Archer C7 V5 and Cascade Access Point Download worse than upload

Installing and Using OpenWrt
1

Hi:
I upgrade my Archer C7 V5 from OpenWrt version 18.06.4 to 21.02.0 and everything is working. I have a 500/100 network and I can managed to have 450Mb in a cascade scenario, my supplier router and Archer C7. On WIFI I can have something like 240Mb

The big issue rises when I try to put other router as AP connected with Archer V7. I have an TL-WDR3600 v1 with OpenWRT 18.06.4 and a TL-WR841N with 15.05 version and on both, trough WIFI and also using a cable the download speed is less that upload speed.

The ratio between Upload and Download are almost 50%, where download is less. I made tests with Speedtest and also with iperf3 and the results are consistent. I have 98Mb upload and something like20 donwnload.

I have the Software based offloading for routing/NAT active but if deactivate the result are the same.

So, why when I connect directly to the router I have all the speed and when I have an AP on the middle the download speed degrades? Any bug on version 21?

2

It's not entirely clear how things are physically connected and how you are testing things (or the actual numbers you are getting -- more specific values for each test scenario would be much more useful than the somewhat ambiguous descriptions you have provided), but fundamentally, that router is OLD. You should just get rid of it. It is 802.11n -- N300, so the max theoretical/advertised bandwidth would be 300 Mbps, but typical N300 devices tend to have a real-world max throughput of around 75Mbps. Plus, the device itself only has 100Mbps Ethernet ports.

If you are routing through the WR841N, you're going to get even lower performance numbers because the SoC is quite underpowered. If you use it as a dumb AP, you might be able to approach the real-world figures of around 75-90Mbps.

So, the WR841N is an old, under-powered, and slow device. AND, you are running an ancient version of OpenWrt on it (15.05 is now 6 years old) which has many security vulnerabilities and has been end-of-life and unsupported for many years.

Try a newer device and you'll be much happier.

3

…and the tl-wdr3600 isn't much better either, while it's fine to use with contemporary OpenWrt, it's severely underpowered for the speeds you mention. Even the Archer C7 v5 technically is (it just has faster 5 GHz wireless and a slightly higher clock rate), as you'll need to use software flow-offloading to achieve your figures in the first place (without that, you'd top out around 200 MBit/s).

4

Hi to all. Many thanks for your reply. I will try to explain better the scenario. Yesterday I was to tired.

So, the diagram are the following:

Supplier Router <-> Archer C7 <-> WDR3600/841N

The connection between Suplier Router and Archer C7 is a LAN <-> WAN and WDR3600 and 841N are connect to LAN port of Archer C7 as an AP.

But, my issue are why the download speed is lower that upload speed when I connect to WDR3600 or 841N? On Archer C7 I have software flow-offloading enabled.

I make many test with that, so lets put it.
Speedtest
Old Desktop Pentium by Cable to Archer C7 - 450Mb
New Laptop DELL i7 by Cable to Archer C7 - 450Mb
Laptop by WIFI to Archer C7 - 215 download and 100 upload
Laptop by Cable to TL-841N - 47 download and 94 upload
Laptop by WIFI to TL-841 - 26 download and 75 upload
Mobile XIAOMI M10 to Archer C7 - 204 download and 105 upload
M10 to TL-841N - 28 download and 70 upload
A old Toshiba Laptop to WDR-3600 by WIFI - Between 50-80 download and 95 upload.

So, with Speedtest the Archer can handle the 450Mb by cable and 200Mb by WIFI on a 5Ghz network but when I move to put and router as an access point front of it, by cable or wifi the download speed is worst that upload speed.

Also, i make tests with IPREF3, lest put some of them.
Between Laptop and Desktop the speed goes from 600-830Mb depending the flow direction, connected by cable and trough Archer C7.
By direction I use the reverse function on IPREF3.

If I put the TL-841 on the middle by cable or by WIFI the speed on both ways is around 80Mb

If I try with my mobile M10 with Desktop I have around 250Mb on both ways.

As I can see, with small differences of 10Mb, the speed between upload and download are smaller.

Now, I put my Laptop on the same zone of the supplier router, meaning that it was connected to a port tagged to the same network where are the WAN port. So, any traffic need to passe by firewall.
Here, I have the same behavior.
If I connect directly to Archer C7 I have:
Desktop to Laptop 424Mb download and 440 upload
Mobile M10 Laptop 135 download and 163 upload

But, when I put the TL-841 on the way I have
Desktop to Laptop 22Mb download and 94 upload
Mobile M10 Laptop 30 download and 72 upload

Once more the download is less that upload.

With SpeedTest I made some test with and with out the software flow-offloading activated but the behavior are the same.

On Openwrt 18 I think I did not have this but on that time I have only a 200Mb/100Mb speed and I did not care to take much of them and did not make tests. Now, I am restructuring all my network to get the most of 500 and I get this,

Could be something regarding QoS or something? The AP have the DHCP and DNS services down.

Many thaks for your help and I hope that I can explain better now the issue.

Best regards

CGR

5

Hi, a small update

I put the 841N connected to the supplier router and I have with SpeedTest
WIFI - around 50Mb download and upload
Cable - 95Mb download and upload.

So, the issue, of having a difference between download and upload happens when I have it connected to Archer C7

Best regards

CGR

6

This indicates you are getting the maximum speeds that the ethernet port on the 841N can handle (100Mbps ports ~= 95Mbps maximum actual throughput).

The wifi, as stated, is probably only good to 50-75Mbps actual throughput, and it looks like you are getting that, in general.

We'd have to see how you have both of the routers configured (the actual config files) to understand more about why you might be seeing a drop when you cascade them. However, why are you cascading the routers in this way? What is the purpose of your 841N in your network? You would probably be best served by setting it up as a dumb ap.

7

Hi, I think I found the issue. The bottleneck looks to be the difference between WAN router connection, 1Gb lan port and 500Mb Internet Speed and the LAN connection to 841N that is 100Mb and also to WDR3600 that is connected to a 100Mb passive Switch.
If I connect the WDR3600 that have 1Gb port to the Archer C7 1Gb port directly I get for instance 240Mb on WIFI. I have to test by cable.

Also, this is a little more strange but I bought a 1Gb passive switch and connected it between the TP-LINK 841N and Archer C7 and now the upload and download are the same. 95Mb by cable and 55Mb by WIFI.

Could be some buffering issue on Archer C7, that have only 1 CPU and cannot sustain the speed correctly? The upload you go from a slow speed to faster, but on download you came from a faster to slower.

With my best regards

CGR

8

Like I said before, the 841N will never achieve speeds higher than 95Mbps in any circumstances. The only potential improvement you could make would be to see if you can improve WiFi speeds. But as I also said, you’re not going to get much more than maybe 50-75mbps WiFi speeds from that device because of the outdated WiFi technology it is using.

9

Hi, thanks for your reply.

But my question was only one, why is download worst that upload. Is that I am inquiring. The rest I am aware that 841N is a very cheap and powerless router. :slightly_smiling_face:
I reminder that by cable using the 841N I have 47 download and 94 upload with a I7 laptop and if that laptop connect directly to Archer C7 I have 450 download and 100 upload (I have 500/100)
Also using IPERF3 I have 22 download and 94 upload.

And also, the TP-LINK WDR3600, better that 841N have the same behavior. According to me I have a buffering issue on Archer C7, that can not sustain to 841 a continuous speed.

Regarding absolute speed I understand that and that is not my issue :slight_smile:

With my best regards

CGR

10

Hi,

what is your speed using non ct drivers? Mine is much better then. See State of TP-Link Archer C7v2|v5 in 2021 for more info.

11

huh maybe should have mentioned that in the original post.

Make sure all your cabling is operating at gigabit. You must use good 4-pair cables. It was kind of common in 10/100 installations to use two pair cable or split a 4 pair cable into two circuits. This will give uncertain results (possibly no link or unreliable link < 100 Mb) with GbE hardware at both ends.

Two PCs connected to the gigabit LAN ports should iperf between each other at gigabit line speed, about 940 Mb. That would be pure hardware switching in a separate chip in the router, not depending on the router CPU.

iperf3 uses a lot of CPU for itself and running it on a device with limited CPU will produce results less than the actual network capacity.

12

Hi, thanks. Regarding that I made tests between a I7 and a Desktop Pentium using Iperf and I managed to have 860Mb with Archer C7 V5. LAN to LAN.

My cable are CAT 6 at least.

Best regards

CGR

13

Hi, thanks. That is new to me! I need to see how I can managed that or change that! I will take a look on that

Can you tell me your speed? With IPerf I can not pass 250-290Mb and with Speedtest I can not pass 210, with a I7 or a Xiaomi M10.

Best regards

CGR

1 Like
14

My speed is only capped by cpu, about ,400-500 mbps

15

Hi, thank you.

I made that change, according with some pages on this forum, replacing the both modules and looks almost the same. I replaced the kmod-ath10k-ct and ath10k-firmware-qca988x-ct
However I think that on 5Ghz the speed is 1 or 2% faster with CT drivers and in 2.4Ghz is 5-7% faster with no CT drivers.
I achieve 204Mb with 5Ghz and 150Mb with 2.4Ghz with NAT and Speedtest

Many thanks for your help.

CGR

1 Like
16

btw I don't use NAT, just dumb AP because I've got a separate firewall/router.

17

Hi, thanks for your reply.

Regarding that I did made some tests on the beginning of this thread and with CT drivers I had around 300Mb. When possible I will try with no CT drivers. I use Iperf3 for that tests.

Btw, with a new Desktop PC that I assembly for a friend, with a entry level Asus Motherboard and Windows 10 I get with Speedtest, trough cable and NAT, 515Mb, the maximum that I get from my Internet supplier.

With my best regards

CGR

1 Like
18

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.