协程似乎并不比 JVM 线程占用更少的资源

问：

我做了一个基准测试（参考答案）来测试线程池中 coroutiens 和线程之间的内存使用情况：

val COUNT = 4_000

val executor: Executor = Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors())// 12 cores on my machine
 fun main(array: Array<String>)  = runBlocking{
     val latch = CountDownLatch(COUNT)
     val start = System.currentTimeMillis()
     repeat(COUNT) {
         launch(Dispatchers.Default) {
             testByCoroutine(latch)
         }
     }
     latch.await()
     println("total: " + (System.currentTimeMillis() - start))


     // testByThreadPool()
 }


fun testByThreadPool() {
    val latch = CountDownLatch(COUNT)
    val start = System.currentTimeMillis()
    for (i in 0..<COUNT) {
        executor.execute {
            val num: Int = request1()
            println(request2(num))
            latch.countDown()
        }
    }
    try {
        latch.await()
    } catch (e: InterruptedException) {
        throw RuntimeException(e)
    }
    println("total: " + (System.currentTimeMillis() - start))
    exitProcess(0)
}

fun testByCoroutine(latch: CountDownLatch) {
    val num = request1()
    val res: Int = request2(num)
    println(res)
    latch.countDown()
}


fun request1(): Int {
    return doGet("https://bing.com")
}

fun request2(token: Int): Int {
    val response: Int = doGet("https://bing.com")
    return response + token
}

fun doGet(link: String): Int {
    try {
        val url = URL(link)
        val conn = url.openConnection() as HttpURLConnection
        conn.setRequestMethod("GET")
        conn.setConnectTimeout(3000)
        return conn.getResponseCode()
    } catch (e: java.lang.Exception) {
        e.printStackTrace()
    }
    return -1
}

线程池案例：

total: 518218

Dispatchers.Default 案例：

total: 561510

Dispatchers.IO 案例：

total: 262719

根据答案：

Coroutines are not designed to be faster than threads, it is for lower RAM consumption 

有一个基准测试表明，线程和协程之间消耗的内存中似乎存在相对恒定的 6：1 比率。

但是从上面的图表中，我没有看到协程和线程池之间的内存使用有任何明显差异。

我设计的基准测试错了吗？