Intermediate results as variables or only base values?

I have a object which has a few base variables (integers as example) and one intermediate variable for further processing. The intermediate variable can be calculated from the base vars. Now the question: Should I always repeat the calculation of the intermediate var from the base vars or should I save the value and only change it if the base vars change? (I need access to every var in pseudo random order.)

Two examples in C++-Code.

Caching the result:

class Test
{
  private:
    int a;
    int b;
    int c;

    int x;

    void calcX ()
    {
        // Just an example for a calculation for x.
        x = ((a + 10) * (b / (c*c + 1)));
    }

  public:
    void setA (int var)
    {
        a = var;
        calcX();
    }
    void setB (int var)
    {
        b = var;
        calcX();
    }
    void setC (int var)
    {
        c = var;
        calcX();
    }

    // Every var has a get methode.
}

Always recalculating the result:

class Test
{
  private:
    int a;
    int b;
    int c;

    int x;

  public:
    // Var a, b and c have get and set methodes.

    int getX ()
    {
        return ((a + 10) * (b / (c*c + 1)));
    }
}

So, which way is considered better? I really don't know. We should not save the intermediate value because it would be redundant, but then again we should not recalculate the value again and again.