Kernel time limit on multi GPU system

We have a multi GPU system with three GTX480 cards with one connected to the monitor. Supposly the one that is connected to the monitor should have the time limit on and the other two should have it off. The problem we are facing now is that all three cards have this “time limit” on even though only one of them is connected to the monitor. Is this a problem with the drivers or is it windows 7?

We also have a second machine with one GTX480 and four C2070 and the same problem there, but we can fix the problem by turning TCC on. Why is the time limit on by default? Is there a way to turn it off for the machine with three GTX480?