Lag is not a speed problem: the gamer's guide to a connection that holds up in a match

Most people who lag in games own a connection that is more than fast enough. They have already paid for a bigger plan, bought a router with antennas on it, and maybe tried a "gaming VPN." The lag stayed, because none of those things touch the four numbers that actually decide whether a match feels crisp or sludgy.

This guide shows you how to measure those four numbers on this site, how to tell which one is hurting you, and what to change in order of cost. Most fixes are free. A few are a cable. The expensive options come last, with honest notes on what they do and do not buy you.

Why your plan speed barely matters

A multiplayer game sends small packets describing where players are and what they did. It does not stream video. Typical online play uses somewhere between 40 and 300 megabytes per hour, which works out to well under one megabit per second of sustained traffic (TestMySpeed). Streaming one HD movie uses more bandwidth than a whole evening of ranked matches (How-To Geek).

That is why upgrading from 300 to 1,000 megabits does nothing for your ping. Bandwidth is how wide the pipe is. Lag is about how long each packet waits, and whether it arrives at all. The exceptions are game downloads, which do scale with speed, and cloud gaming services like GeForce NOW or Xbox Cloud Gaming, which are video streams and need 20 to 45 megabits plus very low latency (mbpstogb).

The four numbers that decide a match

Ping (unloaded latency). The round-trip time between you and a server when nothing else is happening. Research on first-person shooters consistently puts the point where play degrades at around 100 milliseconds, with third-person and sports games tolerating up to about 500 and strategy games up to about 1,000 (Claypool and Claypool, WPI). For competitive play the bar is far lower. In a controlled Counter-Strike study, players at 25 milliseconds scored about 17 percent higher than the same players at 150 milliseconds, and the benefits were linear all the way down, so every 20 or 30 milliseconds you shave off is worth having (Liu et al., WPI).

Ping under load (bufferbloat). The same measurement taken while your connection is busy, for example while someone in the house is downloading a game update or watching 4K video. On many home connections the ping triples or worse the moment the line fills up, because the router or modem holds packets in a long queue instead of sending yours through. This is the number that explains "my ping is 20 in the menu and 200 when the match starts." It is also the number most speed tests never show you.

Jitter. How much the ping wobbles from one packet to the next. A steady 40 milliseconds feels better than a 20 that spikes to 90 every few seconds, because games predict movement based on a consistent rhythm. A 2026 study that varied latency and jitter independently found players rated jitter as the more noticeable of the two (WPI, 2026).

Packet loss. Packets that never arrive. Games cannot wait for a resend, so a lost packet becomes a teleporting enemy or a shot that never registered. Any sustained loss above about 1 percent is worth investigating; above 5 percent most real-time games break down.

The Speed and Network test on this site reports all four, and it is the only one of them you need to care about that the "speed" gauge does not cover.

Run the gaming audit

Each step takes a few minutes. Do them in order; each one rules something in or out.

Step 1: the quiet baseline

Pick a time when nobody else is using the connection. If your gaming device can plug into the router by cable, do that for this test even if you normally play on Wi-Fi, because you want to know what the connection itself can do before adding the wireless variable. Run the test on the device you game on if it has a browser (Xbox has Edge, Steam Deck has one in desktop mode, any PC or phone works). If not, use a laptop in the same spot with the same connection type.

Open Settings first, set your plan speeds so the grade is measured against what you pay for, and choose the "Internet speed and my ISP" focus. Then run the test and write down:

• Latency (Unloaded): the Best figure and whether all four endpoints are similar
• Latency (Loaded): the letter grade
• Jitter in milliseconds
• Packet loss percentage

What good looks like: unloaded latency under 20 milliseconds is imperceptible and under 60 is normal for decent broadband; jitter under 5 is excellent and anything over 30 will be felt in competitive games; packet loss should be zero.

Step 2: the busy-house test

Now make the connection busy on purpose. Start a large download on another device, or stream 4K video on the TV, or begin a cloud backup. While that runs, test again.

If unloaded latency barely moved but the Loaded Latency grade is C, D, or F, you have found the most common cause of game lag in homes with fast internet. Your plan is fine; your router is queueing. Skip ahead to "Fix 2."

Step 3: the Wi-Fi tax

If you game on Wi-Fi, repeat Step 1 twice at the gaming spot: once wired, once wireless. Use the Watch monitor's Spot label so the comparison is saved side by side ("Desk wired" versus "Desk Wi-Fi"). Look at jitter and packet loss rather than speed.

RTINGS tested eleven routers, from budget to premium gaming models, and found that Ethernet produced virtually no added latency or jitter while every Wi-Fi configuration added what they called a "Wi-Fi tax" with periodic spikes, regardless of router model or distance (RTINGS). If your wireless jitter is within a couple of milliseconds of wired, your Wi-Fi is fine for gaming. If it is in double digits, the fix is a cable, not a new router.

Step 4: the evening watch

Lag that only shows up at 8 p.m. is usually neighborhood congestion or household traffic, not your equipment. Start the Watch monitor at 15-minute intervals, leave the tab open from early evening until bedtime for two or three nights, and look at the chart. A connection that is steady all day and degrades every evening is an ISP conversation, and the Evidence Pack packages the timeline for that call. Our ISP call script covers what to say.

Step 5: can other players reach you?

This step is about matchmaking failures, party chat that drops, and "unable to join" errors rather than ping. Those are usually NAT problems, and NAT type has no effect on latency at all; an Open and a Strict NAT on the same line have identical ping (CheckPing). What Strict NAT does is stop other players from connecting to you directly, which means longer queues, failed invites, and no hosting.

Two checks on this site help:

1. Note the public IP shown at the top of your report. Then log in to your router and look at its WAN or Internet IP. If they differ, and especially if the router shows an address starting with 100.64 through 100.127 or 10, your ISP is sharing one public address among many customers (carrier-grade NAT). No router setting can open your NAT in that case; you need to ask the ISP for a public IPv4 address or rely on IPv6 where your game supports it (Tutorialpedia).
2. Run the IPv6 Reachability check. Working IPv6 is a way around carrier-grade NAT for games and consoles that use it, and is worth turning on at the router if it is off.

If you have a real public IP, enabling UPnP on exactly one router (not on both an ISP gateway and your own router) resolves most Strict NAT cases. If you run your own router behind an ISP gateway, put the gateway in bridge mode so there is only one NAT. The inbound port scan will confirm whether a forwarded port actually reaches you from the outside.

Fixes in order of cost

Fix 1 (free): stop the traffic that is drowning you

Before touching any setting, find what else is using the line during matches. The usual culprits are consoles downloading updates in standby, Steam or the Epic launcher updating in the background, cloud photo backups, and security cameras uploading. RTINGS found that ordinary browsing, streaming, calls, and gaming almost never overload a home connection on their own; it is the bulk transfers that do it (RTINGS). Schedule downloads for overnight, and on consoles turn off automatic updates during play hours.

Fix 2 (free, if your router supports it): smart queue management

If your Loaded Latency grade was poor, this is the fix. In your router's settings look for Smart Queue Management, SQM, Adaptive QoS, CAKE, or FQ-CoDel. Enter download and upload limits of about 85 to 90 percent of the speeds you measured in Step 1 (the measured number, not the plan number). Then rerun the busy-house test. A working setup moves the grade from D or F to A or B in one step and keeps loaded latency within a few milliseconds of idle.

Two cautions. First, settings labeled "Gaming Mode," "Game Boost," or "Media Priority" are not the same thing and can make bufferbloat worse; they prioritize traffic they recognize rather than keeping the queue short. Second, SQM costs a little throughput, typically 5 to 15 percent, which is a fine trade since the game needs almost none of it. Our bufferbloat tutorial has brand-specific steps.

Upload is the direction to fix first on cable connections. Cable upload is a fraction of download, so it fills up long before download does, and a filled upload queue delays every packet your game sends.

Fix 3 (a cable): wire the gaming device

If Step 3 showed a Wi-Fi tax, run Ethernet. Where a cable cannot reach, a MoCA adapter pair over existing coax is the next best option and behaves almost like Ethernet; powerline adapters are a distant third and vary with your house wiring. Only if none of those are possible should you spend money on better Wi-Fi, and then the goal is the 6 GHz band (Wi-Fi 6E or 7) with the device in the same room as the access point. Ookla's real-world data shows Wi-Fi 5, 6, and 6E deliver essentially the same latency, and Wi-Fi 7's advantage over 6 is a few milliseconds at most (Ookla).

Fix 4 (free): the TV and the PC

Here is the finding most guides skip. Your screen and your own machine add delay before any packet leaves the house, and that local delay matters more than network delay for aiming. A WPI study found that cutting 100 milliseconds of local latency improved accuracy by 6 percent, three times the gain from cutting the same 100 milliseconds of network latency (WPI). NVIDIA's experiment with more than 15,000 players in an aim trainer found they hit over twice as many targets at 25 milliseconds of system latency as at 85 (NVIDIA).

The biggest single source of local delay for console players is a TV not in Game Mode. In Game Mode, RTINGS measures TVs at an average of about 12 milliseconds at 60 Hz and 7 at 120 Hz; outside Game Mode the same sets are often several times higher (RTINGS). Turn on Game Mode, and enable Auto Low Latency Mode (ALLM) so the TV switches automatically when the console wakes. On PC, turn on NVIDIA Reflex or AMD Anti-Lag in games that offer it, and cap frame rate just below your monitor's refresh rate.

Fix 5 (free): pick the right server

Distance decides ping. Light in fiber covers roughly 100 kilometers per millisecond round trip, so a server 1,500 kilometers away cannot be reached in under 15 milliseconds no matter what you buy. Most games let you choose a region or show per-server ping; pick the lowest stable one, not the one your friends happen to use. The four-endpoint ping matrix on this site is a sanity check: if one endpoint is far slower than the other three, that is a routing or peering issue to raise with your ISP, and the test page names the network to ask about.

The expensive options, honestly

A faster plan

More megabits help in exactly two cases: your household genuinely saturates the line for hours at a time, or you download large games often. If the busy-house test was the only one that failed, SQM is a cheaper fix than a bigger plan, because the queue forms at whatever speed you buy.

A different kind of plan

This is the upgrade that can actually change your ping. Across the FCC's last national measurement, fiber connections idled at 7 to 14 milliseconds, cable at 12 to 24, and DSL at 23 to 34 (FCC, Thirteenth Measuring Broadband America Report). Ookla's 2025 data shows the same gap, roughly 15 milliseconds on fiber versus 25 on cable (Ookla). Fiber also has symmetric upload, which removes the upload-queue problem in one stroke. If fiber is available at your address and you play competitively, it is the one connection change with a measurable payoff.

If you are on Comcast Xfinity, there is a second option that costs nothing. Comcast has deployed Low Latency DOCSIS, based on the IETF's L4S standard, to more than ten million homes as of early 2026. Loaded latency for supported traffic falls to around 18 to 20 milliseconds, close to idle (Livingood, IETF draft). It currently applies to applications that mark their traffic, which at launch included many Steam games, GeForce NOW, and FaceTime (Comcast). It requires a compatible gateway, so ask whether your modem supports it.

Starlink, for those without wired options, reports a median peak-hour latency of 25.7 milliseconds in the United States with fewer than 1 percent of measurements above 55 (Starlink). That is adequate for most online play and short of what a competitive player on fiber sees. Older geostationary satellite service, at 600 milliseconds or more, is not usable for real-time games.

A router

Buy one only if your current router lacks a working SQM option, or if Step 1 shows the router itself is the bottleneck (wired results far below plan speed). RTINGS found no meaningful difference in gaming latency between budget and premium routers when the network was not congested, and that nine of eleven routers fixed congestion latency once QoS or SQM was turned on, so the feature matters and the price tag does not (RTINGS). Look for CAKE or FQ-CoDel in the spec sheet, or a model that runs OpenWrt. Our gear pages list specific models that pass the loaded-latency test.

A VPN

A VPN adds a hop, so it adds ping. In a test of six VPNs inside Counter-Strike 2 on a fixed server, the unprotected connection was 42 milliseconds, the best VPN was 57, and the worst was 67 (Gizmodo). Three situations justify one anyway: your ISP routes badly to a specific game region (test with and without, same server), you are stuck behind carrier-grade NAT and the VPN offers port forwarding, or you stream and want your home IP hidden from viewers. If you use a VPN for that last reason, run the WebRTC leak test with the VPN on, because a leaking browser gives away the address you paid to hide.

Quick reference

Sources

• TestMySpeed. How Streaming and Gaming Affects Data Usage. testmyspeed.com
• How-To Geek. Does Online Gaming Really Use a Ton of Bandwidth? howtogeek.com
• Claypool and Claypool. Latency and Player Actions in Online Games. CACM 2006. wpi.edu
• Liu et al. The Effects of Network Latency on Competitive FPS Players. QoMEX 2021. wpi.edu
• WPI. Evaluating the Effects of Network Delay and Jitter on Gameplay. 2026. wpi.edu
• WPI. Network Latency and Peeker's Advantage. 2022. wpi.edu
• Schmid et al. Small Latency Variations Do Not Affect Player Performance in FPS. 2023. dl.acm.org
• Liu et al. How Player Skill Alters the Effects of Network Latency. 2021. dl.acm.org
• NVIDIA. Improving Player Performance with Low Latency. nvidia.com
• RTINGS. Buying a Gaming Wi-Fi Router Won't Help Your Ping. 2025. rtings.com
• RTINGS. PC Monitor vs TV. rtings.com
• FCC. Measuring Fixed Broadband, Thirteenth Report. 2024. fcc.gov
• Ookla. Wi-Fi 7 Speeds Up in the U.S. 2025. ookla.com
• Comcast. Nation's First Ultra-Low Lag Xfinity Internet Experience. 2025. comcast.com
• Livingood. ISP Dual Queue Networking Deployment Suggestions. IETF Internet-Draft, 2026. datatracker.ietf.org
• Starlink. Network Update. 2025. starlink.com
• Gizmodo. Best VPN for Gaming, Ping Tested in CS2. 2026. gizmodo.com
• CheckPing. NAT Type Explained. checkping.io
• Tutorialpedia. Change NAT Type on PC. tutorialpedia.org
• mbpstogb. How Much Data Does Online Gaming Use? mbpstogb.com

The guide above is enough to find and fix almost any lag. These notes are for readers who want the mechanics and the edge cases.

Advanced notes

• Why bufferbloat survives fast plans. Queueing delay is created wherever traffic arrives faster than the next hop can send it. On a 1 Gbps plan that point is still the cable modem's upstream, which is why upload saturation on cable is the classic trigger. SQM moves the bottleneck into your router, where FQ-CoDel or CAKE can keep each flow's queue short.
• Two kinds of latency variation. Lab studies disagree on jitter: small continuous variation of about 50 milliseconds did not measurably change performance in one experiment (Schmid et al.), while perceived quality fell with jitter in another (WPI, 2026). The practical reading: a steady few milliseconds of jitter is harmless; irregular spikes are what players notice, and those are usually Wi-Fi or congestion.
• Single-stream mode. The site's single-stream speed setting is closer to what one application experiences than the multi-stream figure. If single-stream is far below multi-stream, per-flow fairness or a poor route is in play, which matters more for game downloads than for the match itself.
• Skill amplifies latency. The Counter-Strike studies found experienced players lose more to latency than novices do (Liu et al.). If you are climbing ranks, the fixes above pay off more as you improve.
• The ping you see in-game includes server tick timing and may differ from any browser test. Use this site to diagnose the path; use the in-game net graph to confirm the result.

The sources for every figure cited here are listed at the end of the Simple version of this guide.