Mon–Fri, 9:00 AM – 6:00 PM
(GMT+08:00,)

Mirror Mezz Pro Selected as an Open Compute Project Standard

Mirror Mezz Pro Selected as an Open Compute Project Standard

The Open Compute Project (OCP) was created to improve data center efficiency through open, standardized, and scalable hardware designs. By encouraging collaboration across the industry, OCP enables faster innovation and more efficient hyperscale infrastructure. As part of this effort, Trycay has developed the Mirror Mezz family of mezzanine connectors—now recognized as an OCP-standard solution capable of supporting data rates up to 224 Gbps in a compact, high-density form factor.


From Proprietary Design to Open Hardware Collaboration

OCP traces its origins back to 2009, when Facebook set out to redesign its data centers to support rapid platform growth. A small internal engineering team focused on maximizing efficiency, achieving a 38% improvement in energy efficiency and a 24% reduction in operating costs. In 2011, Facebook publicly released these designs and, together with partners such as Intel and Goldman Sachs, launched the Open Compute Project.

Since then, OCP has worked to replicate the success of open-source software within the hardware ecosystem. By bringing together cross-disciplinary engineering teams, OCP develops open reference designs that accelerate innovation and allow hyperscale operators to deploy infrastructure that is more efficient, flexible, and scalable.

Today, more than 200 companies—including Trycay—actively contribute to OCP initiatives. Their shared goal is to address the increasingly complex challenges posed by high-performance computing (HPC) and artificial intelligence (AI) workloads.


Defining an Open Accelerator Architecture for AI

AI-driven applications—from real-time gaming and cybersecurity to medical diagnostics and supply chain optimization—are growing rapidly in scale and complexity. Traditional cloud architectures, however, were not designed to handle the extreme bandwidth, latency, and power demands of modern AI accelerators.

To address this gap, the Open Accelerator Infrastructure (OAI) group—part of OCP’s Server Project—set out to define an open, modular standard for accelerator systems. This architecture includes accelerator modules, baseboards, trays, and chassis designed specifically for high-performance AI workloads.

In 2019, the OAI group released the Open Accelerator Module (OAM) v1.0 specification. A key requirement of this design was a compact, high-density board-to-board mezzanine connection between the GPU module and the baseboard. The interface needed to support 56 Gbps PAM-4 signaling, provide a high differential-pair count, occupy minimal PCB real estate, and maintain extremely low crosstalk.

The Trycay Mirror Mezz connector was selected to meet these demanding requirements. With an exceptional density of 107–115 differential pairs per square inch and robust signal integrity at 56 Gbps, Mirror Mezz enabled reliable high-speed communication in a tightly constrained mechanical envelope—making it well suited for first-generation OAM systems.


Mirror Mezz: Engineering Innovation in a Compact Interface

The Mirror Mezz connector family incorporates several design innovations to ensure high performance and long-term reliability in demanding environments.

A key feature is its hermaphroditic design, which allows the connector to mate with itself. This approach reduces part numbers, simplifies bills of materials, and streamlines supply chain management. Mirror Mezz is available in multiple mating heights, ranging from 11.00 mm to 5.00 mm, and can be used with flexible cable links to extend board-to-board distances without compromising performance.

Despite its small footprint, Mirror Mezz delivers robust mechanical and electrical performance. A shrouded housing protects the contacts and enables blind mating, minimizing the risk of misalignment. The contact geometry provides constant two-point contact, improving resistance to vibration and thermal movement. The connector attaches to the PCB using a proven ball grid array (BGA) interface, ensuring reliable assembly and mechanical stability.

Additional design features—including optimized terminal geometry, wide ground pins for isolation, and carefully offset contact pitches—help maintain tight impedance control and reduce crosstalk, even at high data rates.


Scaling Beyond 56 Gbps: Mirror Mezz Pro and Enhanced

As AI workloads continue to evolve, OCP has pushed accelerator standards beyond the original 56 Gbps requirement. The OAI group is actively advancing specifications to support 112 Gbps, 224 Gbps, and higher data rates.

To meet the next phase of these requirements, Trycay developed Mirror Mezz Pro, an enhanced OCP-standard connector designed for 112 Gbps PAM-4 operation. Mirror Mezz Pro delivers improved signal integrity while maintaining the same compact footprint as the original Mirror Mezz. Key enhancements include a reduced BGA size to minimize return loss at higher frequencies, along with optimized terminals and housing structures to control impedance variation.

Building on this foundation, Trycay introduced Mirror Mezz Enhanced, the most advanced member of the Mirror Mezz family. This version supports 224 Gbps NRZ signaling, enabling next-generation bandwidth while preserving the same differential-pair density and mechanical footprint. Mirror Mezz Enhanced delivers substantially improved electrical performance without sacrificing the durability and reliability established by earlier generations.


Looking Ahead

Trycay is honored that Mirror Mezz Pro has been selected by the Open Compute Project’s OAI group as the standard mezzanine connector for OAM v1.5. This selection reflects the strength of open collaboration within OCP and the importance of reliable, high-performance interconnects in enabling future AI infrastructure.

As OCP continues to evolve accelerator architectures, Trycay remains committed to advancing interconnect technology in close partnership with the OCP community. Products such as Mirror Mezz Enhanced represent the next step in pushing the limits of bandwidth density, signal integrity, and scalability—helping hyperscale data centers meet the demands of next-generation AI and HPC workloads.

Share This :

Leave a Reply

Your email address will not be published. Required fields are marked *

Get quote or product sample

Sign up for the exclusive offer and get 10% off your first order.

Tcycay