Fiber Speed Test: Getting the Gigabit You Pay For (GPON vs XGS-PON, ONT, Wi-Fi 6)
How to test gigabit fiber correctly: the 940 Mbps ceiling explained, plant types exposed by upload, the full chain test from ONT to Wi-Fi 6, and peak-hour checks.
Quick answer: Fiber changes the rules: symmetric upload, flat latency, and gigabit plans that test within 5% of promise — but only if your ONT, router and Wi-Fi chain can keep up. Testing gigabit is its own skill; at 1000 Mbps, every weak link in your home shows up in the number.
Why fiber tests differently
- Symmetric by design: XGS-PON and GPON plants dedicate upstream capacity — upload tests near download tests is the fiber signature (see the upload guide).
- Flat latency: idle ping 1-8 ms in metro areas; fiber’s bandwidth-delay profile makes bufferbloat rarer (but not impossible on cheap routers).
- Consistency at peak: no shared coax contention — evening runs should match morning runs within a few percent.
GPON vs XGS-PON: the plan-vs-plant truth
| Plant | Typical ceiling | What it means for you |
|---|---|---|
| GPON | ~1-2.4 Gbps down, 1-1.2 up shared split | fine for 500-1000 plans; oversubscribed splits can sag at peak |
| XGS-PON | ~10 Gbps symmetric | the real gigabit+ tier; 2000-8000 plans ride this |
| "Fiber to the cabinet" (FTTC/FTTB) | last meters on copper | 400-900 down but copper upload and latency — not true fiber; test upload to expose it |
Marketing blurs these: "gigabit fiber" on GPON oversubscribed 32-way can test at 600 at peak. The test protocol below separates plant from plan.
The gigabit testing chain (every link must pass)
- Device: a laptop’s 2015 Wi-Fi tops at ~400 Mbps real-world. Test wired with a gigabit+ ethernet port — USB-C 2.5G adapters are cheap.
- Router: must have gigabit WAN/LAN ports (not Fast Ethernet) and a CPU that NATs 1 Gbps without sweating — older routers cap at 300-600.
- Wi-Fi: Wi-Fi 5 tops near 400-700 real-world; Wi-Fi 6/6E reaches 900+ at close range. See the Wi-Fi guide.
- The test: Toolfyra Internet Speed Test with parallel ramped streams and sustained-window averaging — at gigabit, single-shot tests under-report badly; ramping matters (see methodology).
- Repeat at peak hour: fiber should barely move; a 20%+ evening drop points to oversubscription, not your gear.
Reading gigabit results
- 940-980 on wired: textbook — protocol overhead makes 1000 exactly unreachable; anything above ~940 is perfection.
- 700-900 wired: check router NAT throughput and ONT provisioning; some "1 Gbps" plans provision 850-940 at the ONT.
- 500 wired flat all day: wrong port (100 Mbps LAN), old CAT5 cable (not 5e/6), or a 100 Mbps ONT handoff — the cable audit finds it.
- Symmetric 900/900: enjoy — that is XGS-PON doing its thing; verify upload with the upload guide.
After the test: using gigabit well
- Wi-Fi 6/6E router or the gigabit is a wired-only party.
- 2.5G ports on the router if your plan is 2 Gbps+.
- SQM still worth it: even fiber benefits from fq_codel during multi-user saturation (bufferbloat).
- Server-side ceilings: single websites rarely serve at gigabit — do not diagnose your line by one CDN’s mood; use multi-source protocol.
FAQ
Why do I get 940 and not 1000? Ethernet/IP overhead: ~6% of gigabit is protocol tax. 940-980 wired is a perfect result.
Is gigabit worth it over 300 Mbps? For most households, no perceptible daily difference; for creators, 5+ heavy users, or symmetric upload needs — absolutely.
Does fiber lower ping? Usually yes: metro fiber pings 1-8 ms and holds flat under load, and upload symmetry kills the classic upstream-bloat freeze.