Recently, exciting news arrived from the compatibility testing laboratory: several self-developed server motherboards from our company have successfully completed product compatibility certification with two leading domestic CPU platforms and one domestic AI accelerator chip.

Test reports show that the products demonstrated outstanding performance in functionality, stability, compatibility, and overall system operation, fully meeting the requirements for large-scale commercial deployment.

This marks another critical milestone in our strategic expansion within the domestic computing ecosystem.

Why Invest in Domestic Platform Adaptation?

For many years, the global server market was dominated by international giants and x86 architectures.

Whether for enterprise IT infrastructure or hyperscale data centers, x86 was almost always the default choice when purchasing servers.

However, in recent years, with evolving international dynamics and the rapid acceleration of domestic technological innovation, Chinese-developed CPUs and AI chips have entered a new stage of growth.

From breakthroughs in independent instruction set architectures to the gradual maturity of software ecosystems, domestic processors now possess the foundation necessary for large-scale commercial deployment.

As a professional server OEM/ODM manufacturer, we recognized early on that the future market would no longer belong to a single architecture.

Instead, it will become a diversified ecosystem where multiple architectures coexist.

“Our customer base includes not only international internet giants, but also government institutions, state-owned enterprises, and research organizations,”

said Mr. Chen, our Product Director.

“For many of these customers, supply chain security and technological self-reliance are now essential requirements. If we cannot provide server solutions based on domestic chip platforms, we would effectively lose access to a critical market segment.”

Based on this strategic vision, the company launched a dedicated domestic-platform adaptation initiative two years ago.

Today, that initiative has achieved significant phased results.

Deep Adaptation with Domestic CPUs

One of the CPU platforms certified in this compatibility program comes from a leading domestic semiconductor company focused on server and data center applications using its own proprietary instruction set architecture.

But compatibility work involves far more than simply “making the system boot.”

Hardware-Level Signal Integrity Optimization

Different processors have different signal requirements.

Memory routing, PCIe lane allocation, and power delivery topology all require redesign and optimization.

Based on our original motherboard architecture, our hardware engineering team performed three PCB revisions specifically tailored to this domestic CPU platform.

Through simulation and real-world validation, we ensured that:

• Signal reflection

• Crosstalk

• Attenuation

• High-speed transmission integrity

all remained within strict specification ranges.

Firmware-Level Customization

Domestic CPU platforms are often paired with localized firmware ecosystems.

Our firmware engineers worked side-by-side with the CPU vendor’s technical team for nearly two months to complete:

• UEFI porting

• Firmware optimization

• Enterprise management feature integration

The resulting firmware platform now supports:

• Standard boot functionality

• Remote management

• Out-of-band monitoring

• Online firmware upgrades

• Enterprise-grade system management capabilities

Operating System Compatibility Validation

We also completed extensive testing across major domestic operating systems, including:

• Kylin OS

• UOS

• openEuler

Validation covered:

• File systems

• Network protocols

• Virtualization

• Container environments

• Enterprise application compatibility

All test cases passed successfully, and the system maintained stable operation during continuous 72-hour stress testing.

Collaborative Innovation with Domestic AI Chips

Another important achievement is our collaboration with a domestic AI accelerator manufacturer specializing in AI training and inference processors.

Their products have ranked highly in multiple international AI benchmark evaluations.

Unlike conventional servers, AI servers place extremely demanding requirements on inter-chip communication.

Our R&D team worked closely with the AI chip manufacturer to overcome several major technical challenges.

PCIe Topology Optimization

How should eight AI accelerators connect to the CPU?

Should they use direct connections or PCIe switch chips?

Different topologies directly impact:

• Communication latency

• Bandwidth efficiency

• Workload scalability

Through repeated modeling and testing, our joint engineering teams identified an optimized interconnect architecture that reduced inter-chip communication latency by 15%.

Thermal Solution Customization

AI accelerators generate extremely high power density and localized thermal hotspots.

Our thermal engineering team developed:

• Customized vapor chamber solutions

• Optimized thermal conduction paths

• Precision airflow management

These designs ensure that chip temperatures remain below 85°C even under full-load operation.

Driver and Framework Adaptation

AI hardware performance depends heavily on software ecosystem integration.

We assisted the AI chip manufacturer with:

• TensorFlow adaptation

• PyTorch optimization

• Driver integration

• AI framework tuning

Testing on standard datasets such as ImageNet demonstrated that the inference performance of the complete solution reached over 90% of comparable international mainstream platforms.

From “Usable” to “Production-Ready”

Many industry experts understand that the biggest challenge facing domestic chips is not manufacturing capability, but ecosystem maturity.

A modern server involves:

• Hundreds of hardware components

• Multiple firmware layers

• Complex software stacks

Any incompatibility can become the weakest point in the entire system.

Our value lies precisely in serving as the “ecosystem integrator.”

Through system-level engineering and optimization, we integrate:

• Domestic CPUs and GPUs

• Domestic memory modules

• Domestic SSDs

• Domestic NICs

• Domestic power management ICs

into a complete, deliverable enterprise server platform.

Our mission is not only solving whether components can physically work together — but whether they can operate efficiently, reliably, and stably at scale.

For one of the domestic CPU server platforms validated in this project:

• SPEC CPU 2017 benchmark performance improved by approximately 12% compared to the reference design

• STREAM memory bandwidth utilization exceeded 95% of theoretical performance

These results demonstrate that domestic processors are now capable of large-scale deployment in real-world production environments.

Strategic Positioning in the Era of Domestic Computing Ecosystems

Today, the wave of domestic IT ecosystem development is accelerating rapidly across industries.

From government agencies to finance, energy, transportation, and critical infrastructure sectors, organizations are actively building secure and self-controlled IT supply chains.

As part of this ecosystem, server OEM/ODM manufacturers play an irreplaceable role.

We connect:

• Upstream chip vendors

• Downstream industry users

serving as a critical bridge that enables domestic computing platforms to move from laboratory validation to real-world deployment.

“We do not manufacture chips — but we help chips perform at their best,”

Mr. Chen explained.

“No matter how many domestic architectures emerge in the future, we can rapidly adapt and help customers build the computing platforms best suited to their applications.”

The company has already established strategic partnerships with multiple domestic semiconductor vendors.

In addition to the products already certified, several next-generation server platforms based on upcoming domestic processors are currently under intensive development and are expected to launch later this year.

Customer Perspective: Domestic Platforms Are No Longer Optional

At a recent customer technology exchange event, the IT director of a large state-owned enterprise shared his perspective:

“A few years ago, localization initiatives were mainly driven by policy requirements. At that time, performance and stability still had noticeable gaps.”

He continued:

“But things are very different now. Domestic processors have improved rapidly, and with professional OEM manufacturers like you optimizing the entire system stack, the actual user experience is now very close to international products.”

According to him, the company plans to increase the proportion of domestic server procurement from 10% last year to more than 30% this year.

“This is no longer a multiple-choice question. It has become a mandatory direction for the industry. Fortunately, we now finally have strong and reliable answers.”