Argon2 5 threads is slower than 4 threads

Question

I am using https://github.com/charlesportwoodii/php-argon2-ext to play around with Argon2. While i was playing with the parameters I noticed that the speed is not decreasing per core accordingly.

See a graph here: Time is in seconds

Code a ran:

for($threads = 1; $threads <= 8; $threads++) {
    $time1 = microtime(true);

    $options = [
        'm_cost' => 65536,//64MB
        't_cost' => 64,
        'threads' => $threads
    ];

    $password = "secret";
    argon2_hash($password, HASH_ARGON2I, $options);

    $time2 = microtime(true);
    $execution_time = ($time2 - $time1);
    printf("execution time with $threads threads: " . $execution_time . " seconds\n");
}

My system: Debian 9, 8GB RAM, i7-4710HQ @ 2.50GHz

Can somebody explain this or can think of an reason?