It's amazing to believe that something as complex as the game of chess can be made into one big mathematical equation. Computers have no sense of a visual board but they can evaluate a large mathematical matrix of variables. Programmers have to give a numerical value to each of the following variables: material, control of squares, and king safty. A balance must be found between these three elements. For example, it's not uncommon for a pawn to be sacraficed in the opening in order to develope the pieces faster, thus, material can be out weighed by control of squares. Each of the variables must be evaluated for however many moves the program is allowed to think ahead to find the move that leads to the best possible numerical future. In other words, you're just playing a giant calculator that's looking at thousands of positions a second.
Computers do not use artificial intelligence to play chess, they use what's called a brute force algorithm. It just looks at every possible move it can in the allocated time given. I feel proud when I see 6 moves ahead with fuzzy accuracy, but remember, you ignore moves that common sense tells us are pointless. The computer searches everything, no matter how ridiculous, and assigns it a point value several variations deep. When thinking of it this way, we skip thinking about thousands of pointless variations the computer looks through with simple logic. Don't let the position count the computer boasts scare you.
The computer is at a severe disadvantage in that it cannot see long term positions such as end game pawn structure or sometimes long term danger in king side sacrafices. Take for example Eduard Nemeth, a 2100 rated german who beat the 2700 rated computers with a ridiculous openings and knight sacrafice. He played a few that just put the computer into a closed position where the computer thought it was doing better positionally but had no productive moves to penetrate the position so it just kept using it's brute force algorithm and lost on time.
Computer have prevailed as world champion, but considering IBM made a super computer that evaluated 200 million moves per second, it's a wonder Kasparov did so well. The world champion and many other super grandmasters today have a team of "analysts" which are probably aided by super computers. Probably the last "talent" was bobby fischer because he did his own home analysis. By "preparation" the grandmasters mean deep opening memorization to where they know in any variation how to get the upper hand.
Deep fritz 8 vs Deep shredder 9, Hydra vs Shredder? Who knows really. A lot depends on the processor speed so many of the games on chessgames.com by engines do not really say which engine is better or worse. The most recent programs will find excellent moves definitely GM quality given the time. For analyzing your games they are probably better than a grand master and they are a fine replacement for getting crushed at a masters challenge.
Chessmaster offers a variety of different styles of play and different weaknesses with even character profiles that make it seem a little more human. It has built in training lessons thatcan help any strength. Fritz on the other hand is more for pros. You can set a limited depth, but you can do that with the free crafty engine and likely get the same results. It has training modes for 1400 and up called "sparring" to where a light blinks when there's a good move. Both have databases of games, a collection of useless boards, and offer chessy midi style music for background. They are both excellent programs but fritz seems the stronger engine especially for opening and game analysis.