diff --git a/benchmarks/dynamo/distributed.py b/benchmarks/dynamo/distributed.py
index 194b3906de0..b490c48ade9 100644
--- a/benchmarks/dynamo/distributed.py
+++ b/benchmarks/dynamo/distributed.py
@@ -81,8 +81,8 @@ def run_model(args, model, inputs, key):
         if args.verbose:
             dynamo.config.verbose = True
             dynamo.config.log_level = logging.DEBUG
-        if args.dynamo_optimize_ddp:
-            dynamo.config.optimize_ddp = True
+        if args.dynamo_no_optimize_ddp:
+            dynamo.config.optimize_ddp = False
         if args.dynamo == "inductor" and args.fsdp:
             torch._inductor.config.triton.cudagraphs = False
             log.warn("disabling inductor cudagraphs for compatibility with FSDP")
@@ -129,7 +129,7 @@ if __name__ == "__main__":
     parser.add_argument("--trace_file", default="profile.json", help="Run the profiler")
     parser.add_argument("--repeat", default=10, help="Repeats for timing run")
     parser.add_argument(
-        "--dynamo_optimize_ddp",
+        "--dynamo_no_optimize_ddp",
         action="store_true",
         help="Enable dynamo's ddp optimizer",
     )
diff --git a/torch/_dynamo/config.py b/torch/_dynamo/config.py
index 26efff20538..258df9989f8 100644
--- a/torch/_dynamo/config.py
+++ b/torch/_dynamo/config.py
@@ -138,8 +138,11 @@ capture_scalar_outputs = False
 enforce_cond_guards_match = True
 
 # Automatically split model graph into pieces to match DDP bucket sizes
-# to allow DDP comm/compute overlap
-optimize_ddp = False
+# to allow DDP comm/compute overlap.  Disable to allow DDP models to
+# run without graph-breaks, but also without comm/compute overlap.
+# set torch._dynamo.config.log_level to INFO or DEBUG for more info
+# about optimize_ddp behavior.
+optimize_ddp = True
 
 # If True, raises exception if TorchDynamo is called with a context manager
 raise_on_ctx_manager_usage = True