RT-AX88U and Intel 8265 Performance

[Kopas]

I'm running an RT-AX88U on version 384.13. Testing throughput with an Intel 8265 adapter running latest drivers on Windows 10. Both the router and Windows show that I'm connected at 866Mbps:

When I do some testing with iperf to a wired workstation... I get around 380Mbps actual throughput. (Up/Down is the same.) I know you can't expect to get as much throughput as your connected rate... but I guess I expected it to be higher. Anyone care to tell me I'm crazy?

 

[latinkreationz]

Delete

 

[Val D.]

I get around 380Mbps actual throughput.

Sounds about right. Actual throughput is usually slightly less than 1/2 of the link speed. In good conditions I get about 200Mbps on 1x1 AC clients (433Mbps link speed) and about 400Mbps on 2x2 AC clients (866Mbps link speed). Speed varies depending on what is transferred, from/to LAN or WAN, to wired or wireless, how busy the WiFi channels are, how busy the other device is, etc. My ISP speed is 300Mbps, so I'm not going to buy multiple 4x4 AC adapters $100 each any time soon.

 

[latinkreationz]

I just realized my work laptop has the 8265 card in it and I do pull around 600 on a good day. I have 1 gig internet with Cox.


Sent from my iPhone using Tapatalk

 

[Val D.]

8265 card in it and I do pull around 600 on a good day

Intel 8265 is a 2x2 AC adapter (up to 866Mbps link speed) and it is unlikely to reach 600Mbps throughput, no matter how good the day is. Something to read, gives good explanation how things work and what to expect: https://www.duckware.com/tech/wifi-in-the-us.html

 

[latinkreationz]

Intel 8265 is a 2x2 AC adapter (up to 866Mbps link speed) and it is unlikely to reach 600Mbps throughput, no matter how good the day is. Something to read, gives good explanation how things work and what to expect: https://www.duckware.com/tech/wifi-in-the-us.html

Damn Speedtest lied to me then lol.


Sent from my iPhone using Tapatalk

 

[Val D.]

Damn Speedtest lied to me then lol.

Possibly. Speedtest for Windows is way off, for example. Download speed is usually about right, but upload speed is all over the place. I have up to 45Mbps upload speed as per Speedtest, actual up to 20Mbps. Web version is more accurate, but still have to run it 2-3 times to get realistic results.

 

[Greg72]

The speedtest in a browser will be off from the Okla app for Windows 10. You also need to make sure that the Router is set up properly to allow you to get max connection speed from your ISP.

After spending all day farting around with Windows 10 because of some issues that have arisen from the Sept. updates, I have found from searching that Microsoft has screwed up the pipeline between OS and Browser, because they are moving more of the OS towards their version of Linux Kernal, which means that they also screwed up the Intel Wireless connectivity.

I finally got my USB 3.0 to Gig Ethernet stable at 938mbps, which is the max speed with my Gig connection from my ISP. In turn Windows 10 got worse.

 

[sfx2000]

Look for the latency... not the bandwidth.

this is on a simulated 50/10 broadband connection with a QC-Atheros AP121 reference platform - you can buy the same device for 20 bucks on Amazon...

Fun stuff... and no bufferbloat....

 

[Kopas]

Sounds about right. Actual throughput is usually slightly less than 1/2 of the link speed. In good conditions I get about 200Mbps on 1x1 AC clients (433Mbps link speed) and about 400Mbps on 2x2 AC clients (866Mbps link speed). Speed varies depending on what is transferred, from/to LAN or WAN, to wired or wireless, how busy the WiFi channels are, how busy the other device is, etc. My ISP speed is 300Mbps, so I'm not going to buy multiple 4x4 AC adapters $100 each any time soon.

Thanks for the feedback... very appreciated.

I went back through some older notes I had lying around... at one point it seems I was able to pull marginally higher numbers. However, the testing I was doing was just Internet speedtests and I was getting the results I expected. (Only had 300Mbps from Xfinity, which bursted higher during tests.)

Fast forward about 1-2 years and I'm now on the Extreme Pro tier with a 13 year old son who can gobble data faster than I care to admit. Xfinity bumped this tier to 600 Mbps and that's when I noticed I wasn't able to saturate through Internet speedtests anymore. However, my wired Nvidia Sheild showed >700Mbps during a Netfilx test.

At this time, I was already in the middle of upgrading my wireless network. My son is furthest away from the router. (Netgear R7000 running Xwrt-Vortex at the time) The best speeds he was able to pull were 170Mbps through a 2X2 AC adapter on his gaming PC. I upgraded the router to an ASUS RT-AX88U and his adapter to an ASUS PCE-AX58BT (Intel AX200). It worked, where he previously would max out at 170Mbps on Internet speedtests, he was now able to pull around 400Mbps. (I attribute this mostly to the better antennas on the adapter.) But, it was still slower than my wired devices were pulling.

This is when I setup iperf and started digging. I think it was just the timing of everything that made me start doubting my setup. We have 20+ devices on the WiFi and some very heavy users. Glad to hear I'm getting the results I should expect.

 

[Kopas]

After spending all day farting around with Windows 10 because of some issues that have arisen from the Sept. updates, I have found from searching that Microsoft has screwed up the pipeline between OS and Browser, because they are moving more of the OS towards their version of Linux Kernal, which means that they also screwed up the Intel Wireless connectivity.

Very interesting... as all of the devices I'm testing with are Windows 10 with latest updates and Intel adapters.

 

[Kopas]

Something to read, gives good explanation how things work and what to expect: https://www.duckware.com/tech/wifi-in-the-us.html

That was a great read.

 

[Greg72]

Very interesting... as all of the devices I'm testing with are Windows 10 with latest updates and Intel adapters.

I tried a stripped down version of the current 1909 eval and the problem persists in it between OS and browser, but not in the Speedtest app. My Macbook Air runs correct speeds, so it leads me to believe that the start is to strip down Defender protection to see if it is the cause.

 

[Val D.]

I did some speed tests early in the morning during quiet WiFi time. Clients 1x1 AC reach about 250Mbps, 2x2 AC reach about 440Mbps to wired to the router computer. So it can go higher, but not by much and not all the time.

 

[Greg72]

Once they start using hardware level calls vs going through a bunch of hoops with webpages and browsers, thing awill not speed ip.

http://scholar.harvard.edu/files/mickens/files/gibraltar.pdf

Using the Okla Speedtest app vs running any speedtest in a browser, you will only see the highest possible speeds.

 

[Val D.]

you will only see the highest possible speeds

As mentioned above, Speedtest App for Windows is sometimes showing speeds higher than highest possible.

 

[Makaveli]

As mentioned above, Speedtest App for Windows is sometimes showing speeds higher than highest possible.

For me the app is more accurate than the flash based website.

And fast.com told me I had a 3.3Gbps connection...

I've found good numbers with these also

https://www.nperf.com/
http://ovh.net/

Some of the sites seem to be dependent on your ISP and what peering they use.

 

[Val D.]

And fast.com told me I had a 3.3Gbps connection...

Make sure your ISP doesn't know.