[USG Flex 500H + NWA130BE] - Different speedtest value between LAN and Wifi

Options
2

All Replies

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    Options

    Hi @Zyxel_Judy,

    Yes, I've security service applied to LAN_to_WAN policy, but I've also disabled it (all the security service). The speed are invariant between both enabled or disabled.

    The EEE/Green Ethernet was disabled, but also in this case, the speed are invariant between enabled or disabled.
    I've set:

    • Advanced EEE: Disabled
    • EEE Max Support Speed: 2.5Gbit Full Duplex
    • EEE: Disabled
    • Green Ethernet: Disabled
    • Idle Power Saving: Disabled
    • Jumbo Frame: Disabled

    But also in this case, the speed don't change

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    Options

    Hello @PeterUK, @Zyxel_Judy
    I've tried the same USB-A Adapter than yesterday (Realtek Gaming 2.5GbE Family Controller) on a Macbook Pro 2017, with same cable as previous tests.

    Without bridge between LAN_P2 (192.168.2.1/28) and LAN_P3_P12 (192.168.1.1/24):

    • Speedtest on Port P2: more or less 1500/1700Mbit (start at 1800 then slow down at 1500/1600)
    • Speedtest on Port P4: more or less 2000/2200Mbit

    With bridge (P2 and P3_P12 are 192.168.1.1/24):

    • Speedtest on Port P2: more or less 1400/1500Mbit
    • Speedtest on Port P4: more or less 1900/2000Mbit

    So:

    • There is something that affect the port P2, because with or without the bridge, the maximum speed are slowest than the other ports;
    • There are differences between with or without the bridge at port parity (but this is expected);

    Probably now the adapter reach over 2000Mbit because on windows can there is a driver problem (I don't know why).
    Anyway, the maximum speed 2500Mbit is not reachable in this way.

  • PeterUK
    PeterUK Posts: 4,738
    250 Answers 2500 Comments Friend Collector Ninth Anniversary
    image  Guru Member
    edited August 25
    Options

    because your still testing through FLEX you need a baseline test its unclear if the Adapter is a limiting factor such as USB speed

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    Options

    Hi Peter,
    OK, I've done the tests.

    I used iperf3; on PC I have the docking station attached via USB (Realtek USB 2.5GbE Family Controller) and on Macbook I have the USB-A adapter Realtek Gaming 2.5GbE Family Controller.

    This on PC (first as client, second as server):

    image.png

    And this is on MacBook (first as server, second as client):

    image.png

    So the bitrate seems up to 2Gbit more or less…

    For now, I don't have any other external USB-Ethernet adapter; I've get online an Sabrent 5GbE chipset Realtek, tomorrow I take it and retest.

  • PeterUK
    PeterUK Posts: 4,738
    250 Answers 2500 Comments Friend Collector Ninth Anniversary
    image  Guru Member
    edited August 25
    Options

    But the USB port ON the computer you plug that adapter in could be the limiting thing. its why a PCI-E card could be better depending on the card lanes and PCI-E version plus what the PC supports

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    Options

    Hi Perer,

    I have only notebook without PCI-E card.

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    edited August 26
    Options

    Hi Peter,

    after I receive the Sabrent NT-C5GA USB-C adapter, the speedtest between PCs are quite different:

    image.png

    So seems that the internal ethernet on docking station are slowly than the 2.5Gb

    Then this is the speedtest:

    • Without bridge on Port P4:
    image.png
    • Without bridge on Port P2:
    image.png
    • With bridge on Port P4:
    image.png
    • With bridge on Port P2:
    image.png

    Same hardware, same configuration.
    PC ←→ PC the speed are more or less 2.2/2.3Gb
    WAN ←→ LAN the speed are less than PC-PC with significant decrease on bridged mode

    Anyway, in any case:

    • The P2 seems more efficient without bridge, probably because the WAN (P1) and the LAN (P2) work on the same ASIC and this not affect the packet transmission between ASICs and CPUs
    • The P4 it's more efficient with bridge, probabily because the packet transmission is executed splitting the ASICs chipset.

    Anyway…. the pure 2.5Gbit seems to be a dream and mirage

  • Zyxel_Judy
    Zyxel_Judy Posts: 2,753
    Zyxel Certified Network Engineer Level 2 - Nebula Zyxel Certified Network Engineer Level 2 - Switch Zyxel Certified Network Engineer Level 2 - Security Zyxel Certified Network Engineer Level 1 - Nebula
    image  Zyxel Employee
    Options

    Hi @Maverick87 ,

    Thank you for the extensive testing and data you've shared — it's been very helpful for our internal review.

    We've run our own internal test using a similar setup, and our results are actually very close to what you're observing. It looks like the range you're getting is close to the practical achievable throughput for this setup. Factors like USB adapter/controller limitations (as you found with the docking station's internal adapter vs. the Sabrent NT-C5GA), and network protocol overhead, all contribute to the gap from the theoretical 2.5Gb link speed.

    Zyxel_Judy

  • Maverick87
    Maverick87 Posts: 447
    Zyxel Certified Network Administrator - WLAN Zyxel Certified Network Administrator - Nebula Zyxel Certified Network Administrator - Security Zyxel Certified Sales Associate
    image  Master Member
    Options

    Hi @Zyxel_Judy,

    "nice to hear this", if nothing else the tests are similar :)

    Obviously, there is always an adapter with own driver (USB-A/USB-C/Thunderbolt/PCI-Express), it's always an adapter internal or external to the PC. Also, there is always a driver that "translate" the information into human-readable. So can be normal.

    But the strage is, that is not close to 2.5Gb but only 1.9/2.0Gb and the very strange result are the difference between port P2 and P4 with same configuration.
    Without bridge, the results are better, but seems there are present a difference both P2 and P4 of the 500H; so not only with/without bridge but also different port cause different results.

  • Zyxel_Judy
    Zyxel_Judy Posts: 2,753
    Zyxel Certified Network Engineer Level 2 - Nebula Zyxel Certified Network Engineer Level 2 - Switch Zyxel Certified Network Engineer Level 2 - Security Zyxel Certified Network Engineer Level 1 - Nebula
    image  Zyxel Employee
    edited September 9
    Options

    Hi @Maverick87 ,

    As our P3/P4 iperf result (~2.35Gbps, matching P1/P2) confirms the physical ports fully support 2.5G, so this isn't a port issue.

    As your test, a external speedtest to WAN has to pass through routing, NAT, and the security inspection engine. To help narrow down the speed difference between P2 and P4, could you check two things on your side:
    - Are P2 and P4 currently bound to the same LAN-to-WAN security policy and profile? If they sit in different zones, one could still have an inspection service applied that the other doesn't, even with most services disabled.
    - Re-run the external speedtest on both P2 and P4, to the same speedtest server, in the same short time window, with no other devices (especially WiFi clients on the NWA130BE) active on the network.

    Moreover, there are internet-side variance (server load, ISP routing, etc.), which is common with external speedtest compared to a controlled iperf test.

    Zyxel_Judy

Categories

Security Help Center

EnglishTiếng ViệtРусскийไทย中文（台灣）