We present a minimal relational floating-point arithmetic that supports comparison, addition/subtraction, and multiplication/division %. These relations support in a binary, normalized floating-point system with rounding by chopping. The system can be used to solve simple arithmetic problems, quadratic equations, and can reason relationally about overflow. We also show that its runtime generally grows exponentially with respect to precision, and that multiplication runtime grows exponentially with respect to the number of 1’s in the mantissa.