I was going to research this more deeply before making a posting, but, to be honest, I find it too sad.
The chess program, Rybka back in March had beaten GM Jaan Ehlvest +4-1=3 at pawn odds (not pawn and move, just pawn). At the time I felt that the computer still maintained several unfair advantages over the human opponent.
In a recent rematch in July, Ehlvest played the computer even (no material odds), but the computer satisfied all my objections. Rybka's opening book was limited to 3 moves deep - fair enough. There we no endgame tablebases allowed - fair enough.... sigh... Rybka played black in every game. The computer was a "Quadcore pc, limit of 512 MB for hash tables." The time control was 90 minute/30second increment for Ehlvest - 45 minute/15 second increment for Rybka. Ehlvest was allowed two five minute breaks per game... sighhhhh....
The result - 4.5-1.5 in favor of Rybka. It wasn't even a contest.
In August, Joel Benjamin played Rybka at pawn odds. 8 games, 1 hour/ 30 second increment, alternating colors. Benjamin lost +1-2=5.