Run time parallel loop in R

Question

I have access to a quite powerful remote desktop where I ran some parallel code in R. However, I am surprised about the results when I compare the running time between the parallel loop and the simple for loop. Below is my code.

library(rbenchmark)
library(foreach)
library(doParallel)

max.eig <- function(N, sigma) {
  d <- matrix(rnorm(N**2, sd = sigma), nrow = N)
  E <- eigen(d)$values
  abs(E)[[1]]
}

no_cores <- 20
registerDoParallel(cores=no_cores)  
cl <- makeCluster(no_cores) 

benchmark(
  foreach(n = 1:50) %do% max.eig(n, 1),
  foreach(n = 1:50) %dopar% max.eig(n, 1)
)

The speed comparison gives

                                     test replications elapsed relative user.self sys.self user.child sys.child
   foreach(n = 1:50) %do% max.eig(n, 1)          100   3.097    1.000     3.081    0.003      0.000     0.000
foreach(n = 1:50) %dopar% max.eig(n, 1)          100   8.327    2.689     0.943    6.643      3.648    27.523

Notice that the running time of the parallel loop is much longer. This is true regardless of the number of cores I register. In contrast, when I run the comparison on my own computer, the parallel for loop is much faster (with 4 cores). The results give

                                    test replications elapsed relative user.self sys.self user.child sys.child
    foreach(n = 1:70) %do% max.eig(n, 1)          100   11.36    1.939     11.27     0.03         NA        NA
 foreach(n = 1:70) %dopar% max.eig(n, 1)          100    5.86    1.000      2.97     0.35         NA        NA

How is it possible that the parallel loop is slower on the remote desktop? And is there a way to set up the parallel cluster differently so that it is much faster and I can exploit the many cores on the remote computer?