Silicon Labs Proves Matter-Over-Thread Can Scale to 200 Nodes in Real Boston Office Test
Silicon Labs has deployed a 200-node Matter-over-Thread validation network across its Boston corporate campus, demonstrating that the smart-building standard can handle real enterprise-scale deployments without the artificial conditions of a laboratory. The test, announced at the Connectivity Standards Alliance's inaugural Unify event, pushes Matter from a consumer smart-home protocol toward a platform capable of supporting the dense sensor fabrics required by commercial buildings, factories, and multi-dwelling housing.
Real-World Mesh Under Office Chaos
Rather than simulating deployment in a controlled lab, Silicon Labs engineers distributed 40 connectivity clusters of six Wireless Starter Kits each — 240 devices minus four that failed commissioning, netting 200 active nodes — throughout its Boston office and adjacent connectivity lab. The site presented the kind of radio-frequency chaos that laboratory tests never capture: twelve separate Wi-Fi networks, dozens of active Bluetooth peripherals, and a steady stream of employee traffic all sharing the same 2.4 GHz and 5 GHz spectrum that Thread relies on.
Every node was configured as a Full Thread Device and a router simultaneously, a deliberate choice to stress-test the network's routing-table capacity and background Maintenance-SOLIC traffic handling. The hardware under test was BRD4187C development boards powered by EFR32MG24 wireless system-on-chips, running FreeRTOS with thirteen concurrent OS threads managing crypto operations, message parsing, and neighbor-table maintenance.
According to Daniel Cooley, CTO at Silicon Labs, the validation matters because enterprises evaluating Matter have been waiting for proof that it works beyond the living room. "Matter is rapidly evolving from a smart home technology into a platform capable of supporting much larger deployments," he said. "This work demonstrates not only that Matter-over-Thread can scale to thousands of devices, but also how organizations are learning to deploy, manage, and future-proof those networks through innovations spanning Matter, Thread, and Concurrent Multiprotocol technologies."
The 50-metre maximum span between the furthest nodes forced the mesh to rely on multi-hop forwarding for the last clusters installed in a corner hallway, a structural zone with only a single line-of-sight to the main office floor. Concrete columns and perpendicular walls further fragmented the RF topology, creating the kind of physical constraints that building owners cannot redesign around.
Latency Grows Where Concrete Blocks Signals
Mean application-layer latency climbed predictably as node count increased. For 8-byte payloads, average round-trip time rose from 62 milliseconds at one hop to 341 milliseconds across seven hops in controlled unicast testing. The 95th-percentile figures were more telling: 320 milliseconds at eight bytes and 340 milliseconds at 64 bytes in the full 200-node deployment, reflecting the tail-end nodes trapped behind the hallway bottleneck.
The performance spike was not a protocol failure but a topology problem. Automated commissioning scripts brought devices online in network-address order, meaning the final 50 nodes — all physically concentrated in the problematic hallway — arrived last. That single corridor entry point became a congestion chokepoint, forcing multi-hop forwarding and packet flooding to reach every isolated node.
Despite this localised routing stress, the network sustained continuous operation for a three-hour stretch test with less than one percent packet loss across all standard payload sizes. The engineering team's controlled laboratory follow-up — using four-hop to seven-hop topologies inside RF-shielded Ramsey STE330 isolation boxes linked by coaxial cable — confirmed zero packet loss at every hop count, with session-setup latency scaling linearly from 65 milliseconds at one hop to 352 milliseconds at seven hops.
Engineering Past the Commissioning Bottleneck
Commissioning 200 devices proved to be its own engineering challenge. The team's initial strategy relied on Bluetooth Low Energy for individual device pairing, a standard Matter commissioning path. But BLE's limited range and single-device-at-a-time throughput caused automated setup scripts to fail before completing fabric registration across the full fleet.
"We learned that BLE-based commissioning doesn't scale to bulk deployments," explained a Silicon Labs engineer involved in the test. "When you're standing up hundreds of sensors at once, you can't walk a phone past each one."
The fix was an on-network commissioning workflow that bypasses BLE proximity requirements entirely. After devices attach to the Thread mesh using a network dataset distributed through the OpenThread Border Router command-line interface, each node registers a Service Registration Protocol (SRP) instance. The deployment orchestrator then issues on-network commissioning commands over the established Thread link, completing fabric registration securely without physical proximity.
The remediation raised commissioning success to 100 percent across multiple test iterations, and reduced per-node setup time to an average of seven seconds — fast enough that a building with thousands of sensors could be commissioned overnight rather than over days of manual pairing.
What This Means for Building Operators
Commercial real estate portfolio managers and facilities teams have been waiting for a protocol that can handle the density of a true smart-building deployment. A typical Class-A office tower needs temperature, humidity, occupancy, light-level, and air-quality sensors on every floor, plus smart lighting controls, HVAC actuators, and access points — easily reaching into the hundreds of endpoints per building.
Before Silicon Labs' test, Matter-over-Thread had only been proven at residential scale. The 200-node validation shows the standard can handle a mid-sized office building's sensor load while maintaining sub-350-millisecond latency for control-loop traffic — fast enough for lighting and HVAC automation, where response delays above 500 milliseconds start to generate user complaints.
The silicon architecture also matters for battery life. The EFR32MG24 chips used in the test support concurrent multiprotocol operation, letting them listen on Thread while maintaining a BLE connection for mobile-app fallback. That matters for retrofitting older buildings where replacing every light switch with a Thread-native model is cost-prohibitive; the dual-protocol nodes let occupants keep their existing Bluetooth remotes while the backbone runs on Matter-over-Thread.
Industry analysts see the test as a milestone. "Everyone knew Matter could work in theory at scale, but theory and a Boston office full of Wi-Fi interference and concrete are different things," noted one smart-building consultant. "This gives building owners the confidence to specify Matter-over-Thread for new construction rather than waiting for the next spec revision."
The Connectivity Standards Alliance is incorporating Silicon Labs' topology data into the Thread 1.4 certification suite, which will make enterprise-scale validation mandatory for new certifications. That could raise the bar for smart-building hardware manufacturers just as commercial construction starts its five-year modernization cycle.
For more on commercial IoT standards and deployment strategies, see our IoT category coverage and the Semiconductors category for chip-level developments.
Looking Beyond the Boston Test
Silicon Labs is already planning a larger deployment — targeting 500 nodes across multiple floors of its Austin, Texas facility — to test the upper boundary of what Matter-over-Thread can handle when network datasets are distributed across a multi-story building with mixed construction materials. Early results from ongoing Thread 1.4 certification testing suggest the protocol can support up to 2,000 nodes per mesh partition, but no public deployment has yet crossed the 250-node threshold.
The Boston test also surfaced a practical lesson for facilities managers: commissioning speed matters more than raw node count. At seven seconds per node, a 500-node building takes roughly an hour to commission once the network is physically installed. That's within the window of a standard building turnover — a critical threshold for contractors who bid smart-building retrofits with fixed labor schedules.
Full technical documentation from Silicon Labs' 200-node validation, including the per-hop latency tables and multi-hop topology maps, is available through the Connectivity Standards Alliance's Unify event proceedings. For the original announcement and engineering details, see Silicon Labs' 200-node Matter-over-Thread validation deployment.