Editor's Choice

The switch to 400G Ethernet

30 August 2023 Editor's Choice Telecoms, Datacoms, Wireless, IoT

We live in a connected world where communication devices form a global network. We share and consume data freely and rely on telecommunications infrastructure to deliver the performance needed by the latest generation of equipment.

The ever-increasing appetite for data communications drives the need for far greater capabilities. The Internet of Things (IoT), advances in artificial intelligence (AI), and machine learning (ML), and the growth of 5G wireless communications all mean that the existing networks will be unable to scale for future demand.

The move to 400G

The current standard for Ethernet networks is 100 Gigabit Ethernet (referred to as 100G, 100GbE, or 100GbE), the latest in many protocols adopted as data speeds have increased. As the name suggests, it supports data transfer rates of 100 gigabits per second (Gbps) and is governed by the IEEE 802.3 standard. It is widely adopted in data centres, high-performance computing environments, and other applications that process large amounts of data.

The current 100G and next-generation 400 Gigabit Ethernet (400G) are both high-speed networking standards that are used to transfer data between devices. However, there are some key differences between them.

The most apparent difference is speed. As the name implies, 400G Ethernet is four times faster than the existing system. Not only does this result in higher communication speeds, but it also creates greater bandwidth. With the demands being placed on our modern communication systems, greater bandwidth is vital to support the latest data-intensive applications and services.

400G Ethernet has been created with the data centre in mind. It has been designed to handle more data-intensive workloads, such as big data analytics, AI, and high-definition video streaming, which are becoming increasingly common.

Achieving higher performance

The adoption of new signal modulation techniques is key to the improved performance of the 400G network. Traditional data communications have depended on binary communications. Digital signals are transmitted in binary states (zeroes and ones), often referred to as non-return to zero (NRZ). To expand the capacity, the latest systems use a technique known as four-level pulse amplitude modulation (PAM4). PAM4 is a signal modulation technique that uses four different signal levels to represent data instead of two. With this technique, the capacity or bandwidth can be doubled, but at the cost of requiring specialised hardware and cabling. The reason that the hardware needs to be upgraded is to preserve signal integrity.

Signal integrity (SI) is the name given to the techniques and technology applied to preserve the quality of digital signals sent across wired, optical, or wireless media. For a signal to be received accurately, the medium over which it is sent must not degrade it to the point where the difference between the different states cannot be determined.

For traditional NRZ signals that are transmitted in two states, the voltage difference between the states is easier to detect. Therefore, the system is more tolerant of electronic noise or interference. However, PAM4 uses four different states, and the difference in voltage between each state is smaller. Consequently, the integrity of the signal is more at risk of interference. Designers must turn to more advanced technology to protect the SI.

As the medium over which the signals are sent has a significant impact on the performance of the system, the design of the transceivers, connectors, and cabling plays an essential role in preserving SI. The connectors used for 400G Ethernet are evolutions of the types used on older systems, often with new features.

New form factors

One such evolution is Quad Small Form-factor Pluggable Double Density (QSFP-DD), a compact, high-density transceiver module form factor that supports data rates up to 400 Gbps. Developed from the QSFP28 module, QSFP-DD features twice as many channels as the older design, but without an increase in overall size. When combined with the doubling of capacity achieved through PAM4 signalling, QSFP-DD can support the 400G network transmission for the newest systems.

The quest for more incredible speed cannot only be achieved by incorporating more channels into existing connectors. To generate the higher rates needed for 400G Ethernet, the processors require greater power and, thus, more heat.

To cater to this increased performance, the new Octal Small Form-factor Pluggable (OSFP) is a pluggable form factor with eight high-speed electrical channels that support 400 Gbps. It is slightly broader and deeper than the existing QSFP design. It has an integrated heat sink that significantly improves thermal performance, and enables modules with up to 15 W power in a switch chassis without additional thermal management.

These new form factors are enabling system architects to take advantage of higher port densities. The latest-generation transceivers using QSFP-DD and OSFP devices offer higher port densities than previous generations, allowing data centres to conserve space and reduce costs. In addition, the superior performance of these new form factors enables system architects to scale their networks more efficiently to meet the growing demands of applications and services. This helps ensure that data centres continue operating effectively and efficiently as their needs evolve.

Barriers to entry

With so much additional performance available, the benefits of 400G Ethernet are clear. However, moving from 100G to 400G Ethernet is not straightforward. The principal barrier to the adoption of this new technology is cost. Implementing 400G Ethernet requires significant investment in equipment and infrastructure. The cost of new transceivers, switches, and other devices is relatively high compared to previous generations of Ethernet. In addition, as 400G Ethernet is still relatively new, not all vendors can produce the necessary equipment.

The need for additional training also causes a significant cost impact. Making the most of the greater performance on offer requires a new set of skills, and organisations will need to make considerable investments in the training of their staff to deploy and manage 400G Ethernet networks correctly.

Organisations must also consider the compatibility of the existing networks when deploying new technology. The transceivers and connectors required may not be compatible with existing equipment, making it difficult for organisations to upgrade networks to support 400G Ethernet.

Finally, many companies will be wary of adopting new equipment in the early stages of its development, and 400G Ethernet standards are still evolving. Companies may delay purchasing decisions because they are unable to determine which equipment and technologies comply with the latest standards. As this technology is still in its early stages, few existing customers are employing it. The lack of real-world testing can be a concern for companies who feel unable to assess its performance and reliability.

Beyond 400G

As powerful as 400G Ethernet is, it is the latest development on the path to the faster systems of the future. The next step after 400G is already being considered, with 800G and 1,6 Terabit Ethernet. These faster systems will be able to support even more data-intensive applications, such as real-time video streaming, virtual reality, and other applications that require extremely low latency and high bandwidth.

It is also important to remember that this new technology will also play a key part in the roll-out of robust 5G networks, as the high speed and low latency required for 5G communications are similar to the demands of 400G Ethernet.


400 Gigabit Ethernet is a relatively new technology, and its adoption is still in the early stages. Although not yet widely adopted, it is expected to gain broader adoption in the coming years.

The key drivers behind the adoption of this new technology will be the demands created by data-intensive applications, such as AI, cloud computing, big data analytics, and high-definition video streaming. These applications are becoming increasingly popular, and as a result, the demand for high-bandwidth, low-latency connections will also increase.

Some early adopters, such as large internet and cloud service providers, already use this new technology in their data centres. For other organisations, the relatively high cost of new equipment and the skills required to implement them to their full potential are significant concerns.


Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

Products of the year 2023
Editor's Choice News
A summary of products and technologies from 2023 that are currently shaping the electronic engineering industry.

The dream of Edge AI
Altron Arrow Editor's Choice AI & ML
AI technology carries a great promise – the idea that machines can make decisions based on the world around them, processing information like a human might. But the promise of AI is currently only being fulfilled by big machines.

Save space and improve efficiency with a 4-switch buck-boost controller
Altron Arrow Editor's Choice Telecoms, Datacoms, Wireless, IoT
A new generation of 4-switch buck-boost controllers enables power system design with very high efficiency and high power density by using inductor DC resistance (DCR) current-sensing techniques.

Matter 1.2 adds new capabilities
NuVision Electronics Editor's Choice Telecoms, Datacoms, Wireless, IoT
This update introduces new device types and expands the reach of Matter into new markets, while also bringing other improvements that enhance interoperability and user experience.

Personality Profile: Paul Eveleigh
Hiconnex Editor's Choice
With a strong and experienced team behind him, he can put into practice what he has learned during his career and ultimately grow the company towards its strategic objectives.

FPGAs speed up intelligent edge designs
ASIC Design Services Editor's Choice DSP, Micros & Memory
Microchip Technology has added nine new technology- and application-specific solution stacks to its growing collection of mid-range FPGA and SoC support.

Raspberry Pi 5 announced
RS Components SA Editor's Choice DSP, Micros & Memory
The R-Pi 5 comes with new features and is over twice as fast as its predecessor, the R-Pi 4, thanks to an upgrade to the main processor.

Dead-reckoning GNSS module
Altron Arrow Editor's Choice Telecoms, Datacoms, Wireless, IoT
STMicroelectronics’ Teseo-VIC3D module is an easy-to-use dead-reckoning global navigation satellite system (GNSS) standalone module, embedding TeseoIII single-die standalone positioning receiver IC.

Quantum leap in battery technology
Editor's Choice Power Electronics / Power Management
A German start-up has developed a solid-state battery ready for series production with outstanding properties: no cobalt, ten times longer service life, and non-flammable electrolyte.

Simplify your USB-C PD design using a standalone controller
Altron Arrow Editor's Choice
USB PD poses new power requirement challenges because of the variety of voltage and current combinations available to supply the wide range of power the USB PD standard must provide.