diff --git a/src/finn/builder/build_dataflow_config.py b/src/finn/builder/build_dataflow_config.py
index 4a112699ec9bdf126f447fe2244eb01f6f4fa042..7a70398399a9db94ea398b8fd94e3c150a962e91 100644
--- a/src/finn/builder/build_dataflow_config.py
+++ b/src/finn/builder/build_dataflow_config.py
@@ -172,6 +172,13 @@ class DataflowBuildConfig:
     #: that will override the target_fps setting here.
     target_fps: Optional[int] = None
 
+    #: (Optional) Use two-pass relaxation for folding, only relevant if target_fps
+    #: is set. If enabled, parallelization will internally run a second time if the
+    #: target cycles from the first pass could not be achieved, instead using the
+    #: achievable target to obtain a balanced pipeline. If disabled, this can be
+    #: useful for decreasing the latency (even though throughput won't increase).
+    folding_two_pass_relaxation: Optional[bool] = True
+
     #: (Optional) At which steps the generated intermediate output model
     #: will be verified. See documentation of VerificationStepType for
     #: available options.
diff --git a/src/finn/builder/build_dataflow_steps.py b/src/finn/builder/build_dataflow_steps.py
index ce0fc49847344398a2e4709e93f312e31854c9dd..b19f43c9f5810db11090630bc6d865747d582fdf 100644
--- a/src/finn/builder/build_dataflow_steps.py
+++ b/src/finn/builder/build_dataflow_steps.py
@@ -232,7 +232,11 @@ def step_target_fps_parallelization(model: ModelWrapper, cfg: DataflowBuildConfi
     target_cycles_per_frame = cfg._resolve_cycles_per_frame()
     if target_cycles_per_frame is not None:
         model = model.transform(
-            SetFolding(target_cycles_per_frame, mvau_wwidth_max=cfg.mvau_wwidth_max)
+            SetFolding(
+                target_cycles_per_frame,
+                mvau_wwidth_max=cfg.mvau_wwidth_max,
+                two_pass_relaxation=cfg.folding_two_pass_relaxation,
+            )
         )
     return model