Play 2.4, increase thread number, but slow down the processing

I build my play service with Play2.4 and default netty server. I started the service on a server with a 8 cores processor, and use the below command to test.

ab -c 2000 -n 20000 <test_url>

It takes 12.856s to complete when i set parallelism-factor as 1.0.

it takes 39.95s to complete when i change parallelism-factor to 2.0.

CPU and Memory usages are very low for both cases.

The test covers I/O operations, it seems more threads can save I/O waiting time and process more requests. But why more threads slow down the processing? How can I improve the throughput ?

test code as below.

public Result checkPerformance() {
    Random random = new Random();
    int skip = random.nextInt(20000);
    // do some database query
    return ok("request completed");

the thread settings in application.conf as below

akka {
    actor {
        default-dispatcher {
            fork-join-executor {
                parallelism-factor = 2.0  # 1.0
                parallelism-min = 8
                parallelism-max = 40