comparison of direct search and conjugate gradient method

direct search  (Nelder-Mead method): matlab function--fminsearch
conjugate gradient : matlab fun---fminunc
for the same questions to find the next solution:
function f=ff2(x)
f=4*x(1)^2+5*x(1)*x(2)+2*x(2)^2;
the initial value is [1,1];
result of conjugate:
Optimization terminated successfully:
 Search direction less than 2*options.TolX
x =
  1.0e-007 *
   -0.3723    0.4730

fval =
  1.2138e-015

exitflag =
     1

output =
       iterations: 3
        funcCount: 16
         stepsize: 1.9939
    firstorderopt: 3.2647e-007
        algorithm: 'medium-scale: Quasi-Newton line search'
result of direct :
x =
  1.0e-004 *
   -0.4945    0.5283

fval =
  2.3014e-009

exitflag =
     1

output =
    iterations: 47
     funcCount: 90
     algorithm: 'Nelder-Mead simplex direct search'