Device performance under OpenWrt
The same measurement across every ported device, so one board can be compared with another on identical firmware.
Two frame sizes are published here. At 1400 B the gigabit cable itself runs out first -- 1424 B of wire time per frame caps any device on the link at about 87.8 thousand frames per second -- so the packet count at that size measures the WIRE, and these points are shown as bandwidth. At 60 B the link is far from full and the device is the limit, so those points stay in packets per second. The unit is chosen per chart and printed beside every figure. Every size the suite measures, and every qualifier each figure was measured with, stays in the dated artifact these pages are generated from.
DHCP LAN->WAN, 60 B frames
DHCP WAN->LAN, 60 B frames
DHCP+VLAN LAN->WAN, 60 B frames
DHCP+VLAN WAN->LAN, 60 B frames
PPPoE CHAP LAN->WAN, 60 B frames
PPPoE CHAP WAN->LAN, 60 B frames
PPPoE CHAP+VLAN LAN->WAN, 60 B frames
PPPoE CHAP+VLAN WAN->LAN, 60 B frames
PPPoE PAP LAN->WAN, 60 B frames
PPPoE PAP WAN->LAN, 60 B frames
PPPoE PAP+VLAN LAN->WAN, 60 B frames
PPPoE PAP+VLAN WAN->LAN, 60 B frames
DHCP LAN->WAN, 1400 B frames
DHCP WAN->LAN, 1400 B frames
DHCP+VLAN LAN->WAN, 1400 B frames
DHCP+VLAN WAN->LAN, 1400 B frames
PPPoE CHAP LAN->WAN, 1400 B frames
PPPoE CHAP WAN->LAN, 1400 B frames
PPPoE CHAP+VLAN LAN->WAN, 1400 B frames
PPPoE CHAP+VLAN WAN->LAN, 1400 B frames
PPPoE PAP LAN->WAN, 1400 B frames
PPPoE PAP WAN->LAN, 1400 B frames
PPPoE PAP+VLAN LAN->WAN, 1400 B frames
PPPoE PAP+VLAN WAN->LAN, 1400 B frames
WiFi
Every device is charted at each operating point it was measured at, so a newer radio shows the advantage it actually delivers. Each bar names its own operating point -- band, width, standard and cipher; the channel is recorded per measurement in the dated artifact, and is left off the bar because at one band, width, standard and cipher two channels negotiate the same thing -- beside its figure, because bars on one chart may have been taken at different points; where they differ, the gap between them includes that difference and no percentage is claimed. These are end-to-end figures over the device's own WAN, not an isolated radio measurement, and the client, the channel and the air are in them too. A device with no WiFi does not appear here at all.
WiFi -> WAN -- DHCP, 1400 B frames
WiFi -> WAN -- DHCP, 1400 B frames
WAN -> WiFi -- DHCP, 1400 B frames
WAN -> WiFi -- DHCP, 1400 B frames
WiFi -> WAN -- TCP max
WAN -> WiFi -- TCP max
WiFi latency
Round-trip delay over the air, at the 95th percentile -- the tail a user actually notices, not the average. Lower is better on these charts, the opposite of every chart above, so the bars are drawn in a different colour and each states the direction. Each bar names its own operating point. The idle chart is the floor; the loaded ones show what happens to it while another device is using the same radio, which is the case a shared medium makes worst.
WiFi -> WAN, idle
7 WiFi latency point(s) on X111W, X400AXF carry no figure that can be placed beside another device's: the point is BLOCKED on this device -- nothing was measured, so there is no figure to place on an axis
Among the WiFi points published here, 44 further WiFi point(s) on X111W, X400AXF carry no figure that can be placed beside another device's: the cell CONTRADICTS ITSELF as a maximum: it delivered 0.2% of the rate its own radio negotiated -- below the 5.0% floor -- and records limited_by='unknown', i.e. its own producer says nothing in evidence explains the figure. A maximum nobody can account for is not a capability; it is a failure with a number attached. It stays in the dated artifact with every field it was measured with.; the cell CONTRADICTS ITSELF as a maximum: it delivered 1.9% of the rate its own radio negotiated -- below the 5.0% floor -- and records limited_by='unknown', i.e. its own producer says nothing in evidence explains the figure. A maximum nobody can account for is not a capability; it is a failure with a number attached. It stays in the dated artifact with every field it was measured with.; the point is BLOCKED on this device -- nothing was measured, so there is no figure to place on an axis