diff --git a/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.BSB.rst b/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.BSB.rst
index b7656af00bc0ff664844b61b972590aaa0ea7137..b1a9fa5ef03c9556cddca5901d6c3fb0a5d1336c 100644
--- a/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.BSB.rst
+++ b/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.BSB.rst
@@ -11,6 +11,9 @@ mindquantum.algorithm.qaia.BSB
         为了内存效率，输入数组 'x' 不会被复制，并且会在优化过程中被原地修改。
         如果需要保留原始数据，请使用 `x.copy()` 传入副本。
 
+        当使用backend='gpu-int8'时，请注意该后端可能在稠密图或具有连续系数的图上表现不佳。
+        在这些情况下，可以尝试调整参数或考虑使用'cpu-float32'或'gpu-float16'后端。
+
     参数：
         - **J** (Union[numpy.array, scipy.sparse.spmatrix]) - 耦合矩阵，维度为 :math:`(N \times N)`。
         - **h** (numpy.array) - 外场强度，维度为 :math:`(N, )`。
diff --git a/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.DSB.rst b/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.DSB.rst
index b8e6b03d1822845c6ab9ffdae5210a8bfd7246f2..b0e0096c2fa5bd2cdb3b328afe1b8bca049bc063 100644
--- a/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.DSB.rst
+++ b/docs/api_python/algorithm/qaia/mindquantum.algorithm.qaia.DSB.rst
@@ -11,6 +11,9 @@ mindquantum.algorithm.qaia.DSB
         为了内存效率，输入数组 'x' 不会被复制，并且会在优化过程中被原地修改。
         如果需要保留原始数据，请使用 `x.copy()` 传入副本。
 
+        当使用backend='gpu-int8'时，请注意该后端可能在稠密图或具有连续系数的图上表现不佳。
+        在这些情况下，可以尝试调整参数或考虑使用'cpu-float32'或'gpu-float16'后端。
+
     参数：
         - **J** (Union[numpy.array, scipy.sparse.spmatrix]) - 耦合矩阵，维度为 :math:`(N \times N)`。
         - **h** (numpy.array) - 外场强度，维度为 :math:`(N, )`。
diff --git a/mindquantum/algorithm/qaia/SB.py b/mindquantum/algorithm/qaia/SB.py
index a97409a134708bf4d8c94a365e5f2dcebb6c1089..d18376b8185d0c38f71d2280413b9536d327a78f 100644
--- a/mindquantum/algorithm/qaia/SB.py
+++ b/mindquantum/algorithm/qaia/SB.py
@@ -192,6 +192,10 @@ class BSB(SB):  # noqa: N801
         in-place during optimization. If you need to preserve the original data,
         please pass a copy using `x.copy()`.
 
+        When using backend='gpu-int8', be aware that it may not perform well on dense graphs
+        or graphs with continuous coefficients. Please try adjusting parameters or consider
+        using 'cpu-float32' or 'gpu-float16' in these cases.
+
     Args:
         J (Union[numpy.array, scipy.sparse.spmatrix]): The coupling matrix with shape (N x N).
         h (numpy.array): The external field with shape (N, ).
@@ -308,6 +312,10 @@ class DSB(SB):  # noqa: N801
         in-place during optimization. If you need to preserve the original data,
         please pass a copy using `x.copy()`.
 
+        When using backend='gpu-int8', be aware that it may not perform well on dense graphs
+        or graphs with continuous coefficients. Please try adjusting parameters or consider
+        using 'cpu-float32' or 'gpu-float16' in these cases.
+
     Args:
         J (Union[numpy.array, scipy.sparse.spmatrix]): The coupling matrix with shape (N x N).
         h (numpy.array): The external field with shape (N, ).
diff --git a/mindquantum/core/circuit/circuit.py b/mindquantum/core/circuit/circuit.py
index f366c6d352e357eb31f8eac84db9849165d1b51f..ef0963bd5ddd1c8d6e4643945a26a15bf58b0f6e 100644
--- a/mindquantum/core/circuit/circuit.py
+++ b/mindquantum/core/circuit/circuit.py
@@ -21,6 +21,7 @@ import os
 from collections.abc import Iterable
 from types import FunctionType, MethodType
 from typing import List, Optional
+import warnings
 
 import numpy as np
 from rich.box import ROUNDED
@@ -972,6 +973,12 @@ class Circuit(list):  # pylint: disable=too-many-instance-attributes,too-many-pu
         if pr is None:
             pr = ParameterResolver()
         pr = _check_and_generate_pr_type(pr, self.params_name)
+        if pr.is_complex:
+            warnings.warn(
+                "Simulator does not support complex number, the imaginary part will be ignored",
+                UserWarning,
+                stacklevel=2,
+            )
         if self.has_measure_gate:
             raise ValueError("This circuit cannot have measurement gate.")
         if self.is_noise_circuit:
diff --git a/mindquantum/simulator/mqsim.py b/mindquantum/simulator/mqsim.py
index 0ced47391c7042d45a9a1b89a8f89accf3b99865..1a573d88015c0980a163c17c3682bfff0e9c84a2 100644
--- a/mindquantum/simulator/mqsim.py
+++ b/mindquantum/simulator/mqsim.py
@@ -13,6 +13,7 @@
 # limitations under the License.
 # ============================================================================
 """Mindquantum simulator."""
+import warnings
 from typing import Dict, Iterable, List, Union
 
 import numpy as np
@@ -109,6 +110,12 @@ class MQSim(BackendBase):
             if pr is None:
                 raise ValueError("Applying a parameterized circuit needs a parameter_resolver.")
             pr = _check_and_generate_pr_type(pr, circuit.params_name)
+            if pr.is_complex:
+                warnings.warn(
+                    "Simulator does not support complex number, the imaginary part will be ignored",
+                    UserWarning,
+                    stacklevel=3,
+                )
         else:
             pr = ParameterResolver()
         res = self.sim.apply_circuit(circuit.get_cpp_obj(), pr)
@@ -139,6 +146,12 @@ class MQSim(BackendBase):
                     pr = _check_and_generate_pr_type(pr, list(set(sum([p.params_name for p in gate.prs], []))))
                 else:
                     pr = _check_and_generate_pr_type(pr, gate.coeff.params_name)
+                if pr.is_complex:
+                    warnings.warn(
+                        "Simulator does not support complex number, the imaginary part will be ignored",
+                        UserWarning,
+                        stacklevel=3,
+                    )
             else:
                 pr = ParameterResolver()
             if isinstance(gate, Measure):