On the Hacker News homepage today, Wi‑Fi 8 is back, and the reason for a new generation is unusual: the peak figure didn’t grow. The same three bands, the same maximum channel width of 320 MHz, the same 4096-QAM, the same up to eight spatial streams, the same theoretical ceiling of about 23 Gbps, as with Wi‑Fi 7. The IEEE project is called 802.11bn Ultra High Reliability, and the promises from the start are not formulated in gigabits.
Three quarters instead of one number
In the project (PAR) application there are exactly three measurable goals for 802.11be: at least 25% higher throughput at the same signal-to-interference-plus-noise ratio; 25% lower latency at the 95th percentile of the distribution; 25% fewer MPDUs lost — especially when a client roams between access points. The fourth and fifth points are non-quantitative “lower power consumption” and “better peer-to-peer operation.” The same formulations are used in a Samsung Research blog.
The 95th percentile here is not bureaucracy. The average latency on home Wi‑Fi is usually decent — 2–4 ms to the AP. A call and a game fall apart at the tail: every few seconds a packet arrives with over a hundred milliseconds of delay, and you hear precisely that one. Wi‑Fi 8 targets this tail.
What’s already in the draft
The official IEEE 802.11 timelines page (update of August 6, 2026) shows TGbn with the current draft — D2.0. The set of mechanisms is conveniently not divided by marketing sections, but by a single criterion: will it work for you with a single router in an apartment.
Left half — evolution of the link layer: NPCA and DSO about how to occupy spectrum more efficiently, ELR and DRU about range, P-EDCA and LLI about prioritizing latency-sensitive traffic, DPS and MLPM about the battery. Right — multi-point coordination: Co-SR, Co-BF, Co-TDMA, Co-RTWT, Co-CR and seamless roaming SMD.
A noticeable portion of the claimed gains lives in the right column — that is, it requires several APs that coordinate with each other. Experience with Wi‑Fi EasyMesh suggests how this usually ends: coordination works within one brand, while the combination of “ ISP router plus a separately bought repeater” remains the same as before.
NPCA: 140 MHz, which are quiet today
Today access to the medium is tied to the primary channel: a neighbor took it — the station is completely quiet, even if the other 140 MHz out of 160 are free. Non-Primary Channel Access allows you to shift to non-primary subchannels and transmit there.
Dynamic Sub-channel Operation closes the reverse situation: the AP holds 320 MHz, the client can do 80, and such a pair, per MediaTek, loses up to 75% of channel capabilities. DSO allocates a subchannel to the client instead of collapsing the entire transmission to the weakest link. There are no independent measurements and cannot be — hardware does not exist.
Range, promised in decibels
Enhanced Long Range — a 20 MHz frame format with BPSK and QPSK that fixes a long-standing asymmetry: a phone in a distant room hears the AP noticeably better than the AP hears the phone. MediaTek claims a gain in sensitivity of more than 8 dB on the uplink channel compared to the downlink at 802.11g. Distributed-tone RU spreads carriers over a wide band to raise the power per tone within spectral density limits, claiming an expansion of the zone up to 40 meters with four stations on distributed RUs242. This is a chipmaker’s claim participating in standardization, not independent testing.
The calendar that has already been amended twice
PAR was approved in July 2023, the first working group meeting was in November 2023. On the official TGbn page: the ballot for D1.0 was planned for May 2025 and moved to July; the ballot for D2.0 was planned for May 2026 and also moved to July. The schedule has shifted by about two months twice. D3.0 is planned for January 2027, the first SA ballot in May 2027, approval by the working group and 802 EC in March 2028, RevCom and SASB in May. Wi‑Fi Alliance certification timelines vary: Samsung Research cites December 2027, Wi‑Fi NOW issue — CES 2028.
The first devices with Wi‑Fi 8 are expected in 2027 — before the standard is approved. The same story as with Wi‑Fi 6 and Wi‑Fi 7: pre-standard hardware is sold, and compatibility and the full set of features arrive with firmware if they arrive at all. A router will be needed in 2027 — buy it for today’s specs, not for a future upgrade promise.
Check if this is your problem
UHR targets three concrete symptoms. Before waiting for 2028, check if they exist in your setup. Replace the interface name with yours — it will show up with iw dev.
Retractions and losses on the client:
iw dev wlan0 station dump | grep -E 'signal:|tx bitrate|tx packets|tx retries|tx failed'
The ratio of tx retries to tx packets above 10% means frames regularly have to be retried: either the signal is weak or the air is crowded. This is the target of ELR, DRU and NPCA.
Tail latency to the AP (not to the Internet — otherwise measure the provider’s channel):
ping -i 0.2 -c 600 192.168.1.1 \
| awk -F'time=' '/time=/{split($2,a," ");print a[1]}' | sort -n \
| awk '{v[NR]=$1} END{printf "p50=%.1f p95=%.1f p99=%.1f max=%.1f ms\n", \
v[int(NR*0.50)], v[int(NR*0.95)], v[int(NR*0.99)], v[NR]}'
If p50 is about 3 ms and p95 goes to 50–100 ms, you are seeing the tail metric the “−25%” goal is written about.
Neighbor density in the channels:
sudo iw dev wlan0 scan | awk '/freq:/{print $2}' | sort -n | uniq -c | sort -rn | head
Eight or more other BSSs on your frequency — that is the busy primary channel that NPCA is designed for.
What can be fixed today, without waiting for a new standard:
- Tail latency from retries — narrow the channel. In a dense multi-story building, 80 MHz on 5 GHz usually yields better p95 than 160 MHz: fewer intersections with neighbors and fewer reasons to be silent.
- Roaming drops between APs — enable 802.11k/v/r on all APs at once. This isn’t SMD, reassociation remains, but the pause is reduced from seconds to tens of milliseconds.
- Loaded 2.4 GHz — move everything that can to 5 GHz and don’t touch the channel width on 2.4: 20 MHz remains the only sensible option.
About the third band in Russia. In December 2022 the GKChR lifted the ban on the 5.9–6.4 GHz band for indoor Wi‑Fi. Since then it’s stalled: in January 2024, Kommersant wrote that the list of equipment subject to registration with Roskomnadzor was not aligned with the decision, so mass Wi‑Fi 6E did not happen. No fresh official confirmation that the issue was closed could be found — status should be double-checked before buying 6 GHz devices.
Wi‑Fi 8 is not an upgrade people are waiting for, but a recognition that the race to a peak number hit a ceiling, and the coming years will be spent on spectrum discipline. Devices in 2–3 years, mature firmware will take longer. The tail latency in your apartment can be measured and shaved today.
Related on the forum: 26th channel won’t save you: four myths about Zigbee and Wi‑Fi on one band and How to squeeze 5 Gbps from PPPoE on UniFi.
Sources
- IEEE 802.11 Official Timelines — update as of August 6, 2026
- TGbn Status Update — project milestones and schedule shifts
- Samsung Research: IEEE 802.11bn Ultra High Reliability (Wi-Fi 8)
- MediaTek, «Pioneering the Future with Wi-Fi 8», Part 3 — figures on ELR, DRU, DSO
- Tom’s Hardware: Next-gen Wi-Fi 8 focuses on reliability instead of speed (with a link to Qualcomm)
- Wi-Fi NOW: timelines for standard completion and certification
- Habr: Media — in Russia the Wi-Fi 6E standard has been approved and «Kommersant»: legalization of Wi-Fi 6E has been slowed
Measure your p95 to your access point right now — and write in the comments three numbers: range, channel width, and p95. It’s interesting to see who really has tail latency, and who has problems somewhere else.