Rule of Thumb for Wifi Throughput (iperf3) as ratio of Connection (Tx Rate)

rhodos

Well-known Member
Hello, thanks for your thoughts in advance and apologies likewise if this is an easy find.

Having lived with a flaky home network these last 10 years I've taken the last few days to sort out a few issues...all patch cables replaced, all pins checked (simple network tester) and set up an iperf3 server on my main gaming PC to test it out.
Its a simple home network with one main router and then 4 access points / switches around the house all coming back through a 1G switch to the main router via 5e rated cable which was put in 12 years ago with a rewire.
The above tweaks have addressed some <100Mbps readings and every point now gives a solid 950+ Mbps to the iperf server over ethernet :)

So I've moved on to the wifi access points. I expected to see throughput at about 50% of the Tx connection but I am seeing it at 25%, e.g. I have a 866Mbps Tx from this macbookpro I am on now and the iperf3 result is 260Mbps, and on the very same laptop if I switch to a wired connection via a USB C to 1G Ethernet adapter I get 950 Mbps.
The access point is literally 0.5M (50cm) away from the laptop. And of course the access point has a 1G port back to the main switch (hence the 950Mbps on wired).

Is that typical, i.e. throughput about 25% of Tx rate on wifi?

1654333677851.png
 

captain morgan

Active Member
Can’t comment on the tx %, however I get ~1Gb/s on my wired network and ~600Mb/s on 5Ghz wifi @ ~2m & one ceiling
570Mb/s @ ~5m & one ceiling + one floor

Sounds like there’s a issue somewhere
Have you tried running the test at greater distances?
 

rhodos

Well-known Member
Can’t comment on the tx %, however I get ~1Gb/s on my wired network and ~600Mb/s on 5Ghz wifi @ ~2m & one ceiling
570Mb/s @ ~5m & one ceiling + one floor

Sounds like there’s a issue somewhere
Have you tried running the test at greater distances?
Just checking. Is that definitely the measured throughout using iperf you are quoting and not the connection speed?
I get a 860Mbps connection but only a 250Mbps throughput.

Just want to be sure before I go digging any further :)
 

mickevh

Distinguished Member
The Protocol Efficiency of Wi-Fi is of the order of 55-75% so the sort of numbers Captain Morgan cites would be about right for an MCS 9 2-stream 80MHz 802.11AC link with good signalling conditions - 39dBm RSSI and -90dBm noise are what I'd regard as good signalling conditions.

Again as Captain Morgan indicates, sometimes you can be "too close" to an AP and it upsets things (for example, the special streams don't have chance to find alternate physical paths) so as recommended, you could try backing off a meter or two as see if it makes any difference. The relative orientation of the antenna in the AP and client can also make a difference as to how the spacial streams can/cannot form up.

Wi-Fi is an "only one thing at a time can transmit" technology, so anything in the area competing for some "air time" (whether Associated with your AP or not) can rob throughput, as can any non-Wi-Fi sources causing interference. The only way to eliminate any outside interference sources would be to test in an RF shielded environment (like Gene Hackman's character's "lab" ("The Jar" he calls it IIRC) in the movie "Enemy Of The State") but of course that's impractical for most of us.

You might try stepping though the various protocols and protocol options and see if there's something that indicates culpability. For example, set the AP to 802.11A only mode and see what the numbers are (top rate there is 54mbps so throughput would be about 2/3rds that,) Then try N only in 20MHz channel width and test again. Then try again with a 40MHz N channel. And rinse and repeat with 802.11AC at 20Mhz, 40MHz and 80MHz channel width and see how the number work out. Wiki's articles on 802.11A, 802.11N, 802.11AC have tables of how to work out the Link Rates versus channel width, number of spacial streams and so on. Though the more complicated protocols are a bit harder to comprehend due to the number of variables.

You could try using different radio channels and see if that makes a difference - though this is guesswork more than anything. 80MHz channels are formed by binding up 4 x 20MHz channels, so if there's an "issue" in one (or more) of them, it may be that the channel isn't forming properly despite being nominally reported at 80MHz. Equally, some NIC's will just "not bother" trying to grab an 80MHz channel if it thinks it's "only" got a small amount of data (small packet) to send as it's costly in time to seize an 80MHz channel, so sometime it's more efficient (in time usage) to use narrower channels and they just don't try.

You could perhaps try shoving more data across than an shortish iperf test - try copying a DVD rip or something, time it on a watch and do the maths.

A lot of this is guesswork as without some fairly forensic analysis with speciallist tools, it's really hard to figure out what's going on.

Incidentally, I notice your screen grab cites a country code of "DE" (Germany) - if you the in the UK, you ought to set this to UK - not that I think it has much bearing on your problem.
 
Last edited:

rhodos

Well-known Member
Update - looks like I should blush.
If I use iperf client on iPhone I get closer to 500Mb/s. That uses 5 streams which got me looking at the command line options on the Mac client. With a single stream it was doing 260Mb/s but if I use the '-P' argument e.g. '-P 5' then it uses multiple streams and I get about 500Mb/s.
Which for a 5 year old 867ac router is about right on the 2/3 rule.

Thanks for the input. For 10 years my whole network was limited to 100Mbps until I finally found the time to test and address this week - would advise others to do also no matter how tech savvy they are. I assumed just because I had the capable kit it would just work at 1Gbps. [a few wrongly terminated patch cables fixed later and its up to 950Mbps where it should be].
 
Last edited:

mickevh

Distinguished Member
Thanks for the update - I'm glad it's all sorted.

If you haven't already done so, I'd invite you to confirm the "country code" setting" in your AP's. It can effect things like the maximum transmit power and what channels are allowed to be used, though it's much of a muchness as it's common across the most of the EU and even post Brexit the UK is still in line with the EU. But it's one of those little "details" that's worth addressing when spotted lest it comes back and bites you it the future of/when something changes.
 

The latest video from AVForums

Samsung QN95B 4K QLED TV Review
Subscribe to our YouTube channel
Support AVForums with Patreon

Top Bottom