How NVIDIA GB200 Utilizes 800G/1.6T DAC/ACC
The current GB200 has a bidirectional bandwidth of 1800G, and based on a 1.6T configuration ratio, the ratio to
The NVIDIA GB200 NVL72 systems employ OSFP optical modules as expansion interfaces. Each OSFP module supports 16 differential signal lines, enabling 4 NVLink ports per module. A single NVLink switch in the GB200 contains 32 OSFP connectors, cumulatively supporting 128 NVLink 4 ports . These optical modules are integral to the system's high-bandwidth interconnect, allowing each GPU pair to achieve the 1.8 Tbps aggregate bandwidth required by NVLink 5.0 . The external interconnect of the NVL72 rack, however, uses direct copper backplane connections for intra-rack communication, while optical modules are primarily used for expansion and scaling .
While OSFP modules remain standard in current GB200 deployments, NVIDIA is exploring CPO technology to address thermal and power challenges in large-scale deployments. CPO integrates optical engines directly with the switch ASIC, reducing power consumption and improving signal density. Mass production of CPO-based switches is expected to support up to 115.2 Tbps signal transmission, with trial production already underway . This indicates a gradual shift toward CPO for future GB200 systems, but optical modules have not been fully abandoned.
The continued use of OSFP optical modules is driven by:
In summary, GB200 systems continue to use OSFP optical modules for NVLink expansion, particularly for high-bandwidth interconnects. The industry is moving toward CPO solutions to improve efficiency and thermal performance, but this represents a future evolution rather than a complete abandonment of optical modules .

The current GB200 has a bidirectional bandwidth of 1800G, and based on a 1.6T configuration ratio, the ratio to
Introduction The NVIDIA GB200 is a highly integrated supercomputing module based on
Discover the powerful GB200 NVL72 GPU, engineered for AI workloads and next-gen data centers.
NVIDIA also announced the Grace-Blackwell GB200 Superchip. This is a module that has two B200 GPUs wired to
With CX-7 on the GB200 NVL72, each GPU has 400G of bandwidth and is connected to one OSFP cage with the
This is the New NVIDIA GB200 NVL4 At SC24, the NVIDIA GB200 NVL4 platforms with two NVIDIA Grace CPUs and
Huawei AI CloudMatrix 384 – China''s Answer to Nvidia GB200 NVL72 China Abundance of Power, 100% Optics, 0%
Learn about the NVIDIA Blackwell GB200. Enterprise-scale AI with superior performance, advanced networking and
Supermicro NVIDIA GB200 NVL72 SuperCluster features the new advanced in-rack or in-row cooling distribution unit (CDU) and
Learn about the NVIDIA GB200 architecture, performance benchmarks, and use cases.
Nvidia has once again captured the spotlight with the announcement of its latest generation
Pictures of Nvidia''s new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on
The assertions made by financial analysts regarding the transition from optical to copper
This document contains detailed information about the hardware and software stack of DGX Grace Blackwell rack
A: The launch of GB200 is positive for the optical module industry, as it meets the demand for cross-cabinet
In this process, high-speed interconnect devices such as optical modules will continue to play a crucial role in providing
Conclusions: Microsoft''s demand for GB200 surpasses the total orders from other CSPs, with 4Q24 orders
Following the 800G optical modules, as AI server clusters demand higher interconnection speeds, Nvidia has chosen
Each GB200 subsystem has 2 * 18 = 36 NVLink5 Ports. The external interconnect of the
Understanding Your Grace-Blackwell Systems # GB200 NVL is a rack scale solution for GPUs connected using
1.6T optical transceivers are overtaking 800G in 2026 AI clusters because NVIDIA Blackwell GB200 systems require doubled
The GB200 Grace Blackwell Superchip is a key component of the NVIDIA GB200 NVL72, connecting two high
The fabric requirements for GB200 deployments push optical transceivers on latency, density, thermal
Nvidia suppliers building its Blackwell AI server racks have reportedly solved a series of technical hurdles, allowing
In this process, high-speed interconnect devices such as optical modules will continue to play a crucial role in
When NVIDIA unveiled the GB200 Grace Blackwell Superchip, I realized we were witnessing the next evolution of AI compute. As
Multiple power shelves are used in the rack to supply redundant power. The rack power consumption is approximately
Our photonic engineering team can help you select the right connector or splitter for your network.