The simulation rate is essentially as follows: how many times per second are you calculating and updating your objects?
If you're simulation rate is 10 times per second, you're taking all of your objects, updating their position by their velocty/10, and dealing with any forces and collisions. If you up that to 40 times per second, you're moving each object a smaller each time, but you're doing more updates. If everything works properly, a higher simulation rate means more precise physics. Your accelerations will be smoother, and your collisions will be calculated closer to when they should occur. A lower rate means less precise physics (and you might find bounces and accelerations not working as intended)
If you're completing all the calculations in less than 1/40th of a second, a simulation rate of 40 would mean you can display the objects in real-time (ignoring the time it takes to draw them for the purpose of answering this question). The load on your CPU per second would be equal to 40*the time it takes to calculate your physics.
However, if it exceeds your CPU's capabilities, you've got two options: not render in real time (save the locations from each frame), or drop frames. This can potentially wreak havoc (pardon the pun) on the simulation itself as it could miss collisions or not properly calculate accelerations.
So in short, if you need precision, try upping it, but if you find that it's impacting CPU usage, consider dialing it back a bit.