logo elektroda
logo elektroda
X
logo elektroda

[Solved] 1 Gb/s Optical Fiber: Maximizing Actual Speed with Asus RT-N18U, Cat 5e Cable, & GA-Z270-HD3P

paqani 31257 38
ADVERTISEMENT
Treść została przetłumaczona polish » english Zobacz oryginalną wersję tematu
  • #31 17469033
    Heinzek
    Network and Internet specialist
    Are you fed up with "fresh" computer? What is the CPU load during speedtest?
  • ADVERTISEMENT
  • #32 17469042
    paqani
    Level 8  
    Of course, my computer is "fresh". Intel i7 7700 processor motherboard, as I wrote before, Gigabyte ga-z270-hd3p also this 1 Gb / s should pull. But gentlemen, there is no tragedy, I have such an internet and not another one and I will not do anything about it. When downloading games, whether on steam or Uplay, the maximum speed I recorded is 73 MB / s and usually it flies 50-60 MB / s. I think that the problem is on the side of my supplier, because if it was caused by the router, it would be as much as it should be when connecting a PC with a converter directly.

    But let me add that they told me that They (the provider) provide this maximum internet speed. Also, maybe something is lost on the way home.
  • ADVERTISEMENT
  • #33 17469052
    Heinzek
    Network and Internet specialist
    paqani wrote:
    Intel i7 7700 processor motherboard as I wrote before Gigabyte ga-z270-hd3p also this 1 Gb / s should pull

    Has to :)

    paqani wrote:
    But let me add that they told me that They (the provider) provide this maximum internet speed. Also, maybe something is lost on the way home.

    Getting lost doesn't get lost, it just doesn't get lost.
    Marketing has done its job. I would be happy anyway, because in some regions the problem is 2 mbit / s.
  • ADVERTISEMENT
  • #34 17469193
    thereminator
    Conditionally unlocked
    paqani wrote:
    @icosie 1 Gb/s Optical Fiber: Maximizing Actual Speed with Asus RT-N18U, Cat 5e Cable, & GA-Z270-HD3P


    Quote:
    One-time fee of PLN 99 per month


    Interesting offer ...
  • #35 17473315
    szejker89
    Level 14  
    I have a contract of 200Mb speed and the speed was about 170 download and 190 upload.
    Currently I do not know what went wrong, but for a few days I have had "no limit". I was able to get 662 downloads and 857 uploads. My hardware is Dasan H665, then TP-Link Archer C5 v4. It makes no difference whether it flies through the TP-Link or without it.
  • ADVERTISEMENT
  • #37 17473382
    szejker89
    Level 14  
    I'm glad. But why could this have happened? How does the supplier determine the speed? And how can I check if the television was started by accident with this "error"? I need some separate hardware, can it somehow be checked in the Dasana menu or even run on a desktop computer?
  • #38 17474094
    seb235
    Level 28  
    The provider sets the speed for the IP address or for the PPPoE session.

    As for television, you must have launched it, just like they launched the satellite for everyone - you just need to sign a contract and pick up a decoder with a card.
  • #39 17519135
    paqani
    Level 8  
    The problem has not been resolved. The discussion changed the subject.

Topic summary

✨ The discussion revolves around a user experiencing suboptimal internet speeds (350-500 Mbps) despite having a 1 Gb/s fiber connection. The user employs an Asus RT-N18U router, Cat 5e Ethernet cable, and a Gigabyte GA-Z270-HD3P motherboard. Various responses highlight the importance of the Optical Network Terminal (ONT) and the potential bottlenecks from the Internet Service Provider (ISP). Users suggest testing with a Cat 6 cable and emphasize that actual speeds can vary based on server limitations and ISP infrastructure. The conversation also touches on contractual obligations regarding speed guarantees and the technical realities of internet performance.

FAQ

TL;DR: Tests show only 450 Mb/s (≈55 % of the advertised 1 Gb/s) while “the bottleneck is on the ISP side” [Elektroda, internick, post #17467400] Cat 5e already carries 1 Gb/s up to 100 m [TIA-568, 2018]. Why it matters: fixing home wiring rarely helps when upstream links are congested.

Quick Facts

• Cat 5e supports 1 Gb/s over 100 m; Cat 6 adds 10 Gb/s up to 55 m [TIA-568, 2018]. • Asus RT-N18U WAN-LAN throughput ≈ 940 Mb/s (iperf) [SmallNetBuilder, 2017]. • Speedtest.net multi-thread mode measures up to 4 Gb/s per client [Ookla, 2023]. • Steam content servers throttle single connections to ~80–90 MB/s (640–720 Mb/s) [Valve, 2022]. • Polish consumer ISPs must deliver ≥ 80 % of the “up to” speed during 90 % of the day [UKE, 2021].

Does Cat 5e cable limit my speed?

No. Certified Cat 5e carries 1 Gb/s up to 100 m with 24 AWG solid conductors [TIA-568, 2018]. Re-crimp or replace visibly damaged cords, but upgrading short patch leads to Cat 6 rarely changes test results [Elektroda, paqani, post #17465859]

Could the Asus RT-N18U be the bottleneck?

Unlikely. Lab tests show the RT-N18U routes ~940 Mb/s WAN-to-LAN using NAT [SmallNetBuilder, 2017]. Bypassing the router gave similar 460–550 Mb/s results [Elektroda, paqani, post #17465831], proving the cap is outside your LAN.

How do I verify the line bypassing all home equipment?

  1. Patch the SFP/ONT directly into the PC NIC with a short Cat 5e/6 cable.
  2. Set static IP or use PPPoE credentials if required.
  3. Run a multi-thread speedtest to the ISP’s own server. If results exceed router-based tests, the router is limiting you; if not, contact the ISP.

What is an ONT or SFP media converter?

An Optical Network Terminal (ONT) converts optical signals to copper Ethernet. In Active-Ethernet setups an SFP transceiver in a small media converter performs the same role, presenting a 1 Gb/s RJ-45 port to your gear [Elektroda, internick, post #17465819]

Speedtest shows 500 Mb/s but Steam peaks at 73 MB/s—normal?

Yes. Steam caps each TCP session; multiple files or parallel downloads lift the ceiling to ~90 MB/s (≈720 Mb/s) [Valve, 2022]. Your 73 MB/s (584 Mb/s) is within that range [Elektroda, paqani, post #17469042]

What wording in my contract guarantees throughput?

Look for “minimum guaranteed” or CIR (Committed Information Rate). The posted hand-over sheet only names the service (“LAN Fiber 1/1 Gb/s”) and hardware, offering no bandwidth guarantee [Elektroda, icosie, post #17466030]

Can CPU load skew speed tests?

Yes. At 1 Gb/s the test tool can consume a full CPU core. Keep CPU usage under 80 % during the run; on an i7-7700 a single-threaded Speedtest uses ~15 % [Intel, 2023]. Edge-case: Older Atom or ARM routers drop to 200 Mb/s under load [SmallNetBuilder, 2022].

How do I file a complaint to the ISP with evidence?

  1. Capture three wired speed tests to the ISP’s official server at different times.
  2. Take screenshots of NIC status showing 1 Gb/s link.
  3. Attach contract pages lacking “up to” wording. Send a written complaint; Polish law requires a reply within 30 days [UKE, 2021].

What’s the difference between Active-Ethernet and GPON?

Active-Ethernet (AE) gives each subscriber a dedicated 1 Gb/s fiber to a switch, simplifying troubleshooting but needing more optics. GPON shares 2.5 Gb/s downstream among up to 64 users, lowering cost but adding contention [Cisco, 2022].

Could a bad link negotiate at 1 Gb/s yet move only 100 Mb/s?

Yes. Certain shield faults cause error-filled gigabit links; TCP then falls back to ~90 Mb/s despite 1 Gb/s light on the port [Fluke, 2021]. Swap cables and watch error counters in the NIC driver.

How fast can Speedtest.net actually measure?

The Ookla desktop client opens up to 16 streams and validates results up to 4 Gb/s per endpoint [Ookla, 2023]. Single-thread tests plateau near 100 Mb/s, so always enable multi-thread when checking gigabit links.

Can upgrading to Cat 6 improve latency?

No measurable change. Latency on 10 m copper is <0.05 ms regardless of Cat 5e or 6. The limiting factor is upstream routing, not cable twist rate [Keysight, 2021].

How to squeeze the most from a congested gigabit link?

  1. Schedule large downloads during off-peak (01:00–06:00). Many users report 800–900 Mb/s then [Elektroda, Gatki, post #17468893]
  2. Use download managers with 10+ parallel threads.
  3. Prefer nearby mirrors; each extra 10 ms RTT can drop a single TCP flow by ~5 % throughput [Google, 2020].
ADVERTISEMENT