800G QSFP-DD VS OSFP 800G

In today’s fast-paced world, the demand for high-speed data transmission has reached unprecedented levels. The emergence of AI applications and large models has made computational power a critical infrastructure for the AI industry. With the ever-increasing need for faster communication, high-speed optical modules have become essential components of AI servers. This article delves into the evolution of 800G optical modules and their vast potential in the era of artificial intelligence.

The evolution of 800G optical modules.

Optical modules undertake the task of photoelectric signal conversion in network connections, responsible for converting electrical signals into optical signals at the transmitting end, and then transmitting them through optical fibers before converting the optical signals back into electrical signals at the receiving end. With the development and integration of optoelectronic devices, their performance and transmission bandwidth are continuously improving. Optical modules now require higher transmission rates and smaller sizes to adapt to various usage scenarios. The packaging methods are also evolving, with smaller packaging and lower power consumption meaning that optical modules can achieve higher port density on switches, allowing the same amount of power to drive more optical modules.

Ever-increasing bandwidth demand.

The growth in bandwidth demand has had a significant impact on high-speed optical modules. With the continuous emergence of new technologies and the need for large-scale data transmission, traditional 100G, 200G, and 400G optical modules can no longer fully meet market demands. To address the ever-increasing bandwidth requirements, 800G optical modules are becoming the trend.

The growth of LPO technology.

In the era of 800G optical modules, Linear-drive Pluggable Optics (LPO) technology has emerged as a standout. LPO utilizes linear analog components in the data link, eliminating the need for complex CDR or DSP designs. Compared to DSP solutions, LPO significantly reduces power consumption and latency, making it highly suitable for the short-distance, high-bandwidth, low-power, and low-latency data connectivity requirements of AI computing centers. As cloud service providers expand their computing resources, LPO solutions, including 800G LPO, are expected to capture a significant share of the market.

800G optical module packaging.

With the continuous advancement of technology, the packaging forms of optical modules have undergone significant evolution. From the early GBIC packaging to the smaller SFP packaging, and now to the current 800G QSFP-DD and OSFP packaging. This development trend not only reflects the continuous increase in the speed of optical modules but also demonstrates their progression towards miniaturization and hot-swappable capabilities. The application scenarios for 800G optical modules are becoming increasingly widespread, covering various fields such as Ethernet, CWDM/DWDM, connectors, fiber channels, and wired/wireless access.

800G QSFP-DD form factor

The Double Density Quad Small Form-factor Pluggable High-Speed Module. QSFP-DD is currently the preferred packaging for 800G optical modules, enabling data centers to efficiently grow and expand cloud capacity as needed. QSFP-DD modules utilize an 8-channel electrical interface, with each channel capable of speeds up to 25Gb/s (NRZ modulation) or 50Gb/s (PAM4 modulation), providing an aggregate solution of up to 200Gb/s or 400Gb/s.

The advantages of 800G QSFP-DD

• It is backward compatible, supporting QSFP+/QSFP28/QSFP56 QSFP packaging.
• It employs a 2×1 stacked integrated cage connector, which can support both single-height and double-height cage connector systems.
• Through the use of SMT connectors and 1xN cages, it achieves a thermal capacity of at least 12 watts per module. The higher thermal capacity reduces the cooling requirements for the optical modules, thereby cutting down on some unnecessary costs.
• In the design of QSFP-DD, the MSA working group has fully considered user flexibility, incorporating ASIC design that supports multiple interface rates and is backward compatible (compatible with QSFP+/QSFP28), thus reducing port costs and equipment deployment costs.

800G OSFP Form Factor

OSFP is a new type of optical module, significantly smaller than CFP8 but slightly larger than QSFP-DD, featuring 8 high-speed electrical channels. It still supports 32 OSFP ports per 1U front panel, and when paired with integrated heat sinks, it can greatly enhance thermal performance.

Advantages of 800G OSFP

• OSFP modules are designed with 8 channels, directly supporting a total throughput of up to 800G, thereby achieving higher bandwidth density.
• Due to the OSFP packaging supporting more channels and higher data transmission rates, it can provide higher performance and longer transmission distances.
• OSFP modules feature excellent thermal design, capable of handling higher power consumption.
• OSFP is designed to support even higher rates in the future. Because OSFP modules are larger in size, they have the potential to support higher power consumption, thereby supporting higher rates such as 1.6T or beyond.

800G Optical Module Form Factor Comparison

QSFP-DD is typically the preferred choice in telecommunications applications, while OSFP is more suitable for data center environments. The main differences between the two are:

• Size: OSFP is slightly larger in size.
• Power Consumption: OSFP has a slightly higher power consumption than QSFP-DD.
• Compatibility: QSFP-DD is perfectly compatible with QSFP28 and QSFP+, whereas OSFP is not.

Types of 800G Optical Modules

800G = 8100G = 4200G, therefore, based on the single-channel rate, it can be divided into two categories, namely single-channel 100G and 200G. The corresponding architecture is shown in the figure below. Single-channel 100G optical modules can be quickly realized, while 200G places higher demands on optical components. Since the maximum rate currently supported by electrical interfaces is 112Gbps PAM4, for the case of single-channel 200G, a gearbox is required for conversion.

For the multimode case, there are two main standards for 800G optical modules, corresponding to transmission distances less than 100m.

800G SR8

It employs a VCSEL solution with a wavelength of 850nm and a single-channel rate of 100Gbps PAM4, requiring 16 fibers. This can be seen as an upgraded version of the 400G SR4, with the number of channels doubled. Its optical interface is either MPO-16 or two rows of MPO-12, as shown in the figure below. The 800G SR8 optical module is typically used for 800G Ethernet, data center links, or 800G-800G interconnections.

800G SR4

The solution uses 850nm/910nm wavelengths for bi-directional transmission, utilizing DeMux in the module to separate the two wavelengths. The single-channel rate is 100Gbps PAM4 and requires 8 fibers. Compared with SR8, the number of fibers in this scheme is reduced by half. Its block diagram is shown below:

Its fiber optic interface is shown below with MPO-12 pin.

For the single-mode case, there are various standards for 800G optical modules:

800G DR8, 800G 2xDR4, 800G PSM8

All three standards have similar internal architectures, with 8 transmitters and 8 receivers, a single channel rate of 100 Gbps, and the need for 16 fibers.

800G DR8 Optical Module adopts 100G PAM4 and 8-channel single-mode parallel technology, the transmission distance through single-mode fiber can reach 500m, usually applied in data centers, 800G-800G, 800G-400G, 800G-100G interconnections.

The 800G PSM8 adopts CWDM technology with 8 optical channels, each with a transmission rate of 100Gbps, supporting a transmission distance of 100m, making it ideal for long-distance transmission and fiber resource sharing.

800G 2DR4 refers to two “400G-DR4” interfaces, 2DR4 optical interface is two MPO-12, as shown in the figure below, can be interconnected with 400G DR4 optical module, no fiber branch cable, support 500m transmission distance, convenient for data center upgrades.PSM8 and DR8 optical interface is MPO-16. MPO-16.

800G FR8

These two solutions are upgrades of 400G FR4 and LR4 optical modules using CWDM4 wavelengths of 1271/1291/1311/1331nm. 2xFR4 supports transmission distance of 2km, and 2xLR4 supports transmission distance of 10km. their optical interfaces are either dual CS or dual duplex LC interfaces.

Al’s Impact on 800G Optical Module Deployment

First, AI servers require high data rates and low latency, requiring top-of-rack switches that match the underlying bandwidth. These switches may also require latency redundancy, which requires high-speed optical modules. For example, the NVIDIA DGX H100 server comes with eight H100 GPU modules, each of which requires two 200G optical modules. Therefore, each server requires at least 16 200G modules, and the corresponding top-of-rack switch ports require at least 4 800G.

Second, 800G optical chips are more cost efficient and economical. They use 100G EML chips, while 200G/400G use 50G optical chips. Data shows that one 100G optical chip costs 30% less than two 50G optical chips at the same rate.

Nonetheless, 400G optical modules remain significant in the industry. While they may not be able to match the speed of 800G optical modules, they offer a significant increase in bandwidth over older technologies and are the preferred solution for many organizations. Additionally, some applications may not need the full functionality of 800G Ethernet, making 400G Ethernet more practical for them.