why the simulation result is differenct between fork join and begin end?

[Altera_Forum]

this is the testbench in an495. 

why can not use begin ... end. initial   begin  arst <= 1'b0; // resetting the contoller initially  # 250   arst <= 1'b1;   iderst <= 1'b0;  ideen <= 1'b1;  pioiordyen <= 1'b1;  piorqst <= 1'b1;  pioaddr <= 4'b0111;  piodatain <= 16'b1111111100000000;  piowe <= 1'b1;  # 600 piowe <= 1'b0;  ddi <= 16'b0000000011111111;  iordy <= 1'b1;  intrq <= 1'b1;  flag <= 1'b0; //if error occured in the test procedure,then asserted  # 5000  $stop;  end use the fork join,the result is right  fork  arst = 1'b0; // resetting the contoller initially  # 250   arst = 1'b1;   iderst = 1'b0;  ideen = 1'b1;  pioiordyen = 1'b1;  piorqst = 1'b1;  pioaddr = 4'b0111;  piodatain = 16'b1111111100000000;  piowe = 1'b1;  # 600 piowe = 1'b0;  ddi = 16'b0000000011111111;  iordy = 1'b1;  intrq = 1'b1;  flag = 1'b0; //if error occured in the test procedure,then asserted  # 5000  $stop;  join

[Altera_Forum]

The difference between "begin-end" and "fork-join" is that in initial block "begin-end" operations work step-by-step, but in "fork-join" they work simultaneously. You can rewrite the testbench using "begin-end". For example,  

 begin  arst <= 1'b0; // resetting the contoller initially  iderst <= 1'b0;  ideen <= 1'b1;  pioiordyen <= 1'b1;  piorqst <= 1'b1;  pioaddr <= 4'b0111;  piodatain <= 16'b1111111100000000;  piowe <= 1'b1;  iordy <= 1'b1;  intrq <= 1'b1;  flag <= 1'b0; //if error occured in the test procedure,then asserted  ddi <= 16'b0000000011111111;  # 250 arst <= 1'b1;   # 350 piowe <= 1'b0;  # 4400 $stop;  end  I believe, it will work like original "fork-join"

[Altera_Forum]

The begin-end-block also assigns the values to the nets simultaneously if non-blocking assignments are used (as it's the case in the example). The difference is that all timing assignments are with respect to the block entry time in case of fork-join and with respect to the previous timing assignment in case of begin-end.