diff --git a/tests/unit/test_fmha_v3_stage_c.py b/tests/unit/test_fmha_v3_stage_c.py
index be944adb..5f01bb75 100644
--- a/tests/unit/test_fmha_v3_stage_c.py
+++ b/tests/unit/test_fmha_v3_stage_c.py
@@ -323,8 +323,6 @@ class FmhaV3StageCMulti:
             # the missing rescale shows as accuracy drift.
             for kt in range(self.n_kv_tiles):
                 si_handle = s_cons.wait_and_advance()
-                if kt == 0:
-                    cute.printf("SOFTMAX self.n_kv_tiles=%d\n", Int32(self.n_kv_tiles))
 
                 # Load S[kt]
                 tTMEM_LOADrS = cute.make_rmem_tensor(tTMEM_LOADcS.shape, self.qk_acc_dtype)
@@ -403,7 +401,6 @@ class FmhaV3StageCMulti:
 
             # === Final O normalization: O *= 1/row_sum ===
             inv_row_sum = Float32(1.0) / row_sum
-            cute.printf("FINAL row_sum=%f inv_row_sum=%f\n", row_sum, inv_row_sum)
 
             tTMrO = cute.make_rmem_tensor(
                 (tTMEM_LOADcO.shape, 128 // corr_tile_size), self.acc_dtype