G24W RTL9603CVD
Bandwidth, 60 B frames Mbps (cable, L1)
Higher is better: more megabits per second.
Bandwidth, 1400 B frames Mbps (cable, L1)
Higher is better: more megabits per second.
WiFi, 2.4 GHz Mbps (L2)
Higher is better: more megabits per second.
Operating point: 2g/20MHz/HT/CCMP
Signal received by the lab WiFi client
Higher is better: closer to 0 dBm is stronger.
2.4 GHz
| Firmware | This device |
|---|---|
| openwrt | -15.0 dBm |
| stock | -14.0 dBm |
Memory MiB
Lower used memory is better: more RAM is left free.
Snapshot provenance
Physical RAM: 128.0 MiB; source: Recorded bootloader DRAM sizestock firmware: idle snapshot 2026-09-24T20:08:18; image MD5 not recorded; MemTotal 82.2 MiB; outside MemTotal 45.8 MiB
OpenWrt: idle snapshot 2026-10-09T10:49:35; image MD5 9ff739bf9acdf0ad963ad1f642a61a18; MemTotal 111.8 MiB; outside MemTotal 16.2 MiB
CPU clock MHz
Higher maximum is better for CPU-bound work; dynamic scaling spends less at rest.
| CPU clock | stock firmware | OpenWrt |
|---|---|---|
| Dynamic frequency | no | yes (schedutil) |
| Range | fixed 1050 MHz | 525-1050 MHz |
| Hardware range | fixed 1050 MHz | 525-1050 MHz |
| Source | bootloader banner; the kernel exposes no cpufreq (CPU:1050MHz) | cpufreq sysfs (luna); hardware range from cpuinfo |
At rest idle
Lower is better: fewer timer interrupts, a cooler SoC.
| At rest | stock firmware | OpenWrt |
|---|---|---|
| Timer interrupts per second | missing | 221.5 |
Image footprint MiB
Smaller is better: less flash taken by the image.
WiFi stability
Lower is better for loss and delay; higher is better for the download and upload rates.
Alone
| Firmware | Band | Radio settings | Download | Upload | Loss | Mean RTT | p95 |
|---|---|---|---|---|---|---|---|
| OpenWrt | 2.4 GHz | 2g/20MHz/HT/CCMP | 6.00 Mbps | 14.1 Mbps | 7.10% | 1.92 ms | 4.02 ms |
| Stock | 2.4 GHz | 2g/20MHz/HT/CCMP | 104 Mbps | 94.4 Mbps | 0.00% | 7.16 ms | 25.6 ms |
2.4 GHz: FAIL
Another client downloading and uploading at once
| Firmware | Band | Radio settings | Download | Upload | Loss | Mean RTT | p95 |
|---|---|---|---|---|---|---|---|
| OpenWrt | 2.4 GHz | Not captured | |||||
| Stock | 2.4 GHz | Not captured | |||||
Missing data, and why
| Quantity | Firmware | State | Last attempt | Why |
|---|---|---|---|---|
| WiFi latency and loss: WiFi -> WAN, a SECOND WiFi device loading the air (both directions) | stock firmware | MISSING | no attempt recorded yet (no cell stored) | |
| Hardware acceleration, IPv6 | stock firmware | BLOCKED | 2026-11-06 | the root helper could not generate this shape, so nothing was measured (the op exists; it refused): ERR the LAN client holds no routable IPv6 on g24wmv (no prefix is delegated on this rig), so an IPv6 shape cannot be generated AT ALL -- ...the root helper could not generate this shape, so nothing was measured (the op exists; it refused): ERR the LAN client holds no routable IPv6 on g24wmv (no prefix is delegated on this rig), so an IPv6 shape cannot be generated AT ALL -- this must read BLOCKED, never an ONU forwarding failure |
| DHCP LAN->WAN, 60 B frames | OpenWrt | BLOCKED | 2026-09-28 | ours 238.2 Mbps [B@2026-09-28], stock >=321.7 Mbps [B@2026-09-28] (-25.9%): owed a re-measure: stock holds only a LOWER BOUND (capped by unknown) |
| WiFi 2.4 GHz DHCP LAN->WAN, 60 B frames | OpenWrt | FAIL | 2026-11-04 | ours 19.6 Mbps [B@2026-11-04], stock >=28.5 Mbps [B@2026-11-04] (-31.1%): below 70 % of stock |
| WiFi 2.4 GHz DHCP WAN->LAN, 60 B frames | OpenWrt | FAIL | 2026-10-05 | the routed path did NOT forward at this size (short) -- a real small-frame datapath result (the datagrams WERE accounted for and almost none arrived), recorded without a fabricated number: only 181 of 1468793 datagrams arrived (need >= ...the routed path did NOT forward at this size (short) -- a real small-frame datapath result (the datagrams WERE accounted for and almost none arrived), recorded without a fabricated number: only 181 of 1468793 datagrams arrived (need >= 1000) -- the path did not forward at this size |
| WiFi 2.4 GHz DHCP LAN->WAN, 1400 B frames | OpenWrt | BLOCKED | 2026-10-30 | owed a re-measure: stock measured at 6 stream(s), ours at 1 |
| WiFi 2.4 GHz DHCP WAN->LAN, 1400 B frames | OpenWrt | FAIL | 2026-10-30 | ours >=6.0 Mbps [B@2026-10-29], stock >=103.9 Mbps [B@2026-11-04] (-94.2%): below 70 % of stock |
| WiFi latency and loss: WiFi -> WAN, a SECOND WiFi device loading the air (both directions) | OpenWrt | MISSING | no attempt recorded yet (no cell stored) | |
| Hardware acceleration, IPv6 | OpenWrt | BLOCKED | 2026-11-02 | the root helper could not generate this shape, so nothing was measured (the op exists; it refused): ERR the shaped train did not run (icmp6 -> 2001:db8:160::10 size=64): ping: connect: Network is unreachable |
| At rest | stock firmware | BLOCKED | 2026-11-06 | no idle figure could be read: no thermal zone, and no timer among the IRQs (?, IPI_call, IPI_resched, apl_sw, btn,reset, eth0, rtk-gpio-0, rtk-gpio-1, rtk-gpio-2, rtk-tick, serial, wlan0) |
| Time from kernel start until every service is up | OpenWrt | BLOCKED | 2026-11-16 | ours was loaded from 'network': a boot time is measured on the unit's own flash, the way a customer's unit starts |
OLT compatibility
| OLT / firmware | ONU stock | ONU OpenWrt |
|---|---|---|
| HSGQ G008 / HG08 / V2.0.5C | PASS | PENDING |
| HSGQ G01 / V1.2.16C | PENDING | PENDING |
| HSGQ HG16 / V1.3.7C | PASS | PENDING |
| V-SOL V1600G / V2.3.1R.202401251723 | PENDING | PENDING |
| ZTE ZXA10 C610 / V1.2.2B6 patch GC0002 | PENDING | PENDING |
| ZTE Entel | FAIL | PENDING |
| ZTE AXS | FAIL | PENDING |