Hola Rand.
Yeah, my articles at SLcentral tend not to be of quite the same technical depth as some others, but I might also point out that I've never once been paid
Actually, I'd say that two of my articles were as good as, if not better than (in some respects) some of the ones at Aces and Ars. I just don't have as many good ones ;-). The two that I'm thinking of precede their equivalent articles at Ars by over a year (in one case, about two), so I figure I have at least that going for me.
Enough of my trying to save my fragile ego. Go read Ars's "CPU Theory and Praxis" stuff. Hannibal writes good schtuff. Johan at Aces is pretty good too. Just read his a little more carefully -- english isn't his first language, so there's a slipup everynow and again. That said, his english is markedly better than most "enthusiest" sites', most of which are run by 'lil 20 year olds who were never taught how to speel or writ wel </rant mode>
For some other good stuff go to
www.realworldtech.com. Paul DeMone is a phenominal writer, knows his stuff, and they just got a David Wang -- who is a grad student at my school, U of Maryland at College Park -- who also really knows his stuff.
Go lurk the forums (and I do mean lurk) at realworldtech. Paul DeMone will occasionally get into flamewars with the likes of Bill Todd, and a few others, but he's (usually) right. David Wang is the consumate proffessional; I've never seen him get into a flamewar. The reason I say lurk is 'cause if every newbie who had questions posted, the place would get cluttered.
Comp. Arch is also a lot of fun. LOTS of flamewas over there though -- after a couple weeks, you'll be able to tell who knows their stuff, who's being a jack-arse, who is mostly interested in how programming relates to CA, etc.
Um...as for your original question: It's really complicated. The guys above gave some good advice. If you are really into it, do what I recommended above (because, even if it is passive, watching others interact and talk about this is a great way to reinforce the material in the books/sites mentioned by others and myself).
To give a somewhat direct answer....
AMD and Intel implement (mostly) the same Instruction Set Architecture (ISA) -- (well, at least up through the P4 and the Athlon -- the Athlon64 and Itanium families have very, very little in common). This means that you can send the same instruction to any processor running the same ISA, and each one will spit out the same answer. The difference between Intel chips, and AMD chips (and old vs. new chips from their respective companies) is in how they implment it. You can have more cache, less cache, better branch prediction, worse BP, longer pipeline with the goal of a higher frequency, shorter for a smaller mispredict penalty, etc. Think of designing chips as a GIANT profit function (that's why these companies make chips, afterall). All of engineering work is dealing with tradeoffs. Designing a chip is no different -- AMD and Intel choose different tradeoffs.
How's that for vague?
BTW -- I think that IBM has a fab or two that can compete with Intel ;-)