PAM4 vs NRZ: What Are They and Why Is High-Speed Transmission Moving from NRZ to PAM4?｜Technology Year：2021

As high-speed interfaces have evolved from 25G and 50G per lane toward 100G, 200G, 400G, 800G, and beyond, one engineering challenge has become increasingly important:

How can more data be transmitted without continuously pushing signal frequencies higher and making PCB, connector, and cable channels increasingly difficult to manage?

This is one of the main reasons high-speed interconnects have gradually shifted from NRZ toward PAM4 signaling.

NRZ uses two signal levels and carries one bit per symbol. PAM4 uses four amplitude levels and carries two bits per symbol. As a result, PAM4 can approximately double the amount of information transmitted at the same symbol rate.

However, PAM4 is not simply a “faster and better” replacement for NRZ. Its smaller voltage margin increases sensitivity to noise, reflection, crosstalk, jitter, and distortion.

The real engineering question is therefore not:

“Is PAM4 better than NRZ?”

but:

Which signaling method provides the best balance of data rate, channel loss, reach, power, latency, and Signal Integrity for the application?

NRZ vs PAM4: Key Differences

The most important concept is simple:

NRZ transmits 1 bit per symbol, while PAM4 transmits 2 bits per symbol.

PAM4 achieves this by using four amplitude states that can represent:

00, 01, 10, and 11

The “4” in PAM4 therefore does not mean four bits per symbol. Four possible states represent 2 bits of information.

Why Can’t High-Speed Interfaces Keep Using Faster NRZ?

Consider a simplified example.

If a system needs approximately 50 Gb/s of data throughput:

• NRZ may require close to 50 GBaud
• PAM4 can deliver similar information capacity at approximately 25 GBaud

This matters because as symbol rate and signal frequency rise, the complete electrical channel becomes increasingly difficult to maintain.

A high-speed signal may pass through:

SerDes → PCB Trace → Connector → Cable → Connector → PCB → Receiver

As frequency increases, the system becomes more sensitive to:

• Insertion Loss
• Return Loss
• Crosstalk
• PCB Material Loss
• Connector Discontinuity
• Cable Loss
• Jitter
• Equalization Limitations

This is why simply increasing the NRZ baud rate indefinitely is not practical.

PAM4 increases the amount of data carried by each symbol and therefore provides another way to increase bandwidth without making the required symbol rate rise at the same rate.

Bit Rate vs Baud Rate

These terms are often confused.

Bit Rate describes how many bits are transmitted per second.

Baud Rate, or Symbol Rate, describes how many symbols are transmitted per second.

For simplified comparison:

Actual interface throughput also depends on encoding, FEC, protocol overhead, lane count, and PHY architecture.

Why Is Nyquist Frequency Important?

One of PAM4's major advantages is its ability to increase data rate without doubling the Nyquist frequency.

PCIe provides a clear example:

PCIe 6.0 doubles the transfer rate but uses PAM4 so that the fundamental Nyquist-frequency requirement remains at approximately the same level as PCIe 5.0.

This does not mean the PCIe 6.0 channel is easy to design. It means PAM4 avoids the even more severe frequency and channel-loss pressure that would occur if the same data-rate increase were attempted using much faster NRZ signaling.

Why Is PAM4 More Difficult Than NRZ?

PAM4's higher information density comes with a major trade-off:

smaller decision margins.

NRZ uses two principal voltage levels and produces one eye opening in an Eye Diagram.

PAM4 uses four voltage levels and produces three eyes.

The same overall voltage range must therefore be divided into more decision regions.

Conceptually:

NRZ
→ larger single eye

PAM4
→ three smaller eyes

As a result, PAM4 is more sensitive to:

• Noise
• Crosstalk
• Reflection
• Jitter
• Transmitter non-linearity
• Receiver sensitivity
• Impedance discontinuity
• Channel distortion

PAM4 does not eliminate Signal Integrity problems. Instead, it changes the engineering trade-off:

Less pressure on symbol rate and channel frequency, but greater pressure on signal quality, linearity, equalization, and error correction.

This is the key reason PAM4 requires more advanced SerDes and channel design.

Why Is FEC Important for PAM4?

Because PAM4 voltage levels are closer together, raw error performance can be more challenging than comparable NRZ signaling.

High-speed PAM4 systems therefore commonly rely more heavily on:

FEC — Forward Error Correction

FEC adds error-correction capability to recover a certain amount of corrupted data.

PCIe 6.0, for example, introduced PAM4 together with technologies including lightweight FEC, CRC, and FLIT-based operation.

However:

FEC does not replace proper Signal Integrity design.

A severely degraded cable, connector, or PCB channel cannot simply be “fixed” by adding FEC.

A reliable high-speed link still requires:

Signal Integrity + Equalization + Error Correction

to work together.

Is PAM4 Replacing NRZ Completely?

No.

The transition from NRZ to PAM4 should not be understood as:

“NRZ is obsolete and PAM4 is always better.”

Both technologies remain useful.

NRZ may still be attractive when a system prioritizes:

• Simpler architecture
• Larger voltage margin
• Lower implementation complexity
• Short reach
• Lower latency
• Lower signal-processing requirements

PAM4 becomes increasingly attractive when the application needs:

• Higher data rate per lane
• Better bandwidth density
• Reduced symbol-rate growth
• More efficient use of a limited number of electrical lanes

The choice therefore depends on:

Data Rate + Reach + Channel Loss + Power + Latency + BER + Cost

rather than modulation technology alone.

How PAM4 Changed Ethernet and QSFP Interconnects

One of the clearest examples of PAM4 adoption can be seen in Ethernet.

Earlier high-speed Ethernet generations commonly used approximately:

25G/Lane NRZ

The industry then moved toward:

50G/Lane PAM4

which enabled higher aggregate bandwidth without requiring the number of electrical lanes to double.

This can also be seen in QSFP interfaces:

The lane count remains four, but the data capacity per lane approximately doubles.

This is why the transition from QSFP28 to QSFP56 is not mainly about adding more lanes.

It is about increasing:

Data Rate per Lane

through higher-density signaling.

What Does PAM4 Mean for High-Speed Cables and Connectors?

Although PAM4 reduces the symbol rate required for a given data rate, it does not make cable or connector design easier.

Because the voltage decision margin is smaller, the complete channel must control Signal Integrity more carefully.

Key parameters include:

A cable that passes a continuity test is not automatically suitable for a PAM4 link.

Likewise, the connector is not merely a mechanical interface.

At high data rates, its:

• Contact geometry
• Differential pair arrangement
• Ground structure
• PCB launch
• Shielding
• Impedance continuity

become part of the complete high-speed transmission channel.

The correct engineering model is therefore:

SerDes + PCB + Connector + Cable + Receiver

rather than evaluating the cable alone.

Why Is PAM4 Important for AI Data Centers?

AI infrastructure requires enormous amounts of high-speed communication between:

GPUs, NICs, DPUs, switches, servers, and racks.

As network bandwidth grows from:

100G → 200G → 400G → 800G and beyond

simply increasing the number of lanes creates other challenges:

• More connector contacts
• More cable pairs
• Lower port density
• More complex PCB routing
• Higher power requirements
• Greater thermal pressure

Increasing the data rate per lane therefore becomes essential.

PAM4 is one of the key technologies enabling this transition.

Instead of relying only on more lanes or dramatically higher NRZ symbol rates, the industry can increase information density per lane.

This is why PAM4 has become important across:

• High-speed Ethernet
• PCIe
• AI servers
• HPC
• Data center switches
• High-speed copper cables
• Optical interconnects

PAM4 vs NRZ: Key Takeaways

The difference can be summarized simply:

NRZ uses two signal levels and carries 1 bit per symbol.

PAM4 uses four signal levels and carries 2 bits per symbol.

At the same symbol rate, PAM4 can therefore provide approximately twice the information capacity.

This helps high-speed interfaces increase data rate without making Nyquist frequency and channel bandwidth requirements rise proportionally.

The trade-off is greater complexity:

**Smaller Eye Margin

• Higher Noise Sensitivity
• More Difficult Signal Integrity
• Greater Equalization and FEC Requirements**

Therefore, the industry's move from NRZ toward PAM4 is not because NRZ is technically “bad.”

It is because:

As data rate per lane continues to increase, simply raising NRZ symbol rate creates increasingly difficult channel-loss and bandwidth challenges. PAM4 provides another engineering path to increase data density.

High-speed interconnect design must ultimately evaluate:

**Modulation

• Lane Rate
• Signal Integrity
• Cable
• Connector
• Complete Channel**

together.

SyncTronics Solutions

As high-speed interfaces transition from NRZ to PAM4, cable and connector selection must consider more than form factor and nominal data rate. Lane rate, characteristic impedance, insertion loss, return loss, crosstalk, cable length, connector design, and equipment compatibility all affect the complete transmission channel. SyncTronics provides high-speed cable and interconnect solutions including SFP28, QSFP28, QSFP56, Mini-SAS HD, PCIe, and customized Cable Assembly solutions for networking, server, AI infrastructure, and industrial applications.

Need Wire Harnesses, Connectors, Electromechanical Components, or Custom Solutions? Contact Us Today

Whether you are looking for connectors, cables, wire harness assemblies, electromechanical components, or complete electronic interconnect solutions, SyncTronics provides professional technical support and engineering evaluation for product selection, specification verification, New Product Introduction (NPI), OEM/ODM customization, as well as high-speed transmission, high-reliability, and other specialized application requirements.

We welcome you to visit our Contact Us page and share your project requirements and contact information. Our team is committed to responding quickly and professionally, evaluating the most suitable products and electronic interconnect solutions based on your product specifications, application requirements, and project objectives. We also provide technical recommendations and customized engineering support to help shorten product selection and development time, reduce project risks, and accelerate your path to mass production. We look forward to the opportunity to work with you and provide reliable technical support and practical solutions for every stage of your project.

Related Articles

What Is Signal Integrity? High-Speed Cable Loss & Impedance Explained | SyncTronics Technology

QSFP28 vs QSFP56 vs QSFP-DD | Speed & Lane Comparison | SyncTronics Technology

AI Data Center Cables | DAC, AEC, AOC & Optical Transceiver Comparison | SyncTronics Technology

AI Data Center Networking | 400G, 800G & 1.6T Ethernet Explained | SyncTronics Technology

Disclaimer and Technical Exchange Statement

This content is compiled based on publicly available information and industry experience, intended solely for reference during technical exchanges. It does not constitute any commercial, technical, or legal advice or basis. The specifications, materials, manufacturing processes, application methods, and related information described herein may vary due to differing design conditions, environmental factors, or usage scenarios. No guarantees are made regarding the completeness, correctness, or suitability of such content.
Readers should verify the suitability with the responsible design or technical personnel based on actual application scenarios, and independently assess all associated risks.