Fiber, cable, 5G or satellite for gaming: what the FCC data shows

The short version

• Fiber has the lowest lag in the FCC's data. In the FCC's most recent fixed-broadband measurements (the Thirteenth Measuring Broadband America report, data from September and October 2022), the median idle latency during evening peak hours was about 10 ms on fiber, 19 ms on cable and 31 ms on DSL.
• The bigger difference shows up when the line is busy. While a speed test filled the upload, the fiber providers measured 24 to 41 ms, the cable providers 88 to 252 ms and the DSL providers 181 to 720 ms. On the cable lines, a full upload added 77 to 235 ms, many times the 9 ms idle gap between fiber and cable, and much of that is bufferbloat you can reduce at home with SQM.
• 5G home internet and Starlink aren't in the FCC's fixed measurements. Their own published figures: T-Mobile 16 to 28 ms on its main home plans, Verizon 35 to 70 ms depending on plan, Starlink 25 to 60 ms on land. Geostationary satellite can't go below about half a second, by physics alone.
• Distance and your home network can outweigh the technology. Each 1,000 km of fiber to the game server adds about 10 ms to the round trip, about the same as the idle gap between fiber and cable, and a weak Wi-Fi hop adds spikes no provider can fix.
• Caveat: the FCC data is four years old, covers ten providers and only tiers below a gigabit, and some providers have since changed how their networks queue traffic. Use it to compare technologies, not to rank today's providers.

What the FCC measured, by technology

The FCC's Measuring Broadband America (MBA) program puts a small measurement box (a "whitebox") in volunteers' homes, wired by Ethernet to their router, and has it test against servers at ten major internet exchange points. Each box picks the test server with the lowest round-trip time, so the results describe the provider's own network, not the distance to any particular game. The Thirteenth report, posted in August 2024, is the most recent, and the FCC says the MBA program "is currently not collecting additional data". It covers ten providers, 2,025 panelists' results and twelve provider-and-technology combinations.

It measured latency two ways. Idle latency is the round trip of small test packets when the home is otherwise quiet. Latency under load (the FCC also calls it working latency) is measured during the report's 10-second download and upload speed tests, which fill the line. These are the FCC's figures for weeknight peak hours, 7 to 11 p.m.:

The loaded figures come from the FCC's chart data, where it shows them without provider names because 2022 was the first year it reported the metric and the metric wasn't announced before the testing period. Three patterns stand out:

• Idle, the technologies sit close together. The fiber and cable medians are about 9 ms apart (arithmetic: 19.1 minus 9.9). For most games, that's a small part of the total round trip.
• Under load, they spread apart. A full upload added 13 to 31 ms on the fiber providers, 77 to 235 ms on the cable providers and 154 to 696 ms on the DSL providers (our arithmetic, loaded minus idle for each provider). On cable and DSL the upload is the narrow direction (the cable tiers measured had advertised upload speeds of 5 to 50 Mbps), so that's where a queue fills.
• Providers vary as much as technologies. During a download, one fiber provider measured 89 ms while one cable provider measured 47 ms. The technology sets the tendency, not the result for your address.

For scale, the FCC itself treats 100 ms or less as "low latency" in its satellite spectrum-sharing rules (cited in its 2026 Section 706 report). Idle, every terrestrial provider in the sample was far below that. During a full upload, four of the five cable providers and all four DSL providers went above it; none of the fiber providers did.

What the FCC data can't tell you

• It's from late 2022. The FCC notes that "some ISPs have since introduced active queue management (AQM) in their networks leading to improved results compared to the measurements from 2022." The report lists the FCC's contractor, SamKnows, as a program participant only through July 2023, and says data from August 1, 2023 onward isn't available. The newest folder in the FCC's MBA download area is still the 2023 one.
• It covers ten providers and the tiers most of their subscribers used. Gigabit tiers were measured but left out of the averages, so the results describe tiers from 1.5 Mbps DSL up to 800 Mbps cable.
• It measures the provider's network, not yours. The whitebox is wired to the router, and the FCC names in-home Wi-Fi as a possible bottleneck outside what it measures, noting that "some Wi-Fi connections may be limited to tens of Mbps."
• It measures to an exchange point, not a game server. The FCC says its metrics "do not necessarily replicate customer experience", because an application's own servers, content networks and routing differ.
• It doesn't chart jitter. The MBA's technical appendix describes a voice test that records jitter, but the Thirteenth report publishes no jitter results by technology.

Idle latency, latency under load and jitter

Three numbers decide how a connection feels in a game:

A provider can win on the first and lose badly on the second. A game sends small packets steadily, so a household upload (a cloud photo backup, a video call, a game clip) puts every one of those packets behind a queue. That's why the loaded columns in the FCC table matter more for a busy home than the idle one. Our latency and jitter explainer covers what each feels like in a match.

What the broadband label says about latency

Since April 2024 (October 2024 for providers with 100,000 or fewer subscriber lines), FCC rules have required a "broadband consumer label" for every home internet plan on sale, at the point of sale. The FCC's 2022 order (FCC 22-86) requires providers to show "typical upload and download speeds and typical latency". Its July 2026 order (FCC 26-48, adopted July 22 and published in the Federal Register on August 13, 2026) trims other parts of the label but keeps latency. It also makes latency one of the fields phone sales staff summarize aloud instead of reading the label word for word. That change to the label rule is delayed indefinitely: the FCC says it will announce the effective date in the Federal Register, and it hadn't by October 29, 2026.

The label doesn't standardize how latency is measured. Under the 2022 order, providers in the MBA program may use their MBA results; others may use the MBA method, "internal testing, consumer speed test data, or other data regarding network performance"; and mobile providers without such data may publish a "Typical Speed Range" of speeds and latency that most customers can expect. The FCC also declined to tie the reported figures to peak hours, and nothing on the label covers latency under load. These are the published figures for a few plans, checked October 29, 2026:

Read across, fiber's figures are the lowest, cable's and 5G's overlap, Starlink's range overlaps 5G's, and geostationary satellite is in a different league. But the figures aren't comparable to each other or to the FCC's. Verizon's fiber, for example, measured 10.4 ms idle in the FCC's 2022 data and 3.97 ms in Verizon's own 2026 testing: a different year, a different test point and a different method, none of which the label has to state. Treat a label as a quiet-line figure measured the provider's way, and test the loaded behavior yourself.

Each technology, for gaming

Fiber

The lowest idle latency in the FCC data (6.7 to 14 ms by provider) and the smallest penalty under load, with no fiber provider above 89 ms in either direction. The fiber tiers measured also had far higher advertised upload speeds than the cable and DSL tiers (the FCC charts them separately because of the "relatively high upload speeds for optical technology"), so a household upload is less likely to fill the line. If fiber reaches your address, it's the safest choice for low lag.

Cable

Idle latency close to fiber (11.9 to 23.9 ms), but in 2022 most cable providers built up large queues under load, especially on the upload. That weakness is the one SQM targets: shaping the line slightly below its speed in your router moves the queue somewhere it can be kept short. See how to test for bufferbloat before deciding a cable plan is the problem. Check the label's typical upload speed too, since that's the queue that filled.

DSL

The highest idle latency of the wired technologies (22.8 to 34.2 ms) and upload queues of up to 720 ms in the FCC data. Lower-speed DSL tiers were slowest: the FCC's chart data shows 61 ms idle on a 1.5 Mbps tier against about 20 ms on 100 Mbps DSL tiers. Choose it only when nothing wired is faster.

5G home internet

The published ranges (16 to 28 ms for T-Mobile's main plans, 35 to 70 ms for Verizon's) are workable for most games, but they come from the providers, not the FCC, and a cell is shared. T-Mobile lists "how many other customers are attempting to use the same spectrum resources" among the factors that affect performance, and Verizon says that on certain plans "we may prioritize your data behind other traffic". There's also a gaming-specific catch: T-Mobile says its gateways can't change NAT type or forward ports, and that settings some gaming systems need "cannot be configured on the T-Mobile gateway." See which providers support bridge mode and how carrier-grade NAT affects you.

Low-Earth-orbit satellite (Starlink)

Starlink's satellites are close enough that the radio path is short: its latency report puts the speed-of-light travel at "1.8-3.6ms per leg, and usually under 10ms for the round-trip", with the rest coming from ground routing, scheduling on the shared beam and buffers. Starlink reports a US median peak-hour latency of 25.7 ms as of June 2025, with fewer than 1% of measurements above 55 ms, measured from its router to the internet. Starlink also says it added fq_codel queue management to its Wi-Fi router, so that one person's big download doesn't affect another's game. These are Starlink's own measurements, not the FCC's. Two other things matter for gaming: the default plan uses carrier-grade NAT (a public IPv4 address is an option only on Priority plans, and Starlink's own Wi-Fi router doesn't support port forwarding, so opening ports also needs a third-party router), and the dish needs a clear line of sight to the satellites, as an FCC commissioner's statement in the 2026 Section 706 report notes.

Geostationary satellite (Hughesnet, Viasat)

Physics sets the floor. A geostationary satellite orbits about 36,000 km above the surface, by NASA's figure. The ITU's planning value for the path through space between two earth stations at that altitude is 260 ms one way, so a game packet and its reply, each going up and down once, spend about 520 ms in space before any network or processing delay (arithmetic). As a cross-check, 4 × 36,000 km at the speed of light (299,792 km per second) is about 480 ms with the satellite straight overhead. The FCC measured 594 to 612 ms in 2017, and Hughesnet's own support page says satellite adds "about a half-second delay". Hughesnet points customers who need real-time multiplayer games to its Fusion plans, which add a low-latency wireless path. For fast online games, treat standard geostationary service as a last resort.

What matters as much as the technology

• The server's distance. The ITU's planning value for optical fiber is 5 microseconds per kilometer one way, so every 1,000 km of cable path adds about 10 ms to a round trip (arithmetic). Picking the nearest server region in the game often saves more than switching providers.
• Your Wi-Fi. The FCC measures over a cable for a reason. A wireless hop shares airtime with every other device and resends lost frames; see Ethernet vs Wi-Fi for gaming for what published studies measured.
• Your router's queue. SQM can turn the FCC's "during an upload" column into something close to the idle column on your own line, at the cost of a little speed; it can't fix congestion inside the provider's network or the distance. What no home router can fix sets out the limits.
• The address translation. Two routers in a row, or a provider's carrier-grade NAT, can cause strict NAT types and break voice chat or hosting. That's a different problem from lag; see double NAT symptoms.

How to choose

Two common mistakes. First, buying speed for ping. The FCC says higher tiers generally have lower idle latency, "as it takes less time to transmit each packet", but in its tier data each cable or fiber provider's tiers were within about 5 ms of each other, and at Cox and Mediacom the faster of the two tiers measured had the higher idle latency; only the slowest DSL tiers stood out. A gigabit plan speeds up game downloads, not shots. Second, judging a provider by a quiet-line speed test. After installation, run a test that measures latency under load, as the test steps in our latency guide describe, from a wired computer. If the loaded figure is far above the idle one, turn on SQM before blaming the line. Choosing hardware next? Our gaming router comparison lists which models document SQM, and the gaming topic guide collects the rest.