The ring $$R=\mathbb{C}[x_1,x_2,\ldots]/(x_1,x_2,\ldots)^2\cong\mathbb{C}[\epsilon_1,\epsilon_2,\ldots]$$
is non-noetherian (where $\epsilon_i$ denotes the image of $x_i$ in the quotient), and the radical of the (obviously finite) zero ideal $I=(0)$ in $R$ is equal to the ideal $(\epsilon_1,\epsilon_2,\ldots)$ which is not finitely generated, much less finite.
It seems to me to be impossible to say something solely in terms of a generating set for $I$, the structure of $R$ is what determines $\sqrt{I\,\strut}$.
Example 1. Take a rank one non-discrete valuation domain $V$. Note that the maximal ideal $M$ of $V$ is not finitely generated. Now take any nonzero non unit x in $V$ and consider $rad(xV)$. Because $V$ is one dimensional for each $y\in M$ we have $x|y^n$ for some $n$ [Kaplansky, Commutaive Rings Theorem 108]. That forces $rad(xV)=M$.
– mzafrullah Jul 30 '20 at 11:05