Speed Is the Least Important Thing on the Sales Page

The best ISP for gaming is rarely the one selling the biggest number. Above roughly 100 Mbps, extra bandwidth changes nothing about how a match feels, because online games exchange well under 1 Mbps in each direction. What decides your experience is the connection technology, how well the ISP routes to the data centres your games run in, whether latency collapses when someone else on the connection starts a download, and whether carrier-grade NAT breaks peer-to-peer titles. None of that appears in the advertised tier.

Why Speed Tiers Stop Mattering So Early

A competitive shooter sends small packets many times per second. The total is trivial. Where bandwidth genuinely matters is everything around the game: downloading a 150 GB title, streaming your gameplay out, and sharing the line with other people in the house. Those are real reasons to buy more than the minimum, but they are throughput problems, not latency problems.

The practical read: pick a tier that covers your household's simultaneous downloads and uploads comfortably, then stop optimising for speed and start optimising for latency behaviour. A 300 Mbps line with clean routing and no bufferbloat will beat a gigabit line without either, every time, and the sales page will never tell you which one you are buying.

Connection Technology Sets Your Latency Floor

Every access technology has a physical and protocol-level minimum delay before your packet even reaches the ISP's network. You cannot optimise below it. These are typical ranges reported for each technology under normal conditions, not measurements from any specific network, and real figures vary widely by provider, equipment and how far you are from the exchange or node.

Technology	Typical last-mile latency floor	Consistency	Notes for gamers
Fibre to the home (FTTH / PON)	Roughly 1-5ms	Excellent	The best consumer option. Symmetric upload on many plans, minimal congestion sensitivity, low jitter.
Cable (DOCSIS 3.1)	Roughly 5-15ms	Good, but node-dependent	Upstream uses a request-and-grant scheduler that adds delay. Shared neighbourhood node means evening congestion is possible.
VDSL / DSL	Roughly 10-25ms, more with interleaving	Moderate	Distance to the cabinet dominates. Error-correction interleaving trades latency for stability; some ISPs can disable it on request.
Fixed wireless (WISP)	Roughly 10-30ms	Variable	Depends heavily on line of sight, tower load and weather. Quality ranges enormously between operators.
5G home internet	Roughly 15-50ms	Variable	Very sensitive to signal quality, cell load and time of day. Convenient where fixed lines are poor, but rarely competitive with fibre.
Low-earth-orbit satellite	Roughly 25-60ms	Variable, with periodic spikes	Playable for many genres but not competitive-grade. Periodic latency spikes are inherent to satellite handovers.
Geostationary satellite	Roughly 600ms or more	Structurally fixed	The orbital distance alone makes real-time gaming impractical. No configuration fixes this.

Notice the ordering rarely changes even when a slower fibre plan competes with a faster cable plan. A 300 Mbps fibre line will usually feel better in game than a 1 Gbps cable line, which is the core argument in our fibre vs cable comparison.

Peering and Routing: The Invisible Difference

Two customers on different ISPs in the same street can see very different ping to the same game server. The reason is routing. Game servers mostly live in a handful of large cloud regions and data centre hubs. An ISP with direct peering to those networks, or a presence at the same internet exchange, hands your traffic over locally. An ISP that buys wholesale transit and hauls traffic to a distant city first adds tens of milliseconds that no router setting will recover.

Signs of good routing to look for:

Traffic reaches a major exchange point or cloud provider within a small number of hops, in your own region.
Ping to a nearby game region is close to what the technology's floor plus geographic distance would predict.
No sudden geographic detour in a traceroute — traffic leaving your city, crossing the country and coming back is a classic symptom of thin peering.

This is checkable before you sign if you know someone on that ISP nearby. Ask them to run a traceroute to your usual game's region and send you the output.

Bufferbloat: The Problem That Ruins Good Connections

Bufferbloat is when oversized buffers in your modem or the ISP's equipment queue up packets during heavy transfers, sending latency from 20ms to several hundred milliseconds the moment somebody starts an upload. It is extremely common, it is invisible on a standard speed test, and it is the single most likely explanation for a fast connection that plays badly.

When evaluating an ISP, ask whether their supplied equipment implements modern queue management, and whether you are permitted to use your own router. Being allowed to put your own device with Smart Queue Management in the path is often more valuable than any plan feature, because it lets you fix the problem yourself. Our bufferbloat guide explains how to test for it and what to configure; the key point at purchase time is to test loaded latency, not idle latency.

CGNAT, Static IPs and P2P Games

Carrier-grade NAT means the ISP shares one public IPv4 address among many customers. It saves them addresses and it works fine for browsing, streaming and most modern games that connect out to dedicated servers. It causes real problems elsewhere:

Peer-to-peer and player-hosted matches can fail to establish inbound connections, producing failed lobbies or forced client roles.
Port forwarding is impossible, since you do not control the shared public address. Console NAT types often report as moderate or strict.
Self-hosted game servers for friends are effectively off the table.
Shared-reputation issues occasionally trigger anti-cheat or rate-limiting behaviour tied to the shared address.

Fixed wireless, 5G home internet and satellite are the most likely to deploy CGNAT; fibre and cable vary by provider. Ask directly: "do you use CGNAT, and can I get a routable public IPv4 address?" Many ISPs will provide one on request, sometimes free, sometimes as a small add-on. Full IPv6 support also helps for games that support it, since IPv6 sidesteps the address-sharing problem entirely.

A static IP, by contrast, is rarely worth paying for as a gamer. A dynamic but routable public address does everything a game needs. Static addressing matters for hosting services with a domain name, not for playing.

Data Caps and Evening Congestion

Modern game installs and patches are enormous, and a monthly cap that looked generous in a browsing-only household disappears fast when two people are downloading new releases. Check the cap, check the overage policy, and check whether there is a deprioritisation threshold, which is common on wireless plans and functions as a soft cap that degrades you during busy periods.

Congestion is the related question. Cable nodes, PON splits and wireless cells are all shared, so performance at 9pm can differ sharply from performance at 2pm. This is precisely the pattern described in our guide to why ping goes up at night. When you research a provider, everything you learn should be weighted toward peak hours.

The Pre-Purchase Research Checklist

Identify the actual technology at your address. Not the marketing name. Fibre to the premises, fibre to the cabinet, cable, fixed wireless and 5G are very different products, and some "fibre" branding covers a copper final segment.
Ask neighbours on the same provider. This is the highest-value research available and costs nothing. Local forums, neighbourhood groups and building chats will tell you about evening congestion and outages that no review site captures.
Get a traceroute from a nearby customer to your main game's server region. Look for the number of hops before reaching a major exchange or cloud network, and for any geographic detours.
Test loaded latency, not just speed. If you can trial the service or test at the address, run a latency measurement while a large download is in progress. A connection whose ping triples under load will feel bad regardless of its speed tier.
Ask about CGNAT and IPv6 explicitly. Get the answer before signing, especially if you play peer-to-peer titles or host for friends.
Confirm you can use your own router. This preserves your ability to fix bufferbloat and control your own network.
Check the cap, the overage terms and any deprioritisation threshold.
Check the contract length and cooling-off period. A short trial window is your only real protection against a connection that tests fine in week one and congests in month three.
Baseline immediately after installation. Measure on day one so you have evidence if performance drifts later.

How to Verify Once You Are Connected

Run our browser ping and jitter test at several times of day, including a weekday evening, and note latency, jitter and packet loss together rather than only the headline figure. Then check results against the specific game regions you play in, since the route to one publisher's data centre can be much better or worse than to another's. The test measures warm HTTPS round-trips to nearby cloud endpoints — a real-world proxy rather than in-game packets — but it is consistent across runs and across times of day, which makes it a fair way to detect congestion patterns.

Compare what you see against the thresholds in what is a good ping for gaming. If your numbers sit above the floor your technology should deliver, rule out your own network first — Wi-Fi, an overloaded router and background transfers explain far more poor results than the ISP does. If the gap persists on a wired connection with an idle network, you have the evidence you need to escalate, or to switch.