AN 855: PCI Express* High Performance Reference Design for Intel® Cyclone® 10 GX

ID 683504
Date 6/08/2018
Public

1.1.1. Protocol Overhead

PCI Express Gen1 and Gen2 IP cores use 8B/10B encoding. Each byte of data is converted into a 10-bit data code, resulting in a 25% overhead. The effective data rate is therefore reduced to 2 Gbps or 250 MBps per lane for Gen1, and 4 Gbps or 500 MBps per lane for Gen2.

An active link also transmits Data Link Layer Packets (DLLPs) and Physical Layer Packets (PLPs). The PLPs are four bytes or one dword and consist of SKP Ordered Sets. The DLLPs are two dwords and consist of the ACK/NAK and flow control DLLPs. The ACKs and flow control update DLLPs are transmitted in the opposite direction from the Transaction Layer Packet (TLP). If link is transmitting and receiving high bandwidth traffic, the DLLP activity can be significant. The DLLPs and PLPs reduce the effective bandwidth available for TLPs. The format of the TLP illustrates that the overhead if a TLP is seven dwords. The overhead is six dwords if the optional ECRC is not included.

Figure 1.  TLP Format

The overhead includes the following fields:

  • Start and End framing symbols
  • A Sequence ID
  • A TLP header that is three or four dwords long,
  • The link cyclic redundancy check (LCRC).

The rest of the TLP contains 0–1024 dwords of data payload.