diff --git a/pytorch3d/ops/points_normals.py b/pytorch3d/ops/points_normals.py
index e5b0ec337..702c0bb70 100644
--- a/pytorch3d/ops/points_normals.py
+++ b/pytorch3d/ops/points_normals.py
@@ -8,6 +8,7 @@
 
 import torch
 
+from ..common.workaround import symeig3x3
 from .utils import convert_pointclouds_to_tensor, get_point_covariances
 
 
@@ -19,6 +20,8 @@ def estimate_pointcloud_normals(
     pointclouds: Union[torch.Tensor, "Pointclouds"],
     neighborhood_size: int = 50,
     disambiguate_directions: bool = True,
+    *,
+    use_symeig_workaround: bool = True,
 ) -> torch.Tensor:
     """
     Estimates the normals of a batch of `pointclouds`.
@@ -33,6 +36,8 @@ def estimate_pointcloud_normals(
         geometry around each point.
       **disambiguate_directions**: If `True`, uses the algorithm from [1] to
         ensure sign consistency of the normals of neighboring points.
+      **use_symeig_workaround**: If `True`, uses a custom eigenvalue
+        calculation.
 
     Returns:
       **normals**: A tensor of normals for each input point
@@ -48,6 +53,7 @@ def estimate_pointcloud_normals(
         pointclouds,
         neighborhood_size=neighborhood_size,
         disambiguate_directions=disambiguate_directions,
+        use_symeig_workaround=use_symeig_workaround,
     )
 
     # the normals correspond to the first vector of each local coord frame
@@ -60,6 +66,8 @@ def estimate_pointcloud_local_coord_frames(
     pointclouds: Union[torch.Tensor, "Pointclouds"],
     neighborhood_size: int = 50,
     disambiguate_directions: bool = True,
+    *,
+    use_symeig_workaround: bool = True,
 ) -> Tuple[torch.Tensor, torch.Tensor]:
     """
     Estimates the principal directions of curvature (which includes normals)
@@ -88,6 +96,8 @@ def estimate_pointcloud_local_coord_frames(
         geometry around each point.
       **disambiguate_directions**: If `True`, uses the algorithm from [1] to
         ensure sign consistency of the normals of neighboring points.
+      **use_symeig_workaround**: If `True`, uses a custom eigenvalue
+        calculation.
 
     Returns:
       **curvatures**: The three principal curvatures of each point
@@ -133,7 +143,10 @@ def estimate_pointcloud_local_coord_frames(
     # eigenvectors (=principal directions) in an ascending order of their
     # corresponding eigenvalues, while the smallest eigenvalue's eigenvector
     # corresponds to the normal direction
-    curvatures, local_coord_frames = torch.symeig(cov, eigenvectors=True)
+    if use_symeig_workaround:
+        curvatures, local_coord_frames = symeig3x3(cov, eigenvectors=True)
+    else:
+        curvatures, local_coord_frames = torch.symeig(cov, eigenvectors=True)
 
     # disambiguate the directions of individual principal vectors
     if disambiguate_directions:
diff --git a/tests/benchmarks/bm_points_normals.py b/tests/benchmarks/bm_points_normals.py
new file mode 100644
index 000000000..487e71e54
--- /dev/null
+++ b/tests/benchmarks/bm_points_normals.py
@@ -0,0 +1,47 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import itertools
+
+import torch
+from fvcore.common.benchmark import benchmark
+from pytorch3d.ops import estimate_pointcloud_normals
+from test_points_normals import TestPCLNormals
+
+
+def to_bm(num_points, use_symeig_workaround):
+    device = torch.device("cuda:0")
+    points_padded, _normals = TestPCLNormals.init_spherical_pcl(
+        num_points=num_points, device=device, use_pointclouds=False
+    )
+    torch.cuda.synchronize()
+
+    def run():
+        estimate_pointcloud_normals(
+            points_padded, use_symeig_workaround=use_symeig_workaround
+        )
+        torch.cuda.synchronize()
+
+    return run
+
+
+def bm_points_normals() -> None:
+    case_grid = {
+        "use_symeig_workaround": [True, False],
+        "num_points": [3000, 6000],
+    }
+    test_cases = itertools.product(*case_grid.values())
+    kwargs_list = [dict(zip(case_grid.keys(), case)) for case in test_cases]
+    benchmark(
+        to_bm,
+        "normals",
+        kwargs_list,
+        warmup_iters=1,
+    )
+
+
+if __name__ == "__main__":
+    bm_points_normals()