Skip to content

Hardware Acceleration

Note: This guide applies to firmware v4.2 and earlier. For newer versions, please refer to Network Acceleration.

Hardware acceleration (sometimes called hardware NAT, flow offloading, or offloading) reduces CPU load by moving packet forwarding from the CPU to the router's SoC/NIC hardware. This typically increases maximum throughput and lowers CPU utilization, but it has important trade-offs, especially for features that rely on the Linux networking stack (netfilter/iptables/nftables) or the kernel queuing disciplines (qdisc) used by SQM (Smart Queue Management).

When hardware acceleration is enabled, the following functions will not work properly: Client Speed and Traffic Statistics, Client Speed Limit.

Supported Models

Supported Models
  • GL-E5800 (Mudi 7)
  • GL-MT5000 (Brume 3)
  • GL-MT3600BE (Beryl 7)
  • GL-BE6500 (Flint 3e)
  • GL-BE9300 (Flint 3)
  • GL-BE3600 (Slate 7)
  • GL-X2000 (Spitz Plus)
  • GL-B3000 (Marble)
  • GL-MT6000 (Flint2)
  • GL-X3000 (Spitz AX)
  • GL-XE3000 (Puli AX)
  • GL-MT3000 (Beryl AX)
  • GL-MT2500/GL-MT2500A (Brume 2)
  • GL-SFT1200 (Opal)
  • GL-MT1300 (Beryl)
Unsupported Models
  • GL-AXT1800 (Slate AX)
  • GL-AX1800 (Flint)
  • GL-A1300 (Slate Plus)
  • GL-E750/E750V2 (Mudi)
  • GL-X750/GL-X750V2 (Spitz)
  • GL-AR750S (Slate)
  • GL-XE300 (Puli)
  • GL-MT300N-V2 (Mango)
  • GL-AR300M Series (Shadow)
  • GL-B1300 (Convexa-B)
  • GL-X300B (Collie)

Quick Setup

On the left side of the web Admin Panel, go to NETWORK -> Hardware Acceleration.

Hardware Acceleration

Toggle the switch to enable, and click Apply.

Hardware NAT vs. Software NAT

  • You care most about throughput (e.g., multi‑gigabit broadband) and don't need on‑router SQM or per‑client shaping → enable Hardware NAT / Network Acceleration. This delivers the highest throughput and lowest CPU usage.

  • You care about low latency, consistent QoS, per‑client limits, or you rely on SQM (cake/fq_codel) → use Software NAT (disable hardware offload). SQM and QoS require packets to traverse the kernel qdisc stack — offloaded packets bypass this path and are therefore not shaped.

Still have questions? Visit our Community Forum or Contact us.