I see where you are coming from, Puiu, but there is a bigger picture if you step back... Intel has a massive rash of changing socket types and supporting chipsets. As you indicated, people often just buy new computers, which Intel is trying to guarantee will have their chipsets and CPUs in them (double-dip). If they supported the older architectures, they wouldn't necessarily make as much, since the stockpile of existing motherboards could be tapped into. By forcing constant updates and upgrades, they are creating a market for their chipsets by proxy, as the next model won't work with previous boards. Doing this adds a constant development cost and complexity to motherboard builders. (and yes, I know that some of the reason for the changes is due to advances in the bus architectures, but so many variations in a short time is ridiculous).
In contrast, look at AMD. They had the AM2, upgraded to the AM2+, now have AM3. Lots of backwards compatibility, plenty of familiarity and common parts bases, low stress for the motherboard manufacturers to keep churning them out. No requirements for new chipsets every 12 seconds. Gradual movement to newer memory systems (DDR3), rather than forcing you to jump in that respect too. It seems to be a more gentle philosophy, and makes you wonder if AMD didn't think well beyond the box to plan their path (and come up with a good architecture capable of longevity), while Intel is just hopping around making all these parallel innovations that constantly render their previous work irrelevant. Looking at the product paths, it's like AMD is a nice straight stretch of highway, while Intel is whipping around hairpin curves constantly... heh