> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/holoscan/sdk-user-guide/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/holoscan/sdk-user-guide/_mcp/server.

# holoscan::kDefaultTCPIPSegmentSize

> The default TCP/IP segment size used by UCX.

The default TCP/IP segment size used by UCX.

```cpp showLineNumbers={false}
constexpr const char * holoscan::kDefaultTCPIPSegmentSize = "128k"
```

If the user has not explicitly set UCX\_TCP\_TX\_SEG\_SIZE and UCX\_TCP\_RX\_SEG\_SIZE, they will be set to this default value.

UCX's default value of 8k results in slow TCP/IP + cuda\_copy performance. Changing to 128k should substantially improve performance of GPU tensor transfers.

If this value is reduced from 128k, it will be necessary to also reduce kDefaultUcxSerializationBufferSize for [UcxSerializationBuffer](../classes/ucxserializationbuffer) to a smaller value as that buffer must be smaller than this segment size.