RU-NS520 / Spine
Fabric Network SoC
High-density east-west switching for Spine fabrics and large data center systems.
RILON UNITS / PRODUCTS
RU-NS Network SoC family
Three RILON-owned Network SoC models for high-density Spine fabrics, large Leaf aggregation and DCI-style deep-buffer interconnects.
RILON product familyOverseas contracting lineup
01 / Lineup
One silicon family. Three network roles.
Each model is defined around a system role so network equipment teams can start a focused architecture and supply conversation.
RU-NS420 / Large Leaf
Aggregation Network SoC
Large Leaf and aggregation silicon for L3/L4 data center and commercial backbone equipment.
RU-NS820 / DCI
Deep-Buffer Network SoC
DCI-style interconnect silicon for hub and sub-DC traffic with 800G / 1.6T class DCI.
02 / Family profile
Network buffering at the center of the fabric.
The RU-NS family combines high-speed network interfaces, a switching fabric and large on-die embedded SRAM for packet buffering.
Embedded SRAM is designed for network congestion and micro-burst behavior. RU-NS is network infrastructure silicon, not an AI compute accelerator.
03 / Shared family specifications
RILON class language for contracting conversations.
The table describes the public RU-NS family profile. Detailed implementation documents are shared through the appropriate engagement channel.
| Parameter | RU-NS520 | RU-NS420 | RU-NS820 |
|---|---|---|---|
| Primary role | Spine fabric | Large Leaf / aggregation | DCI / deep-buffer interconnect |
| Process and die | 4nm-class · >600 mm² class | 4nm-class · >600 mm² class | 4nm-class · >600 mm² class |
| Embedded SRAM | Large on-die SRAM for network buffering | Large on-die SRAM for network buffering | Large on-die SRAM for deep buffering |
| SerDes and switching | High SerDes · tens-of-Tbps class | High SerDes · multi-Tbps class | High SerDes · multi-Tbps class |
| Network profile | East-west fabric · RoCE-ready | L3/L4 · east-west fabric | L3/L4 backbone · DCI transport |
| Interconnect profile | Spine fabric links | Leaf and aggregation links | 800G / 1.6T class DCI |
The published profile uses RILON class language and does not reproduce another vendor's SKU or port-count table.
04 / Architecture
Leaf–Spine–DCI role model
The diagram shows how the RU-NS family maps into a network system and traffic flow.
The RU-NS family maps from high-density Leaf and Spine traffic into DCI transport roles.
05 / Design priorities
Role differences, clearly stated.
Deep-buffer routing, high-density switching and programmable Ethernet are different design priorities. RILON keeps the RU-NS public story focused on fabric, buffer and supply readiness.
Deep-buffer routing
Long-path congestion
Prioritizes congestion absorption and transport behavior for DCI-style systems.
High-density switching
East-west fabric
Prioritizes throughput and low-latency Spine fabrics for data center traffic.
Programmable Ethernet
Control and flexibility
Prioritizes deployment control; RU-NS keeps the current family focus on fabric and buffering.
06 / Product conversation
Bring the traffic model.
RU-NS model selection starts with topology, traffic shape, buffer behavior and qualification gates.