Home›Industry News›Jingxin Micro JXW8748 Enters Mass Production in Leyan Gateway, Marking First Domestic PCIe Switch in Enterprise-Grade Systems
← Back to Industry News
Industry

Jingxin Micro JXW8748 Enters Mass Production in Leyan Gateway, Marking First Domestic PCIe Switch in Enterprise-Grade Systems

IndustryJingxin Micro JXW8748Enters Mass Productionin Leyan Gateway,…fpga.love
Illustration

A domestic PCIe switch chip has been designed into a mass-production model by an enterprise gateway vendor for the first time, rather than remaining on samples and evaluation boards. Leyan Technology's new-generation gateway uses Jingxin Micro's JXW8748 as the sole high-speed interconnect core of the system, with the upstream directly connected to the main control CPU and the downstream connecting network cards and FPGA acceleration cards. The company says burn-in validation has been completed and volume deliveries are underway.

The interconnect bottleneck in gateway architecture

Leyan Technology builds enterprise-grade high-end gateways for border protection, data center traffic forwarding, compute access, and industrial network aggregation. The new-generation product uses a domestic multi-core main control CPU, and the system must connect multiple high-speed network cards and FPGA acceleration cards, while the CPU's native PCIe lane count is limited and cannot directly support all of them simultaneously. Using discrete expansion devices would mean complex routing, a bloated system, and higher latency, bandwidth contention, BOM cost, and thermal pressure. Leyan says that after multiple rounds of solution comparison, performance testing, and reliability burn-in, it ultimately selected the JXW8748 as the core I/O hub. This shows that when system vendors choose components, lane count and topology simplicity now outweigh path dependency on overseas devices.

Single-chip centralized topology

In the new architecture, the JXW8748 connects upstream directly to the main control CPU and serves as the sole high-speed interconnect core handling data exchange and bandwidth distribution between the CPU and all downstream peripherals; downstream it fans out multiple PCIe lanes to connect separately to Xinchuang high-speed network chips and domestic FPGA acceleration cards. The chip integrates traffic arbitration, dynamic QoS bandwidth scheduling, link fault detection, lane retraining, and hot-plug protection, and can allocate bandwidth according to load and isolate traffic from different peripherals. Jingxin Micro also provides low-level drivers, configuration tools, and fault diagnosis support. The official claim is that a single chip replaces multiple discrete expansion devices, reducing peripheral components, simplifying PCB routing and layer count, and lowering BOM cost and thermal pressure. For board engineers, this means high-speed interconnect is shifting from "stacking devices" to "choosing one switch chip to define the topology."

EventImpactWatch
Illustration

Production deployment and validation boundaries

Leyan says the gateway has completed full-function testing and rigorous burn-in validation, achieved mass production and market delivery, and passed full-load, high/low-temperature, and extreme-condition tests with stable links, no packet loss, and no disconnections. Jingxin Micro says the JXW8748 targets network equipment, compute hardware, and high-performance embedded devices, and that this commercial deployment validates the maturity of domestic PCIe switch chips in high-end enterprise network equipment. The real issue is that the material only provides qualitative statements such as "significantly improved throughput" and "continuously optimized latency," without port count, lane configuration, PCIe generation, or measured latency figures, and no third-party retesting has been published. The judgment that this is a domestic substitution can hold, but the conclusion that performance matches overseas equivalents cannot yet hold.

Three takeaways
  • JXW8748 serves as the sole high-speed interconnect core in Leyan's gateway, upstream to the CPU and downstream to network cards and FPGA cards
  • The company says mass production and delivery have been achieved, with full-load and high/low-temperature burn-in completed, no packet loss and no disconnections
  • Port count, PCIe generation, and measured latency have not been disclosed; performance benchmarking against overseas products still lacks third-party data

What it means for us

For teams working on boards and high-speed interfaces, the practical value of this news lies in the topology reference: using one PCIe Switch to replace multiple discrete expansion devices and fan out the CPU's limited lanes to multiple network cards and FPGA cards, with routing and thermal pressure reduced accordingly. If a project is evaluating domestic PCIe Switches, the JXW8748 has entered system mass production and can be included as a candidate, but before selection, vendors should be asked for specific specifications on port configuration, PCIe generation, hot-plug, and fault tolerance, and latency and concurrent bandwidth should be independently retested. On the toolchain side, Jingxin Micro provides drivers, configuration tools, and diagnostic support; migration cost depends on how much the existing firmware relies on non-transparent bridging and traffic isolation. The supply chain now has one more domestic option, but a single system case is insufficient to judge long-term supply and batch consistency.

PCIe Switchenterprise gatewaydomestic substitutionFPGA acceleration card
fpga.love

Subscribe to fpga.love

Daily Briefing · Weekly Digest · Monthly Deep Dive