about numjac handling exp function

I am facing a problem that when I am using numjac to find the jacobian of a exp function, when my deminsion grows , the error grows super large, how can I fix that? the seting as following:

x=[0:1/(i-1):1]'
t=1;
fty=ff(t,x);
thresh=eps.*ones(i,1);
frac=[];
vectorized=1;
numjac(@ff ,t,x,fty,thresh,frac,vectorized);

where

[ FF ] = ff( t,X )
d=size(X);
n=d(1);
FF=repmat(sum(X),n,1).*exp(X);