The latest version of Rybka is a commercially-available, mass-market program. It has a rating over 3000 ELO and can beat all the super grandmasters without much problem. You could run it well on any decent dual-core or better machine.
How good are the chess programs?

Meadmaker, I believe you are underestimating the power of computers. I don't think any human could ever hope to beat Fritz 11, nevermind Rybka 3. They have surpassed us completely. It would be big news if someone managed a draw against one- probably using an anti-computer strategy.
Computers are definitely stronger than any human. However, they may not always beat every human. Take a look at these pages: http://www.xs4all.nl/~timkr/chess2/honor.htm
http://www.chessgames.com/perl/chessplayer?pid=91647
Unfortunately, they're out of date (I think around 2001 or so), but it's still interesting how this guy can beat Fritz and how Fritz can drop a Queen.

Take away their opening book (which contains centuries of collective human knowledge) and engines would be lucky to win a single game against a Super GM.
Now wipe the Super-GM's memory of all openings.
And normal service is resumed.

Take away both players opening books -- use one of the chess960 positions -- and who wins?
Computer.

the real weakness behind a chess engine is this http://www.chessgames.com/perl/chessgame?gid=1497429&kpage=6#reply145n
engines don't fell that's what a gm like nakamura takes advantage of

the funny thing is that he does that in a blitz game where computers like rybka 3 have a really good reputation

If a computer had the same "handicaps" as a human, well then what? Let me explain. I am just now coming back to chess. I used to know openings, tactics, endings. 30 years later -- I am learning again -- albeit faster.
Take a computer, let the AI know the basic rules and 4 openings. That is about where most players start. It has to play a named opening or ending, etc. 3 times before it can put it in the database (how many times does a beginner lose before THEY remember).
The AI can have the knowledge of one book every two months of playing and can only play one or two games a day (who has time?) Soon the computer would "lean" toward an opening or a tactic and a GM would wallop it.
But today's computer engines have the knowledge of a library built in and the knowledge of some of the best games ever played. No hope for even a good play - it is not on even footing.
Take a dual-core Mac or PC and with only basic rules, most of us could beat the computer. My Mac Computer app I can beat on easy and sometimes on medium. On easy, the computer does not even consider harder things like castling and En passant.
The problem is, the computer starts as a super GM with no work. If it had to learn, I do not think the AI could manage wins against titled players for years and years, and would lean towards openings, etc like regular players that titled players would easily take advantage of.
So, it wouldn't take Deep Blue to knock off the world champion anymore? A regular old PC running mass market software would beat any human?
I would like to think that they would play predictably enough that the top dogs of the Chess world would be able to beat them, but the consensus seems to be no.
At least, tell me they haven't programmed them to use Arnold Schwarzenegger's voice.

I think all of the GMs have quit playing against the computers It is not a matter of "thinking" anymore like Deep Blue had to -- it is a matter of memory of millions of games, opening, endings, tactics and strategies. A GM simply cannot memorize what a computer can.
If I open against my computer program with e4 I ALWAYS get the French Defense. Somewhere it has calculated that that is the best move. Now my program is old and came free with my Mac. I do not know what Fritz would do.
It was Kasparov that got upset that one of the mainframes had ALL his recorded games on file -- and that was an unfair advantage. He played 3 more times against computers and won them all.
Perhaps a top GM could beat Fritz, but there seems to be little interest.

It's stupid saying that IF a computer didn't have access to databases/openings etc that a top GM would beat it (I still beg to differ anyway btw). The point is that the computers do have those resources... it's not a "disadvantage", you're playing against a computer and that's part of it's "knowledge". I could easily say that I am disadvantaged when playing Kasparov because he is so much stronger than me; but that doesn't make it an unfair contest - he will simply win.
Point is that computers are just too powerful for any GM to have any chance of beating a decent engine under normal time controls anymore; the setup that won the world championship is (as far as humans are concerned) unbeatable full stop.

... Point is that computers are just too powerful for any GM to have any chance of beating a decent engine under normal time controls anymore; the setup that won the world championship is (as far as humans are concerned) unbeatable full stop.
Not really -- it depends on what you mean by "powerful". It is an academic question anyway. Pure ability to process, a GM is still more powerful. Add endless memory, a GM STILL has a chance. That will not be the case in 10 more years.
So what? The game is between people. It was interesting computer vs. human during an age.
The movie "The Last Starfighter" (Robert Preston's last movie) was made with Atari graphics and a Cray XP1 super-computer. It looked primitive compared to any science fiction show today.
Eras come and go. Chess is between two human players. Computers are training aids, etc. Life has changed.
Muzio to be a killer for Rybka? It's the first time I hear that. Googling a little gives nothing, so what are your sources Gonno? Unfortunately I don't have Rybka to try it myself ... I find this perplexing. In a super sharp, tactical game like Muzio, Rybka is trashed? Yes, you said without opening database, but lately Rybka gave knight odds to a 2700+ player and managed to draw (so it does play extremely strong even without any opening theory).

Muzio to be a killer for Rybka? It's the first time I hear that. Googling a little gives nothing, so what are your sources Gonno? Unfortunately I don't have Rybka to try it myself ... I find this perplexing. In a super sharp, tactical game like Muzio, Rybka is trashed? Yes, you said without opening database, but lately Rybka gave knight odds to a 2700+ player and managed to draw (so it does play extremely strong even without any opening theory).
Interesting. Why would any GM play a computer today, I do not know. It was a great fun "stunt" for Kasparov for which he made a fair some of money. Today, well, I am not sure why it is done. Maybe my 10 years is off. Maybe it is 10 days.
A question came up at work, and I didn't know the answer. Perhaps someone here might know what it is, or how to find it.
A colleague asked me how many people could beat a chess computer these days? Of course, that depends on the computer, so let me frame the question this way:
Suppose you buy a high end personal computer, running some version of the Windows operating system. It should be the sort of computer that a real user might purchase for his specific use. i.e. the sort of computer that I, as a database and computer software developer, would request when spending my company's money. The best machine I could possibly justify for my sole use, as opposed to a web or departmental server machine. Then, I bought and installed a commercially available, mass market, chess playing program, such as Chessmaster or Fritz. How many people in the world could beat it? Would time controls matter?
My guess was that the computers have become so fast that there are at most a few dozen of the top grandmasters that could actually win a series of games against such a program.