240 LCDs with LVDS, they are all RGB/TTL. if you have a 1024x768 or 800x600 or even 1388x800 you're still ok with one LVDS channel, but when you go to 1440x900 you need 2 LVDS channels, and with 1920x1080 or greater you now need three LVDS channels (which is 30 pins - more than 24-pin RGB/TTL). overall it's simpler and cheaper to do 24-pin RGB/TTL and let the motherboards sort it out from there.