40G Switching Products
Ultra-low Latency 40G Switching Products
NeoMUX 40G multiplexing and 40G Rate Conversion - built for proprietary trading firms and service providers that value every nanosecond.
644 MHz / 16-bit 40G Ethernet IP core · ultra-low latency MAC/PCS
NeoMUX 40G
40G multiplexing with precise control
The NeoMUX 40G enables 9 client ports to connect to a single 40G exchange link with extremely low latency of 52ns. Exchange-to-client operates in Layer 1 mode (2.5ns, FPGA bypass) or ACL mode (per-recipient filtering) configurable per port.
Client-to-exchange Layer 152ns
Exchange-to-client Layer 12.5ns
40G Rate Converter
Lowest possible latency between 40G and 10G
The 40G Rate Converter delivers 50ns conversion between 40G exchange lines and 10G client infrastructure. Unlike regular switches, it uses smart pre-buffering techniques on client-to-exchange direction to avoid full store-and-forward latency. Exchange-to-client direction operates in Broadcast mode (delivering all exchange data to every 10G client port) or ACL mode (per-recipient filtering).
Exchange-to-client Broadcast50ns
10G → 40GSmart pre-buffering
Access Control Layer - Available on Both Products
ACL Features
Hardware-enforced per-port filtering
Each client port has its own independent ACL ruleset applied in hardware at line rate - no software bottleneck, no shared state between clients.
Line rate filtering — zero throughput penalty
ACL decisions are made at full 40G line rate. Enabling ACL mode does not reduce available bandwidth or increase jitter on unfiltered traffic.
72 ACL entries per client port, with wildcards
Each port supports up to 72 independent rules. CIDR wildcards are supported — a subnet such as 233.74.125.0/24 counts as a single entry, not 256.
NeoMUX 40G
Exchange-to-Client Modes
The NeoMUX 40G connects 9 client ports to a single 40G exchange link. The exchange-to-client path is configured per client port in either Layer 1 or ACL mode.
Layer 1 mode
exchange → clientInbound exchange data bypasses the FPGA entirely and is broadcast to all client ports simultaneously - achieving 2.5ns exchange-to-client latency. Ideal for proprietary trading firms where all ports belong to the same customer and data leaks are not a concern.
ACL mode
exchange → clientExchange data passes through per-port ACL filtering before delivery, ensuring each client port only receives data it is entitled to. Essential for service providers connecting multiple independent clients to the same exchange link on shared infrastructure.
40G Rate Converter
Broadcast or ACL
The 40G Rate Converter converts between 40G exchange lines and 10G client infrastructure in two modes. Both share the same ultra-low latency core.
Muxing 10G → 40G with smart pre-buffering
exchange → clientAggregates multiple 10G upstream streams into a single 40G exchange line. Analyzes latency-critical protocols like IPv4 and intelligently releases data to minimize serialization delay, further reducing the latency compared to the conventional store-and-forward conversion.
Broadcast mode
exchange → clientAll exchange data is delivered to every 10G client port simultaneously — no filtering, no FPGA decision logic on the inbound path. Achieves 50ns exchange-to-client latency. Suited for proprietary trading desks where all 10G ports belong to the same recipient.
ACL mode
exchange → clientExchange data is filtered per 10G client port before delivery, ensuring each port only receives data it is authorized to see. Also reduces the total amount of data when actual bandwidth on the 40G link significantly exceeds 10G, allowing 10G clients to receive their data with no drops. Suited for service providers distributing a single 40G exchange feed to multiple independent 10G clients on shared infrastructure.
No packet loss
exchange → clientLarge buffers absorb 40G traffic bursts, ensuring zero packet loss and reliable network performance even during high-traffic periods.
40G Rate Converter — Latency Summary
| Mode | Direction | Latency | Notes |
|---|---|---|---|
Broadcast/ACL | Client-to-exchange | 53ns | Variable, depends on packet size |
Broadcast | Exchange-to-Client | 50ns | - |
ACL | Exchange-to-Client | 64/69ns | Multicast/MAC · Unicast |
BOTH PRODUCTS
Built-in platform features
These capabilities are standard on both the NeoMUX 40G and the 40G Rate Converter, regardless of operating mode.
Out-of-band Kill Switch
Hardware-level port disconnection independent of the data path. Immediately isolates any client port without affecting others on the same device — operable even when the data link is down.
Real-time Performance Monitoring
Per-port network statistics updated in real time. Detect packet loss, errors, and degradation before they impact trading operations. Full SNMP support for integration with existing network management infrastructure.
40G Lane Alignment Assistant
40G is a combination of 4 lanes working at 10G. Assistant visualizes potential alignment delays between lanes, helping to mitigate hidden latency problems.
40G RATE CONVERTER
Full 40G specifications
Select a product and mode below.
| Specification | Value |
|---|---|
Client-to-Exchange Latency | 52ns |
Exchange-to-Client Latency — Multicast/MAC | 57ns |
Exchange-to-Client Delivery — Unicast | 64ns |
Exchange-to-Client Delivery | Per-port ACL filtering — each client receives only its authorized data |
ACL Filtering Bandwidth | Line rate |
Number of ACLs | 72 per client port, with wildcards (e.g. 233.74.125.0/24 = 1 entry) |
Topology | 1 exchange port — 9 × 40G client ports |
Storm Control Protection | Per client port |
Out-of-Band Kill Switch | Yes |
| Specification | Value |
|---|---|
Client-to-Exchange Latency | 53ns |
Exchange-to-Client Latency — Multicast/MAC | 57ns |
Exchange-to-Client Latency — Unicast | 64ns |
Exchange-to-Client delivery | Per-port ACL filtering — each 10G client port receives only its authorised data |
Client-to-Exchange mode | Muxing with smart pre-buffering (IPv4 protocol-aware) |
ACL Filtering Bandwidth | Line rate |
Number of ACLs | 72 per client port, with wildcards (e.g. 233.74.125.0/24 = 1 entry) |
Exchange Port | 1 × 40G |
Fiber Lane Alignment | Fiber Lane Alignment Assistant |
Latency Monitoring | Client 40G port skew and latency monitor |
Port Statistics | All Ports, Real-Time |
Management | SNMP |
Latency of the 16-port option is 3ns higher than the 8-port option.