diff --git a/exercise5/handout_pthreads/wave_2d_pthread.c b/exercise5/handout_pthreads/wave_2d_pthread.c
index 044e3f3..4154c8f 100644
--- a/exercise5/handout_pthreads/wave_2d_pthread.c
+++ b/exercise5/handout_pthreads/wave_2d_pthread.c
@@ -85,10 +85,13 @@ void domain_finalize(void) {
 // Integration formula
 void time_step(int_t thread_id) {
   // BEGIN: T3
-  for (int_t i = 0; i < N; i += 1)
+  // let threads work each on a row for better cache locality (huge speed-up)
+  int_t start_row = thread_id * (N / n_threads);
+  // let last thread handle remainder
+  int_t end_row = (thread_id == n_threads - 1) ? N : (thread_id + 1) * (N / n_threads);
+  for (int_t i = start_row; i < end_row; i++)
     for (int_t j = 0; j < N; j++)
-      if (j % n_threads == thread_id)
-        U_nxt(i, j) = -U_prv(i, j) + 2.0 * U(i, j) + (dt * dt * c * c) / (h * h) * (U(i - 1, j) + U(i + 1, j) + U(i, j - 1) + U(i, j + 1) - 4.0 * U(i, j));
+      U_nxt(i, j) = -U_prv(i, j) + 2.0 * U(i, j) + (dt * dt * c * c) / (h * h) * (U(i - 1, j) + U(i + 1, j) + U(i, j - 1) + U(i, j + 1) - 4.0 * U(i, j));
   // END: T3
 }