Limiting double to 3 decimal places

Doubles don't have decimal places - they're not based on decimal digits to start with. You could get "the closest double to the current value when truncated to three decimal digits", but it still wouldn't be exactly the same. You'd be better off using decimal.

Having said that, if it's only the way that rounding happens that's a problem, you can use Math.Truncate(value * 1000) / 1000; which may do what you want. (You don't want rounding at all, by the sounds of it.) It's still potentially "dodgy" though, as the result still won't really just have three decimal places. If you did the same thing with a decimal value, however, it would work:

decimal m = 12.878999m;
m = Math.Truncate(m * 1000m) / 1000m;
Console.WriteLine(m); // 12.878

EDIT: As LBushkin pointed out, you should be clear between truncating for display purposes (which can usually be done in a format specifier) and truncating for further calculations (in which case the above should work).

I can't think of a reason to explicitly lose precision outside of display purposes. In that case, simply use string formatting.

double example = 12.34567;

Console.Out.WriteLine(example.ToString("#.000"));

double example = 3.1416789645;
double output = Convert.ToDouble(example.ToString("N3"));

Multiply by 1000 then use Truncate then divide by 1000.

If your purpose in truncating the digits is for display reasons, then you just just use an appropriate formatting when you convert the double to a string.

Methods like String.Format() and Console.WriteLine() (and others) allow you to limit the number of digits of precision a value is formatted with.

Attempting to "truncate" floating point numbers is ill advised - floating point numbers don't have a precise decimal representation in many cases. Applying an approach like scaling the number up, truncating it, and then scaling it down could easily change the value to something quite different from what you'd expected for the "truncated" value.

If you need precise decimal representations of a number you should be using decimal rather than double or float.