In the high-speed world of PCI Express (PCIe), the primary function of the Data Link Layer is to ensure the absolutely reliable delivery of Transaction Layer Packets (TLPs) across the link. To maintain a stable system, the PCIe specification requires a strict Bit Error Rate (BER) of no worse than 10−12.
However, when you are transmitting billions of bits per second, transient errors are inevitable. Here is a look at why a single bit error is so dangerous and how the Data Link Layer’s Ack/Nak protocol serves as the ultimate safety net.
The Threat of a Single Bit Error
At gigatransfer speeds, physical interference or electrical noise will occasionally cause a bit to flip. While one bit might seem insignificant, a single bit error will corrupt an entire packet.
Because TLPs carry critical memory read/write commands, complex headers, and exact data payloads, an uncorrected bit error could result in data being written to the wrong memory address or the system crashing entirely. Furthermore, as PCIe link rates continue to increase with each new generation, this problem will only become more pronounced, making errors more frequent and harder to catch.
The Solution: Hardware-Based Reliable Transport
To combat these inevitable errors and maintain that pristine 10−12 BER, the Data Link Layer relies on an automatic, hardware-based mechanism: the Ack/Nak Protocol.
To facilitate this reliable transport, the Data Link Layer takes two crucial steps before sending a TLP:
- Sequence Numbers: It assigns each packet a unique incremental Sequence Number, making it easy to sort out exactly which packet encountered an error out of a massive stream of traffic.
- LCRC Addition: It calculates and adds a 32-bit error detection code called a Link Cyclic Redundancy Code (LCRC) to the end of each TLP.
Catching Errors and Replaying Packets
The first step in error checking is simply for the receiver to verify that the LCRC still evaluates correctly upon arrival.
- If the packet arrives perfectly, the receiver returns an Ack (Acknowledge) DLLP to confirm good reception.
- If the LCRC fails—meaning a bit error occurred—the receiver returns a Nak (Negative Acknowledge) DLLP to indicate a transmission error.
When the transmitter receives a Nak, it knows exactly which packet failed thanks to the Sequence Numbers, and it automatically re-sends (replays) the corrupted TLP in hopes of a better result.
Why a Simple “Replay” Works
You might wonder why simply trying again fixes the problem. The PCIe spec relies on the fact that things that cause a transmission error are likely transient events—such as a brief spike in electrical noise. Because the interference is temporary, a simple replay will have a very good chance of solving the problem and successfully delivering the packet on the second try.
Summary The Ack/Nak protocol is the guardian of PCIe data integrity. By combining Sequence Numbers, LCRC checks, and automatic replays, it ensures that even when transient bit errors corrupt an entire packet, the system can seamlessly recover and maintain reliable, uninterrupted transport.