The Eigenvalue That Went Missing

A dataset has $5$ features. The diagonal of its covariance matrix — the variance of each individual feature — is:

Feature	$f_1$	$f_2$	$f_3$	$f_4$	$f_5$
Variance	$8$	$6$	$5$	$4$	$2$

A PCA run wrote all five eigenvalues of that covariance matrix to a log file, but one entry was corrupted and is unreadable:

$11.5, \quad 4.5, \quad 3.0, \quad 1.8, \quad \square$

You need to keep at least $90\%$ of the total variance. What is the smallest number of principal components $k$ that achieves this?