diff --git a/tensorflow/core/kernels/training_ops_gpu.cu.cc b/tensorflow/core/kernels/training_ops_gpu.cu.cc
index 92496e63e1a..2fd2dbe90bb 100644
--- a/tensorflow/core/kernels/training_ops_gpu.cu.cc
+++ b/tensorflow/core/kernels/training_ops_gpu.cu.cc
@@ -148,41 +148,30 @@ __device__ Eigen::half impl_rsqrt(Eigen::half x) {
 
 template <class T>
 __device__ std::complex<T> impl_sqrt(std::complex<T> x) {
-  T re = x.real(), im = x.imag();
-  T mod_x = sqrt(re * re + im * im);
-  const T root2 = 0.7071067811865475;
+  T a = x.real();
+  T b = x.imag();
+  T r = impl_sqrt(norm(x));
+  // returns sqrt(0.5 * (r - v)) where v may be close to r.
+  auto helper = [&](const T& v) {
+    T diff = r - v;
+    if (diff < T(1e-5) * r) {
+      // |a| >> |b|, use rsqrt(1+x) ~= 1 + x/2.
+      return T(0.5) * fabs(b) * impl_rsqrt(r);
+    }
+    return sqrt(T(0.5) * diff);
+  };
   // We pick the root with the same sign of the imaginary component as
   // the input.
-  T root[2] = {T(sqrt(mod_x + re) * root2),
-               T(sqrt(mod_x - re) * root2 * (im >= 0 ? 1. : -1.))};
+  T result[2] = {helper(-a),
+                 [&](const T& v) { return b >= 0 ? v : -v; }(helper(a))};
   // hcc/clang is really weird with its support of complex in device code;
   // for some reason it does not permit a 2-argument constructor
-  return *(reinterpret_cast<std::complex<T>*>(&root));
-}
-
-template <class T>
-__device__ T rsqrt_helper(T x) {
-  return 0.5 * x + 0.125 * x * x + 0.0625 * x * x * x;
+  return reinterpret_cast<const std::complex<T>&>(result);
 }
 
 template <class T>
 __device__ std::complex<T> impl_rsqrt(std::complex<T> x) {
-  T re = x.real(), im = x.imag();
-  T r = rsqrt(re * re + im * im);
-  T ar2 = re * r * r;
-  const T root2 = 0.7071067811865475;
-  T root[2];
-  // With float, calculating 1+re*r and 1-re*r may result in excessive errors
-  // due to subtraction of two close values. We have to get fancy
-  root[0] = sqrt(r * ((std::is_same<T, float>::value && re * r < -0.98)
-                          ? rsqrt_helper(im * im * r * r)
-                          : 1 + re * r)) *
-            root2;
-  root[1] = sqrt(r * ((std::is_same<T, float>::value && re * r > 0.98)
-                          ? rsqrt_helper(im * im * r * r)
-                          : 1 - re * r)) *
-            root2 * (im >= 0 ? -1. : 1.);
-  return *(reinterpret_cast<std::complex<T>*>(&root));
+  return conj(impl_sqrt(x)) * impl_rsqrt(norm(x));
 }
 
 template <typename T>