diff --git a/ticdc/ticdc-data-replication-capabilities.md b/ticdc/ticdc-data-replication-capabilities.md
index ab67d6338f593..8c9e8cb0c6d1e 100644
--- a/ticdc/ticdc-data-replication-capabilities.md
+++ b/ticdc/ticdc-data-replication-capabilities.md
@@ -48,3 +48,17 @@ TiCDC supports the following types of upstream data changes:
 - TiCDC only verifies the integrity of upstream data changes. It does not validate whether the changes conform to upstream or downstream constraints. If the data violates downstream constraints, TiCDC will return an error when writing to the downstream. 
 
     For example: When a changefeed is configured to filter out all DDL events, if the upstream executes a `DROP COLUMN` operation but continues to write `INSERT` statements involving that column, TiCDC will fail to replicate these DML changes to the downstream because of table schema mismatches.
+
+- For the TiCDC [classic architecture](/ticdc/ticdc-classic-architecture.md), when the number of tables replicated by a single TiCDC cluster exceeds the following recommended values, TiCDC might not work stably:
+
+    | TiCDC version | Recommended number of tables to be replicated |
+    |---|---|
+    | v5.4.0 - v6.5.x | 2000 |
+    | v6.5.x - v7.1.x | 4000 |
+    | v7.5.x - v8.5.x | 40000 |
+
+    > **Note:**
+    >
+    > When replicating partitioned tables, TiCDC treats each partition as a separate table. Therefore, the partition count is included when TiCDC calculates the total number of tables to be replicated.
+
+    If the number of tables to be replicated exceeds the preceding recommended values, it is recommended to use the [TiCDC new architecture](/ticdc/ticdc-architecture.md). The new architecture supports replicating more than one million tables per changefeed, making it suitable for large-scale replication scenarios.