Messages

 Hi All, I'm new to OPNSense, I got it setup last week and I'm curious as to why I'm not seeing a full 2GB connection when running a speedtest. I havent done anything fancy on my setup just yet, other than setup some shaper rules to help with Bufferbloat which isn't having the huge impact I was hoping. I followed this guide: https://docs.opnsense.org/manual/how-tos/shaper_bufferbloat.html I can fiddle more with that later.

However, my main question is why don't I see anything near 2Gb of speed when running a speedtest. My current setup is as follows:

    Dell Optiplex 5055 Ryzen 7 Pro 1700

    32GB RAM

    512GB SSD

    Intel X540-T2 10GbE Dual Port Adapter

    Zyxel XMG1915-10E 8-port switch (2x10Gb SFP's)

    Cable Internet - 2Gb/100Mb - Arris SurfBoard S34 Modem

I have the 2.5Gb port in the modem connected to the 10Gb WAN port in the Intel card, and then the other 10Gb port LAN port connected to one of the 10Gb SFP's on the Zyxel switch. I know my laptop is only connected with a 1Gb connection, but if I run the speedtest from plugin in OPNsense, it still only comes back with a max of around 900MBps.

If I SSH into the router and run 'ifconfig' I notice that the WAN port is showing 1000baseT full duplex. I set the drop down to 10GbaseT in the OPNsense config and it still just shows 1000baseT. If I'm only getting 1Gbps speed should I just contact my ISP and downgrade from the 2Gb to the 1Gb plan? I was really hoping to get that extra speed since I work from home.

Code Select Expand

ix0: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
description: WAN (wan)
options=4803828<VLAN_MTU,JUMBO_MTU,WOL_UCAST,WOL_MCAST,WOL_MAGIC,HWSTATS,MEXTPG>
[Redacted IP and MAC Information]
media: Ethernet 10Gbase-T (1000baseT <full-duplex>)
status: active
nd6 options=23<PERFORMNUD,ACCEPT_RTADV,AUTO_LINKLOCAL>
ix1: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
description: LAN (lan)
options=4803828<VLAN_MTU,JUMBO_MTU,WOL_UCAST,WOL_MCAST,WOL_MAGIC,HWSTATS,MEXTPG>
[Redacted IP and MAC Information]
media: Ethernet autoselect (10Gbase-T <full-duplex>)
status: active
nd6 options=21<PERFORMNUD,AUTO_LINKLOCAL>

[No Shaper rules setup, just base install of OPNSense]

I've read through several articles on this forum and across the interwebs on how to setup the FQ-CoDel stuff to relieve my network of bufferbloat but I still seem to be running into issues.

Previously to setting up OPNSense and just going through my Netgear R7000 router (with Fresh Tomato) I was getting a "D" to "F" report when running the bufferbloat test (https://www.waveform.com/tools/bufferbloat).  I just setup OPNSense this last week with the hardware in my signature. 

I haven't done any tweaking to it other than following this guide (https://docs.opnsense.org/manual/how-tos/shaper_bufferbloat.html).  I have 2Gb down/100 Mb Up cable from Cox Cable and an Arris S34 modem.

I seem to be getting some mixed results that I'm curious about.  I have 2 macbooks pro's (1 has an Intel chip (Work laptop), the other has the M2 chip (Personal Laptop)).
The network is fairly simple right now and the setup is as follows.

Cable into the Arris S43 modem.  Modem's 2.5 Gbs port connected to the 10Gbps WAN port on the OPNSense box with Cat8 cable.  10Gbps LAN port connected to the SPF port (with RJ-45 copper adapter) on the TP-Link switch with Cat8 cable.  Netgear Nighthawk RAX120 WAP connected to 1Gb port on TP-Link switch with Cat6 cable.


No Shaper rules setup, just base install of OPNSense

M2 Results:
Wired (Cat6A cable plugged into USB-C Dongle)

Unloaded = 15 ms
Download Active = +15 ms
Download Speed = 900.2 Mbps
Upload Active = +0 ms
Upload Speed = 118.6 Mbps

Wi-Fi

Unloaded = 15 ms
Download Active = +47 ms
Download Speed = 800.8 Mbps
Upload Active = +0 ms
Upload Speed = 112.7 Mbps

Intel Results:
Wired (Cat6a plugged into Thunderbolt Dock)

Unloaded = 18 ms
Download Active = +336 ms
Download Speed = 738.3 Mbps
Upload Active = +0 ms
Upload Speed = 113.5 Mpbs

Wired (Cat6a plugged into USB-Dongle, dock removed)

Unloaded = 19 ms
Download Active = +419 ms
Download Speed = 644.2 Mbps
Upload Active = +0 ms
Upload Speed = 119.9 Mbps

Wi-Fi

Unloaded = 21 ms
Download Active = +103 ms
Download Speed = 788.1 Mbps
Upload Active = +0 ms
Upload Speed = 115.1 Mbps


Shaper Rules setup according to linked article. 
I set the Download pipe to 1700 Mbit/s (85% of 2Gbs) and Upload pipe to 85 Mbit/s (85% of 100Mbs)

M2 Results:
Wired (Cat6A cable plugged into USB-C Dongle)

Unloaded = 13 ms
Download Active = +16 ms
Download Speed = 882.6 Mbps
Upload Active = +0 ms
Upload Speed = 80.9 Mbps

Wi-Fi

Unloaded = 15 ms
Download Active = +34 ms
Download Speed = 768.1 Mbps
Upload Active = +0 ms
Upload Speed = 80.5 Mbps

Intel Results:
Wired (Cat6a plugged into Thunderbolt Dock)

Unloaded = 17 ms
Download Active = +268 ms
Download Speed = 801.2 Mbps
Upload Active = +0 ms
Upload Speed = 80.5 Mbps

Wired (Cat6a plugged into USB-Dongle, dock removed)

Unloaded = 18 ms
Download Active = +373 ms
Download Speed = 704.1 Mbps
Upload Active = +0 ms
Upload Speed = 80.3 Mbps

Wi-Fi

Unloaded = 20 ms
Download Active = +101 ms
Download Speed = 772.6 Mbps
Upload Active = +1 ms
Upload Speed = 78.0 Mbps

So a few questions.

• Is the difference between the Intel chip and the M2 chip a known issue? Why the huge discrepancy between the 2?  It's not major as they work fine for their particular purpose, just work and putzing around on the internet, but I'm more concerned with my Xbox's for gaming and streaming.
• Why isn't there much of a change after adding in the Shaper rules?

I've tried a few things I've read around on the internet as well as this forum for tweaking things like the Quantum and Interval and all that, even lowering the download/upload pipes to 75%, 65% and even 50% of my max speeds and the numbers don't really change at all, maybe a couple of ms here or there.

Bonus question:
I have 32GB of RAM in the computer, but OPNsense only shows 16GB.  The BIOS reports 32GB, if I boot to a Ubuntu Live USB, it sees 32GB of RAM, why is OPNsense only showing 16?  Could that be a reason why there's not much improvement in the buffer bloat tests?