Reaching 1 gbps using RT-AC86U and PCE-AC68?

[thedude92]

Hi guys,

I was recently forced to surrender my cable connection in favor of a wireless setup for my PC.
Naively, I held onto the belief that sizing both the router and wireless adapter to 1000+ Mbps would ensure zero performance drop for my 1 gbps connection. As a consequence, I acquired a RT-AC86U and a PCE-AC68.

The outcome was a performance drop from an average 930 Mbps on cable to an average 550 Mbps on Wifi and moderate disappointment.

I am trying to understand if there exists a bottleneck in my hardware (ONT -> router -> adapter chain) or if it's literally impossible to keep the 1 gbps performance in Wifi no matter what I do.
Right now there is a discount on a PCE-AC88 which I could swap with my PCE-AC68 if it proves to be the root of all evil. If someone can provide some insight on this matter I would be very grateful!

I took the liberty to run these tests:

1) wired case #1:
Signal path: optical fiber -> ONT (Huawei Life 8121H)'s Gigabit port-> Cat 6 -> computer
Speed test: 930 Mbps download & 500 Mbps upload avg. speeds
(matching the provider's 1000 & 500 maximum speed advertisement)

2) wired case #2:
Signal path: optical fiber -> ONT (Huawei Life 8121H)'s Gigabit port-> Cat 5E -> RT-AC86U-> Cat 6 -> computer
Speed test: 930 Mbps download & 500 Mbps upload avg. speeds

3) wireless case #1:
Signal path: optical fiber -> ONT (Huawei Life 8121H) Gigabit port-> Cat 5E -> RT-AC86U-> PCE-AC68 -> computer
Speed test: 550 Mbps download & 500 Mbps upload avg. speeds

Notes:
- no other devices connected to Wifi during the tests
- router and adapter in the same room, less than 2m apart, in line of sight of eachother
- ONT, router, and pc rebooted prior to measurements
- ONT is not operating in bridged mode, just daisy chained it to the RT-AC86U, unsure if it matters

Would an upgrate from the PCE-AC68 to the PCE-AC88 bring any improvements?

Regards,

 

[ColinTaylor]

Look at the router's Wireless Log and use a WiFi analyser app to confirm the router is operating with 80MHz bandwidth. If it's only using 40MHz those are the speeds you would expect.

 

[wouterv]

Some specification related stuff:
The RT-AC86U router supports 4 data streams at 5 GHz, with 80 MHz channel bandwidth.
The PCE-AC68 adapter supports 3 data streams at 5 GHz where the PCE-AC88 does support 4 data streams and does better match the router.
Theoretical data rate numbers at 80 MHz, these should show in the wireless network properties as Link Speed:
3 data streams = 3x 433.3 = 1299.9 Mbps
4 data streams = 4x 433.3 = 1733.2 Mbps
In practice the final throughput will never reach the data rate.
With 3 streams you will never reach 1000 Mbps, with 4 streams you might.
Local radio conditions, distance and disturbance play a role too.
Maybe the PC itself is the bottleneck and cannot optimally deal with the fast WiFi adapter.
Test the Internet speed at about 3 meter distance from the router, with the router and network adapter free of obstructions: this gives you your landmark.
Further away from the router the speed will only go down.

 

[thedude92]

Hi again guys,

I've played around with the bandwidth setting of the RT-AC86U router.
I couldn't find messages regarding the bandwidth in the router's system log, neither could I find a proper WiFi app on the App Store so I went into the router's settings.
Bandwidth was set by default at 20/40/80 MHz so I tried out all 4 possible wireless scenarios: 20, 40, 80, and 20/40/80 MHz as well as the wired "reference" scenario.
I also added a laptop with an Intel AC 8265 into the mix as a sanity check for the tests, I believe the overall "wireless" results look quite close across the 2 devices.

The measured speeds are a bit lower than yesterday's numbers which were acquired at about 1 AM, I believe it's because the latest data was acquired at about 1 PM.
Just saw wouterv's post after I had finished the measurements, the distance I used was about 1.5 meters, practically no obstructions... I'm considering emptying the room for the next round

Here's the data:

Router notes:
- the router has the latest firmware from ASUS, no factory reset has been run, it was just updated during the initial setup
- I haven't enforced a fixed channel on the router, it is set to Auto

I've now also checked the link speed and it is operating at 1.3 Gbps so that should confirm 3 streams and 80 MHz as described by wouterv.

I am hoping the bottleneck isn't the PC, the hardware is quite recent, I will try to look into this as well if the PCE-AC88 doesn't do the trick.
Indeed, for the laptop's case it seems that it isn't able to match the PC's numbers for the "wired" tests so I assume the laptop hit its limits.

Will drop by the store and get a PCE-AC88 to try it out and return with details.

 

[wouterv]

Anyhow at a data rate of 1300 Mbps, the Internet speed or throughput should be better then ~500 Mbps you found.
Here I have a laptop with two data streams and a RT-AC68U router: data rate 866 Mbps, Internet speed easily hits 500 Mbps.
Your 1300 Mbps should at least result in a throughput of 750 Mbps and 4 streams should bring you up to 1000 Mbps at least.

Although the data rate indicates 1300 Mbps, you might try to change the router:
Advanced Settings - Wireless - Professional - 5 GHz
Airtime Fairness: Disable
Universal Beamforming: Disable

[EDIT]
To find the actual wireless channel and bandwidth go to the Router System Log > Wireless Log page:
Halfway down that page you find like this (from my router):
Chanspec: 5GHz channel 106 80MHz

 

[Klueless]

Full disclosure; take anything from a guy who calls himself @Klueless with a grain of salt : -)

• Our very own @thiggins offers some excellent insights;
  • His Reveiw on the RT86U
  • His test methodologies
  • While much of what THiggins writes is over my head my take away is your wireless speeds are pretty damn good!
• I like that you are calibrating your wireless speed tests against wired but do keep in mind:
  • Wired is Full Duplex
  • Wireless is only Half Duplex
    • Based on your relatively upscale post I can assume you've verified nothing else (e.g., a forgotten security camera) is running during your tests?
    • You are using some kind of Internet Speed Test. Without knowing more one might assume;
      • http overhead?
      • "two way communications" (e.g.; down, here's some data; up, thx, give me more)
    • You might consider a LAN speed test
      • I'm old so I still use "netstress" by "nuts about nets" but that's going to come up on the short side of Gigabit.
      • Many use something called "iPerf"

I anxiously await your PCE-AC88 test results. Best of Luck!

 

[thedude92]

Hello again,

I've updated my setup with the PCE-AC88 and achieved some interesting results.
I confirm that the change yielded a clear speed improvement, but appears to be falling short of expectations.
The average speed test results have been bumped up from 500 Mbps to about 725 Mbps on the "browser speed test", details below.

Overall the results remain mixed:

the good: The results with the PCE-AC88 (2100 Mbps) are about 50% better than with the PCE-AC68 (1300 Mbps) which is in the ballpark of the 60% speed improvement advertised.
the bad : Both the PCE-AC68 and PCE-AC88 appear to be consistently underperforming vs. their specs and other owner's reports in my setup.

Furthermore, @thiggins's RT-AC86U review from @Klueless indicated it should be fit for gigabit performance:
ref: https://www.smallnetbuilder.com/wir...900-wireless-router-reviewed?showall=&start=2
"5 GHz average throughput of 164 Mbps downlink and 263 Mbps up".
"911 Mbps downlink and 943 Mbps up mean you are essentially getting the equivalent of gigabit Ethernet performance. Keep in mind, this is with signal levels you're unlikely to encounter in real-world use."

I have followed up with a few additional investigations of my setup to zero in on the cause of the underperformance.

A) The Test Methodology

It's been very difficult to accurately gauge the impact of my modifications because the browser speed test results vary significantly from run to run.
Because of this run to run variation change can very easily become a false positive/negative modification with my current approach. I am looking into improving my methodology.

@Klueless:
I might have to read some more on this topic before being prepared for @thiggins's test methodologies. It's impressive to see how it's done properly though.
Haven't gotten my hands on a security camera yet, but might in the near future. Yes, I confirm there's no competition on the network during the tests.

The download and upload measurements are done sequentially on the speed test website, they appear to be "half duplex" aware.
However, I noticed that iperf allows the user to choose, keeping it sequential.

A.1. Browser Speed Test Approach

My main method up to this point has been to use 2 speed test websites.
The nice thing about this method is that need one device is sufficient to conduct this test.
The bad thing is I don't have any idea about this method's intrinsic overhead. Any number of external factors could appear in between leading to measurement spread.

The first speed test website is the first results you get on Google for such a website.
The second website is that of my internet provider, which I am hoping translates into 1 gbps courtesy of being in their own network.

A.2. iperf3 - attempts, commands, scenarios

I have tried out iperf3 and ran into the following obstacle.
I require 2 devices to be able to run the test, as I only possess one computer which can support 1 gbps over wifi so I tried out 3 scenarios:

Client side (PC) command : iperf3.exe -c *insert_iperf_server_ip* (direct) and iperf3.exe -c *insert_iperf_server_ip* -R (reverse)
Server side (Laptop) command : iperf3.exe -s

I hope that I have correctly accounted for @wouterv's mention of FULL DUPLEX (wired) vs. HALF DUPLEX (wifi).

Test #1: PC wired - Laptop wired through the router. (reference test)
Results: 949 Mbs/sec (direct), 942 Mbs/sec (reverse)

Test #2: PC wireless - Laptop wired through the router.
Results: 376 Mbs/sec (direct), 369 Mbs/sec (reverse)

Test #3: PC wireless - Laptop wireless
Results: 187 Mbs/sec (direct), 182 Mbs/sec (reverse)

Do these very slow wireless speeds for tests #2 and #3 make any sense??

B) The Software

B.1. RT-AC86U firmware update

I have updated the firmware on the the RT-AC86U router to Merlin RT-AC86U_384.14_2 (released 01.01.2020)
Also followed up with a Hard factory reset using Method #2 from Asus's website.
ref: https://www.asus.com/support/FAQ/1039078

B.2. RT-AC86U configuration

Updated Advanced Settings -> Wireless -> Professional -> 5 GHz per @wouterv's advice
Airtime Fariness : Disabled
Universal Beamforming : Disabled

B.3. PCE AC-88 configuration - device manager

After coming across this thread I started wondering if the PCE AC-88 wireless adapter is configured correctly.
ref: https://www.snbforums.com/threads/ac5300-with-pce-ac88-speed-issue.36138/

I haven't yet come across a detailed list anywhere so I am listing them in full for reference purposes.
I will try experimenting with them if I run out of options with the router. Below are the *default* settings after driver installation.

Device Manager -> Asus PCE-AC88 Properties -> Advanced

20/40 Coexistence : Auto
2G Bandwidth Capability : 2.4G: 20/40 MHz
40MHz Intolerant : Disabled
802.11d Support : Enabled
Antenna Diversity : Auto
ARP Offload : Enabled
Assoc Listen Interval : 1
Beamforming : Disabled
Bluetooth Collaboration : Auto
BSS PLCP Header : Auto
D2 Listen Interval : 5
Enable/Disable LTR : Enabled
Fragmentation Threshold : 2346
Locally Administered MAC Address : Not Present
LTR Active value in microseconds : 60
Minimum Power Consumption : Enabled
NS Offload : Enabled
Power Output : 100%
Priority & VLAN : Priority & VLAN Disabled
RTS Threshold : 2347
VHT Features : 7
Wake On Magic Packet : Enabled
Wake On Pattern Match : Enabled
WFD Channel Number : 11
WiFi Rekeying Offload : Disabled
WMM : Auto
XPress (TM) Technology : Enabled

B.4. ONT configured as router, not as bridge

Found this thread mentioning a potential difference in performance between DHCP with ONT and PPoE with ONT as media converter for a RT-AC68.
ref: https://www.snbforums.com/threads/rt-ac68u-getting-1gbps-wan-lan-throughput.29471/

Also found this thread mentioning ISP needs such as MAC clone, bridge, disable DHCP, disable Jumbo Frame, enable QOS..
ref: https://www.snbforums.com/threads/asus-rt-ac86u-not-getting-maximum-speeds-on-gigabit-fibre.50402/

Any chance that configuring the ONT to bridge mode by calling the ISP lead to a miracle?

B.5. Bitdefender Antivirus + Firewall?

I have kept the Bitdefender security suite up and running throughout my tests. I am assuming this is normal even during test conditions.

C) The Hardware

C.1. PC Specs

Motherboard: ASUS X470 F (AM4 AMD Chipset)
CPU: AMD Ryzen 7 2700, note: no core physical/logical hits 100% load during the speed tests, should outrule CPU
SSD: Samsung NVMe PCIe M2 970 EVO

While swapping out the PCE-AC68 I asked myself if using a PCIe 2.0 1x port was leading to bottlenecking. Also wondered if a PCIe 3.0 port would make matters better.
Came across this post mentioning some details on the Broadcom chip.
ref: https://www.anandtech.com/show/10306/asus-unveils-pce-ac88-4x4-wifi-card

"The PCE-AC88 is a small half-height card that features PCIe x1 interface (the Broadcom chip supports PCIe 3.0, but exact data rate of the card is not given)"
-> moved it to the closest free PCIe 3.0 16x slot to the CPU

"To actually hit the maximum bandwidth, the card has to be connected to routers powered by Broadcom’s latest Wi-Fi SoCs that support TurboQAM (256 quadrature amplitude modulation) and NitroQAM (1024 quadrature amplitude modulation) technologies. "
-> this requirement is satisfied by the RT-AC86U

C.2. antenna type and placement

Will get to this too.. soon..

Do you guys consider/agree that the setup is underperforming in the browser & iperf3 tests given the numbers mentioned above?

Regards,

 

[doczenith1]

For testing wireless with iPerf3 try this client command line:
iperf3 -c 192.168.1.10 -O 5 -i 5 -w 4M -t 30
The -w 4M is the secret sauce that helped me get speeds in excess of 600 Mbps on a 2x2 ac device.
I had mixed results when using -R. I found it best to switch the server to my other device.

 

[wouterv]

All I can say is that something keeps slowing down the wireless performance.
In my much simpler situation with a RT-AC68U and at best a 2x2 adapter built in a few of the available laptops, I easily get 500 Mbps, sometimes hitting the ISP subscription max of 600 Mbps. It doesn't matter much if I test with the laptop closer to the router or as usual with the laptop in the room next to where the router is.
Your RT-AC86U with the PCU-AC88 adapter should be able to perform better.

What is the 5 GHz channel that is in use?
Is it a fixed channel or set automatic?
If automatic: does it long time stay the same or does it hop?
How is your most ideal test setup and situation? Are there possible disturbing factors? Other wireless devices? Bluetooth? telephones? Strong 4G? Neighbors with strong Wifi signals?
I assume in your best setup the router antenna's and the adapter antenna's have a free line of sight to each other, far away from concrete walls or steel panels but also not too close to each other (I would say about 1.5 to 3 meter should do).