Page fetching time increasing exponentially

[Mideen]

When I tried with Aquarium, I hit Aquarium (It is running on my localhost:8050) with 100 parallel threads.
It takes the time exponentially from 4 sec to 40 sec. My machine has 4 CPU(s) core. I tried with one Splash instance and 5 Splash instance. There is no much time difference between them.

My docker-compose.yml is

version: '2'

services:
    haproxy:
        image: haproxy:1.7
        ports:
            # stats
            - "8036:8036"

            # splash
            - "8050:8050"
        links:
            - splash0
            - splash1
            - splash2
            - splash3
            - splash4
        volumes:
            - ./haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro

    splash0:
        image: scrapinghub/splash:3.2
        command: --max-timeout 3600 --slots 50 --maxrss 1200 --verbosity 1
        expose:
            - 8050
        mem_limit: 1680m
        memswap_limit: 2160m
        restart: always

    splash1:
        image: scrapinghub/splash:3.2
        command: --max-timeout 3600 --slots 50 --maxrss 1200 --verbosity 1
        expose:
            - 8050
        mem_limit: 1680m
        memswap_limit: 2160m
        restart: always

    splash2:
        image: scrapinghub/splash:3.2
        command: --max-timeout 3600 --slots 50 --maxrss 1200 --verbosity 1
        expose:
            - 8050
        mem_limit: 1680m
        memswap_limit: 2160m
        restart: always

    splash3:
        image: scrapinghub/splash:3.2
        command: --max-timeout 3600 --slots 50 --maxrss 1200 --verbosity 1
        expose:
            - 8050
        mem_limit: 1680m
        memswap_limit: 2160m
        restart: always

    splash4:
        image: scrapinghub/splash:3.2
        command: --max-timeout 3600 --slots 50 --maxrss 1200 --verbosity 1
        expose:
            - 8050
        mem_limit: 1680m
        memswap_limit: 2160m
        restart: always

Is there any way to speed up the Splash on parellel hitting.