Paying for Gigabit, Getting 50 Mbps: Where the Speed Goes

You pay for 1,000 Mbps. Your phone in the back bedroom reports 48. Somewhere between the street and your hand, most of what you bought went missing, and the usual advice — reboot the router, buy a faster router, buy an extender — is a guess at which stage lost it.

There are six stages, each with its own ceiling. Only one of them is the reason you see 50 Mbps, and twenty minutes of testing tells you which.

Throughput is a chain, not an average#

Every packet passes through six ceilings in a row, and your result is the lowest one — not the average, not the sum. A gigabit plan behind a 100 Mbps port is a 94 Mbps connection, and no other upgrade in the chain moves that number.

In most homes the missing speed is in stages 4 through 6. Work down the list anyway: the first three are quick, and finding a 100 Mbps port in five minutes saves you a weekend.

Stages 1 to 3: everything before the antenna#

Start with the number you bought, up and down — asymmetric shapes like 1000/50 are normal, and upload is what calls and backups live on.

Then plug a laptop into a LAN port, turn its Wi-Fi off, and run a speed test twice. That result is the best your house can do, and it collapses stages 1 through 3 into one number. Close to your plan, and every remaining suspect is wireless. Well under it, and the full wired-versus-wireless procedure is in how to tell the ISP from your Wi-Fi.

Three things quietly cap the wired path:

• A 100 Mbps port anywhere in the chain. Older laptops, cheap USB adapters, forgotten five-port switches, some gateway ports. The symptom is unmistakable: results pinned near 94 Mbps whatever you change.
• A patch cable that is not what you think. Cat5e and above carry gigabit; the category is printed along the jacket. A crushed cable often still links, at a lower rate.
• The modem or ONT generation. An older cable modem with few bonded channels cannot reach gigabit however good the line is, and plans above 1 Gbps need a 2.5 GbE WAN port.

The router is stage 3, and the box art will not warn you about it. Names like AX3000 or AX5400 add up theoretical radio rates and say nothing about how fast the box moves traffic between the internet and your LAN. Basic routing is hardware-accelerated on most modern routers, which is why they hit gigabit out of the box. Switch on smart queue management, a VPN client, parental filtering or security scanning and many models drop off that fast path onto the CPU, landing in the low hundreds of Mbps. A VPN client is the heaviest of them.

So test by comparison, not by number: run the wired test, turn the suspect feature off, run it again. If the number jumps, you have priced that feature. A newer router raises this ceiling and stage 4's and does nothing for stage 6, which is why a new router so often changes nothing.

Stage 4: the Wi-Fi link#

Wi-Fi never delivers its negotiated link rate. The medium is half-duplex and shared, so usable throughput lands at roughly half the rate your device reports — the mechanics belong to why full bars still buffer. What matters here is the ceiling before signal strength enters the picture.

Treat those as rules of thumb. Two things explain most complaints.

A device that fell back to 2.4 GHz, or only speaks it, cannot beat roughly 60-90 Mbps whatever you pay per month. If your phone reports 50 Mbps in a far room, check the band before anything else.

Channel width is the second. Dropping from 80 MHz to 40 MHz halves the ceiling — often worth it in a dense building where the narrower channel is cleaner, as choosing a channel in a crowded building explains. On 2.4 GHz, stay at 20 MHz and use channels 1, 6 or 11.

Stages 5 and 6: your device, and the room it is standing in#

The radio inside the client is a ceiling you cannot configure away. Flagship phones and laptops are usually two-stream; budget phones, tablets, smart TVs, streaming sticks and most smart home gear are commonly single-stream, which halves everything in the table above. An old Wi-Fi 4 laptop tops out near 100 Mbps in perfect conditions.

No phone will validate a gigabit plan; 900 Mbps takes a cable and a machine that can receive it. Test two devices in the same spot: when one is three times faster, you are looking at stage 5, not at your network.

Stage 6 is the room. Signal level converts into throughput — a weak link negotiates a slower rate, then loses part of that to retransmissions.

-67 dBm is the practical floor for reliable video and voice, not the floor for speed — what each dBm band actually means covers the scale itself. Speed starts falling long before that floor. Getting there is geometry: an ordinary interior wall costs 3-8 dB at 5 GHz, brick or block 10-16 dB, reinforced concrete 15-25 dB. The material-by-material breakdown has the rest.

The ladder: six tests in twenty minutes#

Run these in order and stop at the first stage that fails. Use the same speed test and the same server throughout.

1. Write down the plan you pay for, up and down.
2. Wired at the router, laptop Wi-Fi off, twice, keeping the better result.
3. Read the WAN link speed on the router's status page. 1000 Mbps full duplex, or you have found it.
4. Turn off QoS, VPN and security features, then repeat the wired test.
5. Wi-Fi in the same room as the router, on your best device, noting which band it is on.
6. Wi-Fi in the problem room, same device, same test, then again on a second device.

Then read where the number first falls off.

When nothing is actually broken#

Only the last row of that table describes a problem an extender or a mesh node can fix. Stop anywhere earlier and more radios change nothing — an extender repeats the coverage it receives; it cannot create bandwidth that never arrived.

Then check whether the number costs you anything. A 4K stream needs 15-25 Mbps. An HD video call needs 1.5-3 Mbps each way and cares more about latency and jitter than Mbps. Cloud gaming wants 25-35 Mbps with a stable ping. A steady 50 Mbps runs all of that at once, and the honest advice then is to stop running speed tests.

The number matters when it misses the bar: large uploads, a NAS, several 4K streams, a workstation moving big files. There the fix is a wire, or a node on wired backhaul — not a repeater. A single-radio extender roughly halves the throughput of everything behind it, and even a good dual-band unit with a reserved backhaul band costs 10-30% per hop. If the failure is confined to one place, what to do when one room is slow picks up where this ladder ends.

Settling the last stage on a floor plan#

Stage 6 is the only one you cannot test from a single spot. Range Up does it on a floor plan instead: sketch your rooms, pick the material preset that matches the build — wood frame, brick or concrete — put the router where it actually stands, and read a predicted room-by-room coverage map against the same dBm windows used above.

If the map shows the failing room comfortably covered, stage 6 is eliminated and your answer sits in stages 1 through 5. If it shows a real hole, the automatic search ranks hundreds of candidate positions for a second unit by how much each lifts that room, and compares coverage before and after for the winner. It simulates your layout rather than measuring your live network — the six tests are still yours to run — but it settles the geometry before you spend anything. The placement walkthrough shows what that looks like; the app itself runs on-device, with no account.