G24W RTL9603CVD

Stock firmware against OpenWrt on the SAME unit of this model. Every pair charted below was measured with the same instrument, at the same operating point -- that is the condition for appearing on a chart at all. A pair whose two halves were taken with different instruments, or at operating points that cannot be held together, is not charted: each figure keeps its qualifiers in the dated artifact these pages are generated from, and the certificate publishes the two of them as separate, individually attributed rows rather than as a difference. Only paired figures appear on this comparison page. Missing or incomparable halves are named without publishing standalone values; the dated artifacts and certificate retain their measurements.

← all ported devices

Two frame sizes are published here, 60 B and 1400 B, and both as bandwidth on the cable: megabits per second counted the way the wire carries them, each frame plus its checksum, preamble and inter-frame gap (L1). Measured the same way at both sizes, every wired bar reads against the same gigabit link, so a 60 B bar and a 1400 B bar can be compared directly. Whether a given bar is AT the link's capacity or far below it is what the bar itself says, and the caption claims nothing about which is which. WiFi figures have no cable framing and are counted at the frame itself (L2). Every size the suite measures, and every qualifier each figure was measured with, stays in the dated artifact these pages are generated from.

Bandwidth, 60 B frames Mbps (cable, L1)

6 of these 12 points are drawn but NOT compared: stock's figure is a LOWER BOUND -- it is bounded by the unknown, not by the device -- so it establishes a floor and not a magnitude; stock's figure is a LOWER BOUND -- it is bounded by the generator, not by the device -- so it establishes a floor and not a magnitude. Both figures stand exactly as they were measured; what is withheld is their difference.

11 of the drawn bars are LOWER BOUNDS, a floor and not a magnitude: stock firmware DHCP LAN->WAN (322 Mbps, bounded by the unknown, not by the device); stock firmware DHCP WAN->LAN (408 Mbps, bounded by the unknown, not by the device); OpenWrt DHCP WAN->LAN (460 Mbps, bounded by the unknown, not by the device); stock firmware PPPoE CHAP WAN->LAN (164 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE CHAP WAN->LAN (134 Mbps, bounded by the generator, not by the device); stock firmware PPPoE CHAP+VLAN WAN->LAN (88.2 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE CHAP+VLAN WAN->LAN (165 Mbps, bounded by the generator, not by the device); stock firmware PPPoE PAP WAN->LAN (102 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE PAP WAN->LAN (118 Mbps, bounded by the generator, not by the device); stock firmware PPPoE PAP+VLAN WAN->LAN (132 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE PAP+VLAN WAN->LAN (71.1 Mbps, bounded by the generator, not by the device). The figure stands as measured; nothing is subtracted from it.

Bandwidth, 1400 B frames Mbps (cable, L1)

9 of these 12 points are drawn but NOT compared: stock's figure is a LOWER BOUND -- its own producer recorded it as one -- so it establishes a floor and not a magnitude; stock's figure is a LOWER BOUND -- it is bounded by the generator, not by the device -- so it establishes a floor and not a magnitude; OpenWrt's figure is a LOWER BOUND -- its own producer recorded it as one -- so it establishes a floor and not a magnitude. Both figures stand exactly as they were measured; what is withheld is their difference.

15 of the drawn bars are LOWER BOUNDS, a floor and not a magnitude: stock firmware DHCP LAN->WAN (999 Mbps, its own producer recorded it as a lower bound); OpenWrt DHCP LAN->WAN (999 Mbps, its own producer recorded it as a lower bound); stock firmware DHCP WAN->LAN (1000 Mbps, its own producer recorded it as a lower bound); stock firmware DHCP+VLAN LAN->WAN (997 Mbps, its own producer recorded it as a lower bound); OpenWrt DHCP+VLAN LAN->WAN (994 Mbps, its own producer recorded it as a lower bound); stock firmware DHCP+VLAN WAN->LAN (998 Mbps, its own producer recorded it as a lower bound); stock firmware PPPoE CHAP WAN->LAN (332 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE CHAP WAN->LAN (408 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE CHAP+VLAN LAN->WAN (999 Mbps, its own producer recorded it as a lower bound); stock firmware PPPoE CHAP+VLAN WAN->LAN (606 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE CHAP+VLAN WAN->LAN (645 Mbps, bounded by the generator, not by the device); stock firmware PPPoE PAP WAN->LAN (377 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE PAP WAN->LAN (513 Mbps, bounded by the generator, not by the device); stock firmware PPPoE PAP+VLAN WAN->LAN (258 Mbps, bounded by the generator, not by the device); OpenWrt PPPoE PAP+VLAN WAN->LAN (199 Mbps, bounded by the generator, not by the device). The figure stands as measured; nothing is subtracted from it.

WiFi, 2.4 GHz Mbps (L2)

Bars on this band: stock firmware, OpenWrt. Operating points differ across these bars: 2g/20MHz/HT/CCMP, 2g/40MHz/HT/CCMP. No difference is claimed where only one firmware has a figure.

1 of these 1 points are drawn but NOT compared: stock's figure is a LOWER BOUND -- it is bounded by the air, not by the device -- so it establishes a floor and not a magnitude. Both figures stand exactly as they were measured; what is withheld is their difference.

2 of the drawn bars are LOWER BOUNDS, a floor and not a magnitude: stock firmware WiFi -> WAN -- 1400 B (102 Mbps, bounded by the air, not by the device); OpenWrt WiFi -> WAN -- 1400 B (50.8 Mbps, bounded by the air, not by the device). The figure stands as measured; nothing is subtracted from it.

No comparable stock/OpenWrt pair is available for: WAN -> WiFi -- 1400 B. Standalone values remain in the dated artifacts.

Radio transmit power

Transmit power reported by the device at its operating point, against the ceiling the regulator permits there. One row per operating point, so the two firmwares are read side by side; the SSID beside each figure says which radio produced it. Where the two firmwares run different channel widths (each one's own default), the row names both and each figure says its width: a declared difference, not a like-for-like pair.

Memory MiB

Read from the running device's own kernel, in the two states the suite captures: Available is what a new workload can take (MemAvailable), Free is MemFree, Given is what the kernel was handed (MemTotal). The two firmwares run different kernels and are handed different amounts of RAM, so the total each was given is drawn beside the rest and no difference is claimed from these figures.

Image footprint MiB

Weighed with the board unplugged -- a property of the firmware, not a reading taken from a running device. Compressed is the root filesystem image as each firmware ships it (ours squashed from the staged tree with the image's own options, the vendor's read from its flash bank); uncompressed is the bytes of regular files in the extracted tree, the same rule on both.

No comparable stock/OpenWrt pair is available for: Installed, by package; Web UI closure, installed; Packages installed; Files belonging to the web UI. Standalone values remain in the dated artifacts.

Not published on this page, and why

Everything this page can carry is listed on it. What is missing is NAMED here rather than left out: an absent figure is either a measurement we have not taken yet, or a part this device does not have -- it is never a result about the device, and never a zero.