diff --git a/per4m/example1.py b/per4m/example1.py
index 14134cf..8c88378 100644
--- a/per4m/example1.py
+++ b/per4m/example1.py
@@ -1,21 +1,35 @@
+# same as example1, but without explicit viztracer calls
 import threading
 import time
-import viztracer
-import time
 
+# if we don't run with gil_load, we just skip it
+import gil_load
+try:
+    gil_load.init()
+    gil_load.start()
+    use_gil_load = True
+except RuntimeError:
+    use_gil_load = False
 
-def run():
+def some_computation():
     total = 0
     for i in range(1_000_000):
         total += i
     return total
 
 
-with viztracer.VizTracer(output_file="example1.json"):
-    thread1 = threading.Thread(target=run)
-    thread2 = threading.Thread(target=run)
+thread1 = threading.Thread(target=some_computation)
+thread2 = threading.Thread(target=some_computation)
+def main(args=None):
     thread1.start()
     thread2.start()
     time.sleep(0.2)
     for thread in [thread1, thread2]:
         thread.join()
+    if use_gil_load:
+        gil_load.stop()
+        stats = gil_load.get()
+        print(gil_load.format(stats))
+
+if __name__ == "__main__":
+    main()
diff --git a/per4m/example2.py b/per4m/example2.py
index 8c88378..c206650 100644
--- a/per4m/example2.py
+++ b/per4m/example2.py
@@ -1,6 +1,7 @@
-# same as example1, but without explicit viztracer calls
+# shows high gil load, but no wait
 import threading
 import time
+import numpy as np
 
 # if we don't run with gil_load, we just skip it
 import gil_load
@@ -11,25 +12,35 @@ try:
 except RuntimeError:
     use_gil_load = False
 
-def some_computation():
+
+N = 1024*1024*32
+M = 4
+x = np.arange(N, dtype='f8')
+
+
+def some_numpy_computation():
     total = 0
-    for i in range(1_000_000):
-        total += i
+    for i in range(M):
+        total += x.sum()
     return total
 
 
-thread1 = threading.Thread(target=some_computation)
-thread2 = threading.Thread(target=some_computation)
+thread1 = threading.Thread(target=some_numpy_computation)
+thread2 = threading.Thread(target=some_numpy_computation)
+
+
 def main(args=None):
     thread1.start()
     thread2.start()
-    time.sleep(0.2)
+    total = 0
+    for i in range(2_000_000):
+        total += i
     for thread in [thread1, thread2]:
         thread.join()
+
     if use_gil_load:
         gil_load.stop()
         stats = gil_load.get()
         print(gil_load.format(stats))
-
 if __name__ == "__main__":
     main()