Think dynamic programming: dp[i] will be the answer for array A[0], ..., A[i-1].

For j = 1 .. k that keeps everything in bounds, dp[i] is the maximum of dp[i-j] + max(A[i-1], ..., A[i-j]) * j .