With FB-DIMMS you can run the memory a bit faster with less problems and more secure data transfers and at higher capacity (i.e. more memory). There is much more to it than this, physically, I am hoping someone with more knowledge can help out. David C1??? -- you seem to be more knowledgable in this area.
The problem with DDR/DDR2 is that its hard to have enough memory capacity, and that's a must for servers. I read that when there was DDR200, four DIMMs a channel was possible, but with DDR400, it went down to two DIMMs. Likely, DDR2-400 allows four, but DDR2-800 and speeds between will support lower amount of DIMMs. The problem is well explained here:
Part 1:
http://www.theinquirer.net/?article=15167
Part 2:
http://www.theinquirer.net/?article=15189
Part 3:
http://www.theinquirer.net/?article=15214
Basically, it is the shared bus architecture of the DIMM itself, a stub-bus configuration. What is that? Imagine a comb, with the bus being the spine of a comb, a dram on each tine, and the controller on one end. Each point where the tine touches the spine, you get an electrical discontinuity. Each of these little discontinuities is like a speed bump, it causes the signal to degrade.
Just like those nasty pieces of roadway, each of these stubs causes more problems the faster the signal goes. The simple solution is to decrease the number of DRAMs on the bus, so that while cranking up the speed makes each discontinuity worse, it means there are fewer of them. Speed or capacity, pick one.
So basically, by adding a buffer that allows point-to-point connection, the problem with stub bus goes away. That means, you can have many more DIMMs than previously possible.
FB-DIMM also has extensive features to help with data reliability, so the differentiation between ECC and non-ECC now disappears.
It's also good for motherboard manufacturers, since the FB-DIMM DDR2 only requires 69 pin per DIMM. That means vastly reduced design time, and unequal wire traces could be used, unlike now, which needs equal wire traces to ensure signal integrity, and complicates motherboard design.
Basically its an ULTIMATE server memory standard.
For desktop users, it seems we can probably keep going with the normal DDRx and so on. Non FB-DIMM is probably better for us anyway since we don't need huge capacity DIMMs or want to sacrifice latency for capacity.
Now here's a possible advantage that people will care. My understanding is that, it may be possible to not change the memory controller and have support for new memory standards as long as the buffer supports it. In reality, it will probably be more complex, but it looks good.