LAN, Wi-Fi or powerline: what lowers ping?

Your ping spikes in the final circle, and the internet's first answer is "buy a gaming router". Cheaper and more reliable: measure first. Twenty minutes with a terminal beat a month of guessing.

The short answer

What lowers ping: LAN, Wi-Fi or powerline? A cable usually gives the steadiest local link, because an Ethernet cable joins exactly two ports and in full-duplex mode the collision handling is switched off [2][3]. Wi-Fi shares the air with every device on the channel and has to wait its turn [9]. Powerline shares your house wiring, so its quality depends on that wiring [20][21]. Which of the three fixes your lag is still a measurement question. Ping the gateway (your router) 100 times over each link, compare the worst values and the packet loss instead of the average, and repeat once while the line is busy. Our assessment: if you can lay a cable, do that; if you cannot, the numbers decide between Wi-Fi and powerline.

How it works

1. What you measure. Ping sends a small ICMP echo request and reports the round-trip time of the answer [11]. Three things matter for games. The typical value, the spread between packets (jitter) and packet loss. The IETF defines delay variation as the difference in delay between selected packets of a stream [16]. A spike is a few packets that arrive late, and a game feels it as a rubber-band moment even when the average looks fine. The gateway is the router, the first device beyond your PC. Pinging it covers only your own hop, without your provider and without the game server. The ping manual page recommends the same order for fault isolation: nearby hosts first, then gateways and hosts further away [12].
2. Why a cable is steady. The Ethernet standard IEEE 802.3-2022 specifies shared-medium (half-duplex) operation and full-duplex operation [2]. On a point-to-point link in full-duplex mode, the CSMA/CD algorithm is disabled and collisions cannot occur [3]. A cable can still be slow, for example with a damaged plug or a congested switch, but its timing has fewer sources of random delay than the other two links.
3. Why Wi-Fi wobbles. Wi-Fi is specified in IEEE 802.11. The edition on the IEEE page is 802.11-2024, published on 28 April 2025 [1]. Its basic access method is "listen before talk": a station that finds the channel busy waits for a random backoff time before it tries again, and frames without an acknowledgement are retried [9]. Your neighbours' networks on the same channel count as well. The waiting time is jitter. A two-month field study on 47 access points of one university campus found that the Wi-Fi hop took more than 60 percent of the round-trip latency in more than half of the cases, and that more than half of the TCP packets at some access points waited longer than 20 ms on the Wi-Fi hop [10]. The same study links airtime utilisation above 55 percent to a high chance of bad latency, and moving one access point to a quieter channel cut its worst latency from 250 ms to 50 ms [10]. Be careful with that source: it is a campus from 2016, not your living room, and it shows a mechanism rather than a number for your home.
4. Which band helps. Microsoft describes 2.4 GHz as longer range but more congested, because microwaves, cordless phones and Bluetooth devices use it. 5 GHz has more non-overlapping channels and is less congested. 6 GHz adds low congestion and lower latency, but has the shortest range and needs a Wi-Fi 6E router plus a PC with a matching adapter and driver [8]. The Wi-Fi Alliance defines Wi-Fi 6E as Wi-Fi 6 in the 6 GHz band [7]. Wi-Fi 7 is based on IEEE 802.11be and adds multi-link operation, which the Alliance lists as a way to reduce latency [6]. We found no independent measurement of that gain in homes on 25 Oct 2026, so we give no number. None of this changes what a concrete floor does to a radio signal.
5. What powerline is. Powerline sends data over the mains wiring. The IEEE standard for it is IEEE 1901-2020 [4]. The IEEE page names a successor, IEEE 1901-2026; we have not reviewed it. HomePlug AV2, published in January 2012, extends the band to 86 MHz and adds MIMO over line, neutral and protective-earth pairs [5]. The headline numbers on the box are physical-layer rates under ideal conditions. The overview paper lists 197 Mbit/s for HomePlug AV, 556 Mbit/s for IEEE 1901, 1012 Mbit/s for AV2 and 2024 Mbit/s for AV2 with MIMO as theoretical maximums [5]. They say nothing about latency. Manufacturers explain why real links fall short: power strips and surge protectors filter the signal [20][21], motors and chargers add noise [20], and adapters on different phases of the house wiring lose performance or fail to connect [21][22]. All adapters in one powerline network share the bandwidth, and mixed adapters run at the speed of the slowest one [22].
6. What ping cannot fix. A local cable does not shorten the way to the game server. And when your line is full, queues in modem and router fill up and delay rises on every link, cable included. That effect is called bufferbloat. The bufferbloat project treats a rise of less than 15 to 25 ms under load as well under control [18]; the fix is queue management in the router, not another cable [17]. A German retailer guide from 20 Sep 2026 lists local latency ranges per link type without naming a measurement method [23]. We do not reuse those numbers, because they differ from flat to flat. Measure yours.

Measure it: 100 packets per link

Choose a fixed time, close downloads and streams, and run the same steps over each link you can build: Wi-Fi at the place where you play, powerline if you own adapters, and a cable even if it lies across the floor for the test.

1. Note the link. On Windows, netsh wlan show interfaces lists the wireless interface [13]; look for the lines with radio type, channel and signal. We did not run it on a Windows PC for this article. On Linux, iw dev wlan0 link shows the connected access point, the frequency, the signal and the last transmit bit rate [14]. Replace wlan0 with your interface name.
2. Find the gateway. Windows: run ipconfig and read the line "Default Gateway". Linux: ip route show default [15]; the address after via is the gateway. The addresses below are placeholders.
3. Ping it 100 times. Windows: ping -n 100 192.168.178.1, where -n sets the number of echo requests (default 4) [11]. Windows prints whole milliseconds, and a good cable shows "<1ms", so it cannot resolve small differences. Linux: ping -c 100 -i 0.2 -q 192.168.178.1. -c stops after 100 packets, -i 0.2 waits 0.2 seconds between packets (only the super-user may go below 2 ms), and -q prints just the summary [12]. The summary line shows loss and min/avg/max/mdev. The manual describes mdev as the standard deviation of the round-trip times and says that the higher it is, the more variable the round trip is [12]. It is not the RFC 3393 delay variation [16], but it is a fair measure of spread.
4. Get the spread on Windows. This PowerShell snippet reads the 100 answers, matches English and German output, and prints minimum, average, maximum, the number of answers and an approximate 95th percentile. We reviewed the syntax but did not run it on Windows. A "<1ms" answer counts as 1. The snippet is below the list.
5. Repeat under load. Local load: start iperf3 -s on a wired computer in your network and run iperf3 -c SERVER-IP -t 30 from the PC under test, while the gateway ping runs in a second window [19]. Internet load: ping an outside host while you run one of the speed tests that the bufferbloat project lists, and watch whether the times jump during download or upload [17].
6. Log it. Run each link at least three times, at different times of day. Copy the table below the list and fill it with your own values and dates.

PowerShell snippet for step 4:

$t = ping -n 100 192.168.178.1 | Select-String '(?:time|Zeit)[=<](\d+)\s?ms' | ForEach-Object { [int]$_.Matches[0].Groups[1].Value }
$t | Measure-Object -Minimum -Average -Maximum
"Answers: $($t.Count) of 100"
($t | Sort-Object)[[math]::Ceiling($t.Count * 0.95) - 1]

Log table for step 6:

Reading the result

Compare each link with your own cable run and with its own idle run. The sizes of the differences matter, not a magic threshold. The only external yardstick we found is the bufferbloat project's 15 to 25 ms rise under load [18].

A scenario, not our data

Imagine a PC two rooms away from the router. The cable run (placed across the floor for the test) shows no loss and a tight spread. Wi-Fi on 5 GHz shows no loss, but the maximum is several times the average, and it gets worse when a TV streams in the next room. Powerline shows 3 lost packets out of 100 whenever the washing machine runs. Decision: the cable wins, so you route a flat cable along the skirting board. If that were forbidden, Wi-Fi on 6 GHz or 5 GHz with the access point moved closer would be the next test, and powerline would be the last option, bought with a return option. We have not measured these three links in our own lab for this article. The method is the deliverable.

Checklist

• Measure the gateway ping with 100 packets on every link before you buy anything.
• Write down band, channel, signal and the time of each run.
• Compare maximum and loss, not only the average.
• Repeat each link three times at different times of the day.
• Run one test with local load and one with internet load.
• Plug powerline adapters directly into wall sockets, never into power strips or surge protectors [20][21].
• Check the router for queue management or QoS if the outside ping rises under load.
• Buy hardware that you can return, and re-measure after every change.

Related pages: TCP, DNS, 2.5 GbE in the homelab. Our assessment: a faster cable class helps file transfers, not the ping of a game that sends little data.