Where the Gap Comes From
| Factor | Why It Lowers Real Speed | Best Baseline Test |
|---|---|---|
| Wi-Fi overhead and contention | Wireless is a shared medium with retries and signal loss | Ethernet test on same device |
| Router or modem limit | Ports, CPU, firmware, or old standards cap throughput | Check equipment specs against plan speed |
| Protocol overhead | TCP/IP and Ethernet headers consume part of the link | Expect 2–5% less than raw line rate |
| Shared infrastructure | Cable nodes and DSL line cards serve many customers | Compare morning vs evening results |
| Distance from DSLAM or node | Signal attenuation raises errors and lowers sync rate | Check modem sync speed in admin page |
| ISP throttling | Specific services or general congestion limits traffic | Compare speeds across multiple services |
What "Up To" Really Means
Every residential internet plan carries the phrase "up to" before the speed figure. This is not purely marketing language — it reflects a genuine engineering reality. ISPs provision their networks assuming not all customers use their full speed simultaneously. The "up to" rate is what the access line can carry under ideal conditions: good signal, no congestion, capable equipment, and a nearby fast server. All those conditions rarely align at once in a real home.
Regulatory bodies in several countries have pushed ISPs to provide minimum guaranteed speeds or typical speeds in addition to the maximum headline figure. If your ISP advertises a "typical evening speed" or a minimum guaranteed rate, that is the more meaningful number to test against during peak hours.
Protocol Overhead: Why Raw Line Rate Is Never the Full Story
Every TCP/IP packet carries headers that are not payload. A standard IPv4 TCP segment has a 20-byte IP header and a 20-byte TCP header. Ethernet adds a 14-byte frame header and a 4-byte CRC. On a 1 Gbps line, this overhead reduces usable throughput by roughly 2–5% depending on packet size. Large transfers using big packets see less overhead than small-packet traffic like VoIP or gaming.
On cable DOCSIS connections, additional framing and error correction further reduce what the raw line rate translates to in application throughput. A 1 Gbps DOCSIS plan reliably delivering 940–960 Mbps in a clean wired test is behaving correctly, not underperforming.
Wi-Fi Overhead and Contention
Wi-Fi cuts measured throughput far more than protocol headers do. Wireless is a half-duplex shared medium: only one device transmits at a time on a given channel. Every device in range — including neighbors on overlapping channels — competes for airtime. Retransmissions, guard intervals, preambles, and management frames all consume airtime that could otherwise carry data.
Wi-Fi 5 (802.11ac) at 80 MHz channel width and reasonable signal may deliver 300–500 Mbps of real throughput on a good day. Wi-Fi 6 (802.11ax) improves efficiency with OFDMA and better multi-device handling, but the wireless medium still cannot deliver what an Ethernet cable does. Testing on Wi-Fi and blaming the ISP for a slow result is one of the most common misdiagnoses in home networking.
Shared Infrastructure: Cable Nodes and DSLAM Line Cards
Cable internet uses a hybrid fiber-coaxial (HFC) architecture. Fiber runs from the ISP's headend to neighborhood nodes, and then coaxial cable distributes the signal to homes. All customers sharing a node share its available capacity. A lightly loaded node during off-peak hours may deliver near plan speed to every subscriber simultaneously. A congested node at 9 pm, with dozens of households streaming 4K video, will show noticeably lower speeds for everyone on it.
DSL uses the telephone network's copper wire pairs. The speed a DSL customer receives depends heavily on the distance between their home and the DSLAM (Digital Subscriber Line Access Multiplexer) at the telephone exchange. VDSL2 can deliver over 100 Mbps within 300 meters of the DSLAM, but drops to 40–50 Mbps at 1 km and much lower at 2 km. Customers at the far end of a long copper run will never see speeds close to what the ISP advertises for the plan, regardless of how good their equipment is.
ISP Throttling: Service-Specific vs General Congestion
There are two types of reduced performance that look like throttling but are different in cause. General congestion is unintentional: the ISP simply does not have enough capacity on a segment at peak times to serve all customers at full speed. Service-specific throttling is intentional: the ISP identifies traffic to certain services — video streaming, peer-to-peer, or specific ports — and rate-limits it regardless of available capacity.
You can test for service-specific throttling by comparing speed test results (which use HTTP/TCP on standard ports) against the same content accessed through a VPN. If speeds improve dramatically through the VPN, traffic shaping on specific services is a likely explanation. If speeds are similarly low with or without a VPN, general congestion or a plan limitation is more likely.
How to Test Fairly
To get a result that fairly reflects your ISP plan rather than local variables:
- Use a wired Ethernet connection directly to the modem or router, not Wi-Fi.
- Test from a single device with no other background traffic on the network.
- Close browser tabs, cloud sync apps, and anything that uses the network.
- Run tests at off-peak hours (early morning) and again at peak hours (evening) to identify congestion.
- Use a server geographically close to you on a well-known test platform.
- Run at least three tests and take the median result.
- Check your modem or router admin page for the sync speed or line rate — if the sync rate is already below the plan speed, the problem starts at the access line or DSLAM, not your router.
What to Do If Speed Consistently Under-Delivers
If repeated wired tests at off-peak hours, using good equipment and a nearby server, consistently show speeds far below your plan rate, the problem is worth escalating. Before contacting your ISP:
- Document results with timestamps, server used, and the device tested.
- Record the modem sync speed from the admin interface.
- Run tests across multiple days to show a pattern, not just one bad result.
- Note whether the problem is constant or only appears at certain hours.
When you contact the ISP, share the documented evidence. Intermittent issues at peak times are often handled by network upgrades and may require reporting from multiple customers before action is taken. Persistent low speeds at all hours, especially when the modem sync rate is below plan speed, typically indicate a line fault or misconfiguration that a technician can investigate.
Frequently Asked Questions
Should I get exactly my advertised speed?
Not usually, and almost never on Wi-Fi. A wired test at off-peak hours landing within 10–15% of the plan rate is generally acceptable, accounting for protocol overhead and normal provisioning variation. If wired results are consistently 50% or more below the plan rate, that is worth investigating.
Why is upload much lower than advertised download?
Most cable, DSL, fixed wireless, and 5G home internet plans are asymmetric by design. The headline figure on the plan is download speed. Upload has a separate, much lower rate — sometimes 10–20% of the download speed on cable plans. Fiber plans (GPON, EPON) are more commonly symmetric.
When should I complain to my ISP?
When repeated wired tests with good equipment, at multiple times of day, consistently show speeds far below the plan rate. Bring timestamps and results. The ISP can check line quality, modem provisioning, and node load from their side, but they need documented evidence to justify a truck roll or infrastructure change.