NOKIA BEACON

What should my Wi-Fi speed be?

on , , , , , .These are starting guesses. Tap any to match your house.

2.4 GHz: 195 Mbps, the most 2.4 GHz can do at 20 MHz on Wi-Fi 6, 5 GHz (Auto usually picks this): 20 MHz, 161 Mbps – 224 Mbps; 40 MHz, 254 Mbps – 423 Mbps; 80 MHz, 430 Mbps – 796 Mbps.

You should see about

Effective width 20 MHz on 2.4 GHz, 20, 40 or 80 MHz on 5 GHz · device max 20 MHz on 2.4 GHz, 80 MHz on 5 GHz

Anywhere from

161–796 Mbps

  1. 2.4 GHz

    ASSUMPTION: Tx backoff 0–3 dB at 1024-QAM and above, not given by any filing. ASSUMPTION: 1024-QAM and above means MCS 10 and up.

    195 Mbps median, the most 2.4 GHz can do at 20 MHz on Wi-Fi 6

    2.4 GHz 20 MHz: median 195 Mbps; 8 in 10 runs 195 Mbps, middle half 195 Mbps; max rate in 92% of runsmax rate in 92% of runs195125150175200225
  2. 5 GHz · Auto usually picks this

    ASSUMPTION: Tx backoff 0–3 dB at 1024-QAM and above, not given by any filing. ASSUMPTION: 1024-QAM and above means MCS 10 and up.

    20 MHz:

    224 Mbps median

    5 GHz 20 MHz: median 224 Mbps; 8 in 10 runs 161–224 Mbps, middle half 179–224 Mbps; max rate in 60% of runsmax rate in 60% of runs161179224150175200225250

    40 MHz:

    381 Mbps median

    5 GHz 40 MHz: median 381 Mbps; 8 in 10 runs 254–423 Mbps, middle half 339–423 Mbps; max rate in 44% of runsmax rate in 44% of runs254339381423200250300350400450

    80 MHz:

    637 Mbps median

    5 GHz 80 MHz: median 637 Mbps; 8 in 10 runs 430–796 Mbps, middle half 478–796 Mbps; max rate in 28% of runsmax rate in 28% of runs430478637796400500600700800900

thin line = 8 in 10 runs, thick bar = middle half, white tick = median

Mbps. Each row has its own scale; compare the numbers.

On Auto width the Beacon may use any of the listed widths, and this app can't tell which. Left out because this device can't use them: 40 MHz on 2.4 GHz; 160 MHz on 5 GHz.

Your phone's bars depend on the band it picked; distance, walls and floors set these speeds.

Plug a computer into one of the router's LAN ports and a speed test should read about 1 Gbps. That is the wired limit, not the Wi-Fi.

Show the math

One device, two spatial streams, one band at a time. AX3000 sums a 574 Mbps 2.4 GHz radio and a 2,402 Mbps 5 GHz radio. That headline is not what one device gets.

Each band gets its own link budget. The Beacon's transmit power (EIRP, from its FCC filing), plus a hand-held device antenna gain (below 0 dBi), minus path loss over the straight-line distance (the IEEE 802.11ax TGax residential model: free-space slope 2 up to a 5 m breakpoint, slope 3.5 after it, the same at every band), minus 5 dB per interior wall and 10 dB per exterior wall at every band, and per floor 5 dB (2.4 GHz) or 7 dB (5 and 6 GHz) for wood, or 18.3 dB for one concrete floor and 33.5 dB for two, is the signal at the device. Subtracting the noise floor (thermal noise for the channel width plus a 10 dB noise figure) gives the SNR, and the SNR picks the highest MCS the IEEE 802.11 sensitivity table allows.

Speed is the PHY rate at that MCS × the measured TCP efficiency for the device class and channel width (phone 0.85 at 20 MHz, 0.82 at 40 MHz, 0.78 at 80 MHz, 0.68 at 160 MHz; laptop or tablet 0.68 at 20 MHz, 0.55 at 40 MHz, 0.35 at 80 MHz, 0.21 at 160 MHz; 320 MHz is assumed equal to 160 MHz) × a generation factor relative to the Wi-Fi 6 the efficiency was measured on (Wi-Fi 6 or 6E × 1, Wi-Fi 7 × 1.031, Wi-Fi 5 × 0.928) × a neighbor factor, capped by the plan and the internet port.

The top PHY rate is the highest MCS for the negotiated generation: Wi-Fi 5 MCS 9 (MCS 8 at 20 MHz) with a 0.4 µs short guard interval, Wi-Fi 6 MCS 11 and Wi-Fi 7 MCS 13 with a 0.8 µs one, their shortest.

Neighbors are other Wi-Fi, not walls: 2.4 GHz loses the most to them, then 5 GHz, then 6 GHz, and wider channels lose more than 20 MHz.

Each range runs from the far edge of your distance (d + u, the ± you entered; with ± 0 both edges are d) at its 10th-percentile throughput over log-normal shadowing (σ = 5 dB, TGax residential), so 90% of spots there do at least this well; throughput only rises with signal, so that percentile is the speed at the 10th-percentile signal, 1.28 × σ = 6.4 dB below the median, up to the near edge at its median throughput. Both ends use one channel width per band: the smallest of the router width, the device max and the band's widest channel, with no narrowing by distance. Out-of-range checks and the Auto rule use the faded signal. A single number means both ends land on the same speed, and its note says why. Auto marks the highest band (6 GHz, then 5 GHz) whose far-edge signal is better than -65 dBm, the band-steering threshold in HPE Aruba's ClientMatch docs; if none is, it marks 2.4 GHz. Expected error: about ±40% of the predicted speed at 1σ for 20–80 MHz, and more at 160 MHz, from ±5–8 dB of signal error at an unknown home and the spread between the efficiency studies. Multi-link operation, mesh hops, and other clients on this Beacon are not modeled. Sources for every number are in docs/beacon-settings-audit.md.

Wi-Fi 6 · 2.4 GHz and 5 GHz · 20–80 MHz · 2×2 · up to MCS 11 · 0.8 µs GI

  • Band is unknown, so this is the slowest to fastest across 2.4 GHz and 5 GHz.
  • Where a band lists several widths, the Auto band mark uses the 80 MHz line's far edge, or the widest listed width when 80 MHz isn't listed.
  • No plan or WAN cap — the link budget and neighboring Wi-Fi set this wireless speed.

Theoretical maximum

287 Mbps – 1.2 Gbps

Top PHY rate before the link budget or caps

LAN output

1 Gbps

Ethernet link speed of this port. Wi-Fi does not change it.

Speed = PHY at that MCS × the measured TCP efficiency for a phone at that width × 1 generation × the neighbor factor for the band and width.

2.4 GHz budget: 195 Mbps, the most 2.4 GHz can do at 20 MHz on Wi-Fi 6

2.4 GHz obstacles: 1 interior × 5 + 0 exterior × 10 + 0 wood floors × 5 = 5 dB.

2.4 GHz near edge, 23 ft at 20 MHz: 30.7 dBm EIRP − 3.9 dBi handset − 59 dB path − 5 dB obstacles (median, no fade margin) = −37.1 dBm; noise −91 dBm; SNR 53.8 dB → MCS 11 (needs 39 dB) → 286.8 Mbps × 0.85 × 1 × 0.8 = 195 Mbps.

2.4 GHz far edge, 23 ft at 20 MHz: 30.7 dBm EIRP − 3.9 dBi handset − 59 dB path − 5 dB obstacles − 6.4 dB fade margin = −43.5 dBm; noise −91 dBm; SNR 47.4 dB → MCS 11 (needs 39 dB) → 286.8 Mbps × 0.85 × 1 × 0.8 = 195 Mbps.

5 GHz budget: 20 MHz, 161 Mbps – 224 Mbps; 40 MHz, 254 Mbps – 423 Mbps; 80 MHz, 430 Mbps – 796 Mbps

5 GHz obstacles: 1 interior × 5 + 0 exterior × 10 + 0 wood floors × 7 = 5 dB.

5 GHz near edge, 23 ft at 20 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles (median, no fade margin) = −46.5 dBm; noise −91 dBm; SNR 44.5 dB → MCS 11 (needs 39 dB) → 286.8 Mbps × 0.85 × 1 × 0.92 = 224 Mbps.

5 GHz far edge, 23 ft at 20 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles − 6.4 dB fade margin = −52.9 dBm; noise −91 dBm; SNR 38.1 dB → MCS 10 (needs 37 dB) → 258.1 Mbps × 0.85 × 1 × 0.92 = 202 Mbps.

5 GHz near edge, 23 ft at 40 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles (median, no fade margin) = −46.5 dBm; noise −88 dBm; SNR 41.5 dB → MCS 11 (needs 39 dB) → 573.5 Mbps × 0.82 × 1 × 0.9 = 423 Mbps.

5 GHz far edge, 23 ft at 40 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles − 6.4 dB fade margin = −52.9 dBm; noise −88 dBm; SNR 35.1 dB → MCS 9 (needs 34 dB) → 458.8 Mbps × 0.82 × 1 × 0.9 = 339 Mbps.

5 GHz near edge, 23 ft at 80 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles (median, no fade margin) = −46.5 dBm; noise −85 dBm; SNR 38.5 dB → MCS 10 (needs 37 dB) → 1080.9 Mbps × 0.78 × 1 × 0.85 = 717 Mbps.

5 GHz far edge, 23 ft at 80 MHz: 26.8 dBm EIRP − 2.2 dBi handset − 66 dB path − 5 dB obstacles − 6.4 dB fade margin = −52.9 dBm; noise −85 dBm; SNR 32.1 dB → MCS 8 (needs 32 dB) → 864.7 Mbps × 0.78 × 1 × 0.85 = 573 Mbps.

Internet on this Ethernet port

A speed test on that cable is the minimum of the plan, the internet port, and the jack. A copy between two wired devices can use the full jack and ignores the plan.

Speed test on the cable
1 Gbps
Computer to computer
1 Gbps
Your plan
Your Beacon
Distance from the Beacon

Straight line, in feet. Add ± if you're not sure: the Monte Carlo below samples evenly between the two ends.

Interior walls in the way
Exterior walls in the way

Brick or siding, no foil-backed insulation.

Floors in between
Other Wi-Fi nearby

Neighbors, not walls.

Fine tuneauto width · 1 Gbps Ethernet
Wi-Fi band

Auto covers every band this phone can join, slowest to fastest.

Router channel width

As set in the Beacon's Wi-Fi settings.

2.4 GHz

5 GHz

5 GHz channel

As set in the Beacon's Wi-Fi settings. Sets the filed transmit power for that U-NII band.

Channel 100 is a DFS channel. If the Beacon detects radar it must leave the channel (47 CFR 15.407(h)) and may move to a non-DFS channel or a narrower width. This estimate does not model that.

Transmit power

As set in the Beacon's Wi-Fi settings.

The device

A typical phone is Wi-Fi 6 and stops at 80 MHz.

Device max on this Beacon: 2.4 GHz 20 MHz · 5 GHz 80 MHz. Speeds use the phone TCP efficiency.

Wired LAN port (router to device)

The LAN jack your computer plugs into — not the WAN/internet uplink, and not the Wi-Fi speed.

Runs the speed test. A laptop helper below adds your Wi-Fi link to the same reading.

Testing over
Speed testDownload, upload, delay and jitter (about 25 s)

How this test works. Your browser downloads for about 10 seconds, then uploads for about 10 seconds, using 6 requests at a time to Cloudflare's nearest site. We skip the first moments, while the connection speeds up, and report the steady speed after that, with the range of its middle 80% of quarter-second slices. We also show the plain 10-second average, which is how M-Lab reports speed.

Delay is how long a tiny request takes to come back: first with the line quiet, then while it is busy. A big jump while busy means your router or line is queueing traffic. Jitter is how much the quiet delay varies from one try to the next.

  • Results depend on your device, Wi-Fi and browser, so a fast line can read lower than its plan.
  • It measures your whole path: Wi-Fi, your internet line and the route to Cloudflare, not the open internet.
  • Uploads are at most 256 KiB per request (free-plan CPU limit), and one test sends at most 1,000 speed requests, so above about 400 Mbps the upload stops early.
  • Shared limit of 20,000 speed requests a day for all visitors; after that the test says "daily test limit reached" until 00:00 UTC.
  • We do not store or publish your IP address.
M-Lab speed test (optional)ndt7, one connection, published by M-Lab

Before you start: this test runs on Measurement Lab (M-Lab) servers, not ours. M-Lab publishes every test's results as open data, including your IP address, the time and the measurements. Beacon Bench keeps the summary on this page only and does not send it anywhere. M-Lab privacy policy.

Optional. The main speed test above stays on Cloudflare; this one is only run when you press the button. Delay before, during and after is a tiny request to this page's own server, median (10th–90th percentile). RTT, retransmits, cwnd and BBR come from M-Lab's server (TCP_INFO, download).

Slow-server diagnosis

Runs the M-Lab test several times on the server picked above and reports the median and spread (10th–90th percentile). Single stream only: ndt7 has no multi-stream mode. A browser cannot choose IPv4 or IPv6, so the version used is recorded.

How to read the server's DSCP (tshark)

Start a capture, run the test, stop the capture, then paste its output below. Use the server IP shown in your result. Run it once on Wi-Fi and once wired. Fields: ip.dsfield.dscp (wireshark.org/docs/dfref/i/ip) and ipv6.tclass.dscp (wireshark.org/docs/dfref/i/ipv6).

Windows (install Wireshark with Npcap; list interfaces with tshark -D; Wi-Fi is usually "Wi-Fi", wired "Ethernet"):

tshark -i "Wi-Fi" -f "host <server_ip>" -T fields -e ip.src -e ipv6.src -e ip.dsfield.dscp -e ipv6.tclass.dscp

macOS (Wi-Fi is usually en0; a USB/Thunderbolt Ethernet adapter is en5 or similar; check with tshark -D):

sudo tshark -i en0 -f "host <server_ip>" -T fields -e ip.src -e ipv6.src -e ip.dsfield.dscp -e ipv6.tclass.dscp

Linux (Wi-Fi wlan0 / wlp…, wired eth0 / enp…; check with ip link):

sudo tshark -i wlan0 -f "host <server_ip>" -T fields -e ip.src -e ipv6.src -e ip.dsfield.dscp -e ipv6.tclass.dscp

Or save a pcap with -w test.pcap and read it later: tshark -r test.pcap -T fields -e ip.src -e ipv6.src -e ip.dsfield.dscp -e ipv6.tclass.dscp.

DSCP to Wi-Fi access category, downstream (RFC 8325 §4.3 Figure 1, UP→AC from Figure 2 / IEEE 802.11 Table 10-1). The "common default" column is the 3-bit mapping described in §2.3. Which one your Beacon uses is not known (UNSOURCED).

DSCPRFC 8325Common default
DF (0)UP 0 AC_BEUP 0 AC_BE
CS1 (8)UP 1 AC_BKUP 1 AC_BK
AF11 (10)UP 0 AC_BEUP 1 AC_BK
CS2 (16)UP 0 AC_BEUP 2 AC_BK
AF21 (18)UP 3 AC_BEUP 2 AC_BK
CS3 (24)UP 4 AC_VIUP 3 AC_BE
AF31 (26)UP 4 AC_VIUP 3 AC_BE
CS4 (32)UP 4 AC_VIUP 4 AC_VI
AF41 (34)UP 4 AC_VIUP 4 AC_VI
CS5 (40)UP 5 AC_VIUP 5 AC_VI
EF (46)UP 6 AC_VOUP 5 AC_VI
CS6 (48)UP 0 AC_BEUP 6 AC_VO
CS7 (56)UP 0 AC_BEUP 7 AC_VO
LE (1)UP 0 AC_BE UNSOURCEDUP 0 AC_BE

A/B check, DSCP 0 vs CS1 on one server: launch a Linux server with scripts/dscp-ab-server.sh as its user-data (port 8080 untagged, 8081 marked CS1), then download its 250MB file on Wi-Fi and wired from each port while capturing. Steps are in docs/dscp-capture.md.

Conditions (simulate)Things your router settings can't change

Link rate from 802.11 math for Beacon 3.1 and this client. Only combinations both can do are offered.

Client Wi-Fi standard
Band
Client spatial streams

Beacon 3.1 has 2, so the link uses at most 2.

Channel width
Guard interval
MCS index

Auto runs a Monte Carlo over your distance, walls and fading and uses the median MCS. Set the MCS or the RSSI to override it.

Computing the MCS distribution…

MCS source: Auto · computing the distribution

Measured efficiency (throughput ÷ link rate)

Unset by default. John's readings: 52% and 64%.

Link rate no link

Throughput computing

HE (802.11ax) · 2 streams · 80 MHz · GI 0.8 µs

The distribution runs in your browser after the page loads.