Question : Problem: Are $100+ motherboards worth the premium?

In general, I can always find a motherboard for my servers that is about $50-$75 with everything I want (AM2, 8GB RAM, 1GB NIC, onboard Video, SATA RAID, etc...)

Why would I pay 2x-3x the price for a more expensive motherboard?

Are they worth it? Why?

Are they faster? How?

Answer : Problem: Are $100+ motherboards worth the premium?

I'll answer your questions by number ==>

1 ==>  Every memory chip on an unbuffered module [NOT each module ... each memory CHIP on the module] presents an electrical "load" to the address and data bus.   This cause some distortion of the waveform, and the higher the load, the more distorted it gets.   Many desktop boards slow down memory access when 4 modules are installed, or add an additional wait state to accesses to allow more time for signal stabilization.    I NEVER install more than 2 modules on a desktop board => so, for example, with the board you noted, I would limit the installed memory to 4GB (2 x 2GB modules).    Buffered RAM modules (used in virtually all server boards ... but not common on desktop boards) have an electronic buffer between the chips on the board and the address and data bus => so they present ONE load per MODULE to the bus, rather than the EIGHT to SIXTEEN loads per MODULE that a desktop board presents.   So this results in a much more reliable memory subsystem ... and also allows boards that support MUCH more memory than desktop boards (I've seen several boards that support 16 modules).

So your question, "... what kind of improvements am I going to see with this? " ==> A more reliable memory subsystem.

2 ==>  ECC is, as you already know, error correcting memory which will automatically correct for single bit errors.   The result is fault tolerance.   I agree that modern memory is very reliable ... but it's even more so if you use ECC.   A reasonable expectation for uncorrected soft memory errors (that ECC would have simply corrected) is about 1 bit error per gigabyte per month.   That's not many ... but an error on the wrong bit could certainly cause some havoc.   ECC simply eliminates these errors in virtually all cases.

So your question, "... What are benefits of ECC? " ==> A fault-tolerant and more reliable memory subsystem.

Side Note for 1 & 2 ==> I think it's safe to say that MANY "Windows" errors are really just memory errors that cause Windows to malfunction (crash; hang; etc.).   I've found that my systems with ECC memory have far fewer "weird" crashes than those that don't.

3 ==>  "... what makes a chipset better than the next? " ==>  Better integration of functions; faster (and more) communications path between the various controller elements (memory controller, bus controller, I/O controller, etc.);  support for multiple GPU's (not a factor for you); more stable drivers (the best chipset makers ... Intel & nVidia ... are very good at providing stable chipset drivers); and better control features [this can allow the BIOS to provide more flexible hardware control operations].

4 ==>  Capacitor-based failures have become so well known that many manufacturers are advertising when they use higher-quality capacitors on their boards, so you may simpy be able to tell from their advertisements (taken, of course, with a grain of salt).   But in general it's true (as you noted) that the higher cost boards have better quality capacitors (and other components as well).

Regarding your comments on RAID ==>  If you're happy with the performance of Windows-based software RAID, then that's fine.   It works well, and does provide excellent protection.   And the overhead is not all that bad as long as you stay with simple mirrored arrays (rather than a RAID-5 with the inherent computational overhead).   Most large RAID arrays on servers are very performance-oriented, so a dedicated RAID controller is a much better idea ... but if that doesn't apply to your applications, then what you're using is fine.

Reliability ==>  Reliability is often in the eye of the beholder.   If you're satisfied with the reliability of the inexpensive systems you're using, that's fine.   The question to ask is what are the consequences of a system failure??   If you're superbly backed up, and downtime isn't a significant concern (within reason of course), then for your particular applications that's probably fine.   You're certainly not the only person who's using an inexpensive board designed for desktop use as a server :-)

The ONE thing I would do if I was going to use a relatively inexpensive desktop board in a server capacity is to be sure I used a board that supported ECC memory ... and only install 4GB (2 modules of 2GB each).  That single feature will make it a much more reliable system.