diff --git a/include/ck/utility/type_convert.hpp b/include/ck/utility/type_convert.hpp
index ad2acdf0ad..0f273affec 100644
--- a/include/ck/utility/type_convert.hpp
+++ b/include/ck/utility/type_convert.hpp
@@ -58,20 +58,6 @@ inline __host__ __device__ constexpr bhalf_t bf16_convert_rtn<bhalf_t, float>(fl
         uint32_t int32;
     } u = {x};
 
-    uint32_t sign = u.int32 & 0x80000000;
-    uint32_t exp  = u.int32 & 0x7F800000;
-
-    // 处理超出bf16指数范围的情况（溢出）
-    if(exp > 0x47800000)
-    { // 原float指数超过bf16最大正常数
-        return sign ? uint16_t(0xFF80) : uint16_t(0x7F80); // 钳位到无穷大
-    }
-    // 处理下溢（此处简化为零，实际可能需要调整尾数）
-    if(exp < 0x38000000)
-    { // 原float指数过小
-        return sign ? uint16_t(0x8000) : uint16_t(0x0000);
-    }
-
     const uint32_t first_bf16_mantisa_bit = ((u.int32 >> 16) & 1);
     constexpr uint32_t rounding_bias      = uint32_t((1 << 15) - 1);
     union