Wi-Fi performance bad on RB4011 - possible misconfig

[santyx32]

Can get up to 450mbps using btest in my hap ac2, this is my config if it helps

/interface wireless security-profiles
set [ find default=yes ] supplicant-identity=MikroTik
add authentication-types=wpa2-psk disable-pmkid=yes eap-methods="" \
    group-key-update=1h mode=dynamic-keys name=main supplicant-identity=""
/interface wireless
set [ find default-name=wlan1 ] amsdu-limit=2048 amsdu-threshold=2048 band=\
    2ghz-b/g/n country="united states" disabled=no distance=indoors frequency=\
    2462 installation=indoor keepalive-frames=disabled mode=ap-bridge \
    multicast-buffering=disabled multicast-helper=disabled security-profile=\
    main ssid=network wireless-protocol=802.11 wps-mode=disabled
set [ find default-name=wlan2 ] amsdu-limit=2048 amsdu-threshold=2048 band=\
    5ghz-a/n/ac channel-width=20/40/80mhz-Ceee country="united states" \
    disabled=no distance=indoors frequency=5660 installation=indoor \
    keepalive-frames=disabled mode=ap-bridge multicast-buffering=disabled \
    multicast-helper=disabled security-profile=main ssid=network-5g \
    wireless-protocol=802.11 wps-mode=disabled

PD: I live in front of some old buildings with lots of WISP and 4G antennas, the non-DFS spectrum is extremely crowded as 2.4ghz is, so using a free DFS channel helped me

[ksuuk]

Can get up to 450mbps using btest in my hap ac2, this is my config if it helps

I don’t know why, but there is no “one for all good” wifi configuration, if even two same type of models, with same configuration, acts totally different way, not to mention of different models.

Yea, I know, now some will write a long list of all kind reasons (different environment, different clients, different users, different weather etc), but somehow other brands do not have such problems. Maybe Mikrotik builds hardware from disqualified parts and as their tolerance is very high, then users must tune and tune to achieve the basic conditions, when some other brands almost does not include any tuning options and still they work well and perform better.

[rkrisi]

These 3 devices are used for testing. Totally random screenshot, so I have not adjusted anything.

Ok , rkrisi (Kristof) , thanks for sharing. There is normally a constant or even real time monitoring needed, but lets start with a little comment on this registration table.

Elements to consider:

• the ultimate max we can expect from this equipment (AP + clients) is “866.7 Mbps/80 MHz/2S:SGI”, this with a stron clean signal in both directions with 256 QAM encoding at VHT MCS 9.
  We are operating in the yellow part of this generic 802.11 table: https://d2cpnw0u24fjm4.cloudfront.net/wp-content/uploads/802.11-Data-Rates-and-SNR-Requirements-download.pdf
  Telling us that we should have at least a SNR of 35, what comes with a signal of -51 dBm. (Mikrotik has its own table , and even documents receive sensitivity in the product specs)
  Whith that 866Mbps interface rate one can expect around 500 Mbps-600 Mbps net throughput in one dedirection.
• a second element is a mismatch in the RouterOS commands and a 4x4 device. RouterOS GUI only handles 3x3 (ch0,ch1,ch2). Everything in ch3 is missing. Even more, the wiki says if you want to use the 4 chains, leave “Data rates” in default. Only recent change logs did mention 4x4 support through CLI.

Just about your table, the observation learns something (but of course here the location and orientation, and the numbers on the client devices are missing). Also missing is the rate of variation of those numbers. I thought I found a tool to log this over time, but lost track of it. There are script to log this in the wiki.

Device 1 : iPhone
-transmit rate selector is at (390 Bbps) HT MCS12/MCS04/16QAM. The selector has dropped till a robust signal (16QAM)
-receive rate is set at (585 Mbps) HT MCS14/MCS06/64QAM. This requires a signal of -59 dBm and a SNR ratio of 26, we have a signal of only -70 dBm, coming from the combined ch signals, including ch3. I don’t know if the 2 bests are combined, or all 4.

• the retransmits are acceptable (249000/216000) but of course eat away some of the capacity (P-throughput : supposed to be averaged over 5 seconds) 305Mbps

Interpretation: quite good, iPhone is at the end of its wifi range (unless the iPhone is conserving batteries and has reduced its transmit power). The 12% misses make the router drop the MCS rate. An iPhone is normally handheld, and that hand together with the human body is an absorber for wifi signals. Unclear what the signal strength is that the iPhone receives.

Device 2: iPAD
-transmit 585Mbps HT MCS14/MCS06/64QAM
-received the same numbers but with a similar signal of -68 dBm but the SNR in the Mikrotik is still good (36 dB). ch3 must have good signal, as the combination of ch0,ch1 and ch2 is very weak.
-retransmits 192000/130000 becoming 30% already. But 5 sec average P-throughput is 508Mbps

Interpretation: similar to iPhone. Stronger (better SNR) but more distorted signal. MCS rate goes higher but has more misses.

Device 3 : computer
-transmit 520 Mbps HT MCS 13/MCS05/64QAM. Looks reasonable but there is something wrong here. (P-throughput only 92 Mbps)
-received strong signal -59 dBm in the range that is considered a good signal. The rate is accordingly 702Mbps what is VHTMCS08/2S/LGI/256QAM. SNR 45 dB. For 256QAM you need a pure signal.
-retransmits 2589000/2143000 as the iPhone

Interpretation.:
This not good. What is going wrong here in the path Mikrotik->computer? The other way is a very good signal. (Computer have a TX power which is much higher than a tablet or smartphone. It also depends on the adaptor setting in the computer) What is the computer receiving? What is it’s location compared to iPhone/iPad ? TX and RX paths in the air are supposed to be the same. What causes this poor CCQ of 64 and P-throughput of 92Mbps? The information for the path Mikrotik->computer is in the computer only. (If the computer is replaced with a Mikrotik in station mode, then the registration would also have shown the received TX signal information.

Next steps? Trial and check how those numbers move.

• position of devices, the router placement is very important
• collect wifi information from devices (will also learn if there is room to lower the TX power at the Mikrotik)
• test with 2 chains and with modified TX power at the router (devices also adapt to the announced TX power in the beacons of the Mikrotik)
• check how agressive the TX rate selector is. If it is very aggressive lower the hw_retry=3, and check again

Ok thanks!
I know that only in 80MHz mode the theoretical max is 860Mbps… However I will be happy with the half of that… (My ISP is only 500Mbps and usually handheld Wi-Fi devices use just the internet, not local network).
I don’t know what’s wrong with the Computer. It is “almost” in clear sight of the router, around 10-15m away from the router, it needs to pass only max 1 wall. The iPad and iPhone was on the table (locked so that’s why it drops the rates - I saw that) which is almost 1m away from the computer antennas. However from the table where the phone was, you have a clear sight to the router (I can clearly see that router and nothing crosses this path).

The problem might be which I experienced just today:
The router is located in a corner, as I said, around 1.5m above the ground, and around 20-20 cm from each wall (it is a corner!). Also there is a cabinet next to it which might be a problem.
How I know that: I have an older Mikrotik router also (RB751) acting as a Wi-Fi client, because I have some devices at the other end of the flat which needs internet connection and they only has wired connection, so that I use this old router as a Wi-Fi client.

Today I rotated the router just a little bit (only a few degree and cm) so I might think that this won’t affect the performance, but:
The 5GHz performance stayed the same or almost the same on every device, however that AP Client Mikrotik router (which uses 2.4GHz) lost connection, the connection became really instable. CCQ dropped to 30/30% (and yes here I can see both TX/RX CCQ), and I was barely able to connect to it with Winbox. Rotating the router back, just a few degree, CCQ changed back to 70/40%!!
This is not still not the best, but these devices don’t need that much speed - they only send a few bytes every minute - so I don’t really care. But Winbox works instantly now. Just by rotating this little…
Which why I ask (and I have read this in many places): it is better to place the router in an upper position? I’m able to place the router on the top of that cabinet which is around 2.1m high. I wonder if this will make sense. Lots of article says that you should try to place it as high as you can… Unfortunately, I can’t do anything about the corner - but this cabinet is also a little bit further away from the corner. So I might try that.

Regarding the PC speed: I don’t know why that happens, checking now, it dropped to 50Mbps. Yesterday I got 210Mbps. I only tried rotating the router as I said, I didn’t change anything in the config…

Do I understand well that that RB751 is close to the RB4011? That can probably be a problem, even if they are on total different frequencies. Electronic boards with radio’s on do emit a lot at different frequencies. Even the passive antenna will work as Yagi antenna for the RB4011. Placing the RB4011 on top of that cabinet is a great idea. It’s further away from anything else there.

That corner is not a lucky thing, it will surely disturb the wifi signal.. But you have no other choice That RB751 as station is a transmitter like an AP (it just does not send out beacons), if you are not using other virtual interfaces as AP (because then it is just an AP also). That RB751 would be a good test device for wifi. The “registration” would also give the remote signal strength in the RB4011. Unfortunatly 2.4 GHz only.

Did you try “snooper” yet? I had a case where a nearby wAP ac appeared in snooper but at phantome frequencies. The wAP ac was only 50 cm from the testing hAP ac2 on the same table. Moving the devices removed the phantom frequency immediatly.

Keep measuring. Even life-measurements when you move or rotate things. If you click on the registration, and open it, the “signal” tab gives even more information.
Most of the Mikrotik quality indicators cannot be trusted if there is not enough traffic. A connection that is not actively used will show bad performance numbers!

https://7signal.com/10-wi-fi-antenna-placement-mistakes/

Ok thanks! I was able to solve the P throughput problem on my PC, I have relocated its antennas. Now I have P Throughput around 350000. I think this seems much better. TX rate 390Mbps (almost same) RX rate 702Mbps. Signal is still -61 dB
True throughput still around 210Mbps.

No, the RB751 is around halfway as my computer (in distance and in direction as well), so it is not that close, also in a different room.
In the weekend I will try to relocate the router (I think this might be the root problem, because even small changes of the router location at that point causes significant differences).
Also I will try Snooper and see if I can understand something better.

[rkrisi]

Can get up to 450mbps using btest in my hap ac2, this is my config if it helps

/interface wireless security-profiles
set [ find default=yes ] supplicant-identity=MikroTik
add authentication-types=wpa2-psk disable-pmkid=yes eap-methods="" \
    group-key-update=1h mode=dynamic-keys name=main supplicant-identity=""
/interface wireless
set [ find default-name=wlan1 ] amsdu-limit=2048 amsdu-threshold=2048 band=\
    2ghz-b/g/n country="united states" disabled=no distance=indoors frequency=\
    2462 installation=indoor keepalive-frames=disabled mode=ap-bridge \
    multicast-buffering=disabled multicast-helper=disabled security-profile=\
    main ssid=network wireless-protocol=802.11 wps-mode=disabled
set [ find default-name=wlan2 ] amsdu-limit=2048 amsdu-threshold=2048 band=\
    5ghz-a/n/ac channel-width=20/40/80mhz-Ceee country="united states" \
    disabled=no distance=indoors frequency=5660 installation=indoor \
    keepalive-frames=disabled mode=ap-bridge multicast-buffering=disabled \
    multicast-helper=disabled security-profile=main ssid=network-5g \
    wireless-protocol=802.11 wps-mode=disabled

PD: I live in front of some old buildings with lots of WISP and 4G antennas, the non-DFS spectrum is extremely crowded as 2.4ghz is, so using a free DFS channel helped me

Thanks! I also think there is no copy paste solution, but I have some things which is really different from yours. And I don’t know if it will make any sense.
Like amsdu-limit/threshold, multicast-helper. buffering, etc… Why did you set those?

[ksuuk]

Like amsdu-limit/threshold, multicast-helper. buffering, etc… Why did you set those?

There is no clear answer, seems it helps in some cases stability and some clients (ios) connection, for me amsdu 2048 made things worst. For example my current config is:

set [ find default-name=wlan1 ] adaptive-noise-immunity=ap-and-client-mode band=2ghz-g/n basic-rates-a/g=12Mbps,24Mbps basic-rates-b="" channel-width=20/40mhz-Ce country=no_country_set disabled=no \
    distance=indoors frequency=2437 frequency-mode=manual-txpower hw-retries=5 installation=indoor keepalive-frames=disabled max-station-count=300 mode=ap-bridge multicast-buffering=disabled \
    multicast-helper=disabled rate-set=configured security-profile=xb ssid=xb supported-rates-a/g=12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps supported-rates-b="" wireless-protocol=802.11 wmm-support=\
    enabled wps-mode=disabled
set [ find default-name=wlan2 ] adaptive-noise-immunity=ap-and-client-mode antenna-gain=1 band=5ghz-a/n/ac basic-rates-a/g=12Mbps,24Mbps channel-width=20/40/80mhz-Ceee country=no_country_set \
    disabled=no distance=indoors frequency-mode=manual-txpower installation=indoor keepalive-frames=disabled max-station-count=300 mode=ap-bridge multicast-buffering=disabled multicast-helper=\
    disabled rate-set=configured security-profile=xb ssid=xb supported-rates-a/g=12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps wireless-protocol=802.11 wmm-support=enabled wps-mode=disabled

And like I said, strange and unlogic, but true in my case, disabling route cache and fasttrack gives better speed for wifi, however, as my wan is 4G, which speed is much lower then Your, then it may not help You at all. But You never know, until You try.

What makes me wonder is Mikrotik stability problem, for example, yesterday evening I played a bit with amdsu settings and speed was 5 m away kind of 30/30 (yes, it’s still pathetic, as wan is about 80/50). Today morning speed was 5/9, tried about 5 times, same result, then via Mikrotik app I opened same amdsu setting, there was 8192, deleted last octet, inserted again and saved (so actually I saved config without any actual change) and speed was again 25/25.

What is going on? Does Mikrotik fall into some standby mode or what?

[rkrisi]

These 3 devices are used for testing. Totally random screenshot, so I have not adjusted anything.

Ok , rkrisi (Kristof) , thanks for sharing. There is normally a constant or even real time monitoring needed, but lets start with a little comment on this registration table.

Elements to consider:

• the ultimate max we can expect from this equipment (AP + clients) is “866.7 Mbps/80 MHz/2S:SGI”, this with a stron clean signal in both directions with 256 QAM encoding at VHT MCS 9.
  We are operating in the yellow part of this generic 802.11 table: https://d2cpnw0u24fjm4.cloudfront.net/wp-content/uploads/802.11-Data-Rates-and-SNR-Requirements-download.pdf
  Telling us that we should have at least a SNR of 35, what comes with a signal of -51 dBm. (Mikrotik has its own table , and even documents receive sensitivity in the product specs)
  Whith that 866Mbps interface rate one can expect around 500 Mbps-600 Mbps net throughput in one dedirection.
• a second element is a mismatch in the RouterOS commands and a 4x4 device. RouterOS GUI only handles 3x3 (ch0,ch1,ch2). Everything in ch3 is missing. Even more, the wiki says if you want to use the 4 chains, leave “Data rates” in default. Only recent change logs did mention 4x4 support through CLI.

Just about your table, the observation learns something (but of course here the location and orientation, and the numbers on the client devices are missing). Also missing is the rate of variation of those numbers. I thought I found a tool to log this over time, but lost track of it. There are script to log this in the wiki.

Device 1 : iPhone
-transmit rate selector is at (390 Bbps) HT MCS12/MCS04/16QAM. The selector has dropped till a robust signal (16QAM)
-receive rate is set at (585 Mbps) HT MCS14/MCS06/64QAM. This requires a signal of -59 dBm and a SNR ratio of 26, we have a signal of only -70 dBm, coming from the combined ch signals, including ch3. I don’t know if the 2 bests are combined, or all 4.

• the retransmits are acceptable (249000/216000) but of course eat away some of the capacity (P-throughput : supposed to be averaged over 5 seconds) 305Mbps

Interpretation: quite good, iPhone is at the end of its wifi range (unless the iPhone is conserving batteries and has reduced its transmit power). The 12% misses make the router drop the MCS rate. An iPhone is normally handheld, and that hand together with the human body is an absorber for wifi signals. Unclear what the signal strength is that the iPhone receives.

Device 2: iPAD
-transmit 585Mbps HT MCS14/MCS06/64QAM
-received the same numbers but with a similar signal of -68 dBm but the SNR in the Mikrotik is still good (36 dB). ch3 must have good signal, as the combination of ch0,ch1 and ch2 is very weak.
-retransmits 192000/130000 becoming 30% already. But 5 sec average P-throughput is 508Mbps

Interpretation: similar to iPhone. Stronger (better SNR) but more distorted signal. MCS rate goes higher but has more misses.

Device 3 : computer
-transmit 520 Mbps HT MCS 13/MCS05/64QAM. Looks reasonable but there is something wrong here. (P-throughput only 92 Mbps)
-received strong signal -59 dBm in the range that is considered a good signal. The rate is accordingly 702Mbps what is VHTMCS08/2S/LGI/256QAM. SNR 45 dB. For 256QAM you need a pure signal.
-retransmits 2589000/2143000 as the iPhone

Interpretation.:
This not good. What is going wrong here in the path Mikrotik->computer? The other way is a very good signal. (Computer have a TX power which is much higher than a tablet or smartphone. It also depends on the adaptor setting in the computer) What is the computer receiving? What is it’s location compared to iPhone/iPad ? TX and RX paths in the air are supposed to be the same. What causes this poor CCQ of 64 and P-throughput of 92Mbps? The information for the path Mikrotik->computer is in the computer only. (If the computer is replaced with a Mikrotik in station mode, then the registration would also have shown the received TX signal information.

Next steps? Trial and check how those numbers move.

• position of devices, the router placement is very important
• collect wifi information from devices (will also learn if there is room to lower the TX power at the Mikrotik)
• test with 2 chains and with modified TX power at the router (devices also adapt to the announced TX power in the beacons of the Mikrotik)
• check how agressive the TX rate selector is. If it is very aggressive lower the hw_retry=3, and check again

Ok thanks!
I know that only in 80MHz mode the theoretical max is 860Mbps… However I will be happy with the half of that… (My ISP is only 500Mbps and usually handheld Wi-Fi devices use just the internet, not local network).
I don’t know what’s wrong with the Computer. It is “almost” in clear sight of the router, around 10-15m away from the router, it needs to pass only max 1 wall. The iPad and iPhone was on the table (locked so that’s why it drops the rates - I saw that) which is almost 1m away from the computer antennas. However from the table where the phone was, you have a clear sight to the router (I can clearly see that router and nothing crosses this path).

The problem might be which I experienced just today:
The router is located in a corner, as I said, around 1.5m above the ground, and around 20-20 cm from each wall (it is a corner!). Also there is a cabinet next to it which might be a problem.
How I know that: I have an older Mikrotik router also (RB751) acting as a Wi-Fi client, because I have some devices at the other end of the flat which needs internet connection and they only has wired connection, so that I use this old router as a Wi-Fi client.

Today I rotated the router just a little bit (only a few degree and cm) so I might think that this won’t affect the performance, but:
The 5GHz performance stayed the same or almost the same on every device, however that AP Client Mikrotik router (which uses 2.4GHz) lost connection, the connection became really instable. CCQ dropped to 30/30% (and yes here I can see both TX/RX CCQ), and I was barely able to connect to it with Winbox. Rotating the router back, just a few degree, CCQ changed back to 70/40%!!
This is not still not the best, but these devices don’t need that much speed - they only send a few bytes every minute - so I don’t really care. But Winbox works instantly now. Just by rotating this little…
Which why I ask (and I have read this in many places): it is better to place the router in an upper position? I’m able to place the router on the top of that cabinet which is around 2.1m high. I wonder if this will make sense. Lots of article says that you should try to place it as high as you can… Unfortunately, I can’t do anything about the corner - but this cabinet is also a little bit further away from the corner. So I might try that.

Regarding the PC speed: I don’t know why that happens, checking now, it dropped to 50Mbps. Yesterday I got 210Mbps. I only tried rotating the router as I said, I didn’t change anything in the config…

Do I understand well that that RB751 is close to the RB4011? That can probably be a problem, even if they are on total different frequencies. Electronic boards with radio’s on do emit a lot at different frequencies. Even the passive antenna will work as Yagi antenna for the RB4011. Placing the RB4011 on top of that cabinet is a great idea. It’s further away from anything else there.

That corner is not a lucky thing, it will surely disturb the wifi signal.. But you have no other choice That RB751 as station is a transmitter like an AP (it just does not send out beacons), if you are not using other virtual interfaces as AP (because then it is just an AP also). That RB751 would be a good test device for wifi. The “registration” would also give the remote signal strength in the RB4011. Unfortunatly 2.4 GHz only.

Did you try “snooper” yet? I had a case where a nearby wAP ac appeared in snooper but at phantome frequencies. The wAP ac was only 50 cm from the testing hAP ac2 on the same table. Moving the devices removed the phantom frequency immediatly.

Keep measuring. Even life-measurements when you move or rotate things. If you click on the registration, and open it, the “signal” tab gives even more information.
Most of the Mikrotik quality indicators cannot be trusted if there is not enough traffic. A connection that is not actively used will show bad performance numbers!

https://7signal.com/10-wi-fi-antenna-placement-mistakes/

Now I have tried snooper.
It reported a few channels which don’t have any bandwidth at all, I have selected one and I get a slightly better results now. Around 10% better. But still far from what I want.
Next I will try to move the router a little bit…

[mmbeck]

I also did some tests on the RB4011 5 GHz performance this weekend (using ROS 6.47). It looks like it does not like to transmit data. Using iperf3 over the wifi-connection, the RB4011 was able to receive about 300 MBit/s from a 2x2 client (Intel AC7260), but it only managed to send 80 Mbit/s back to the client. Both under almost perfect radio conditions. I also checked the same client’s performance on several last-generation 802.11ac wave 1 APs from different manufacturers, and always got better performance.

Connecting two RB4011, standing on the same table, on 5 GHz, it looked that the one in AP mode was really slow in increasing TX rate, while the one acting as station seemed to behave about normal. Running Ping Speed over the 5 GHz link for a few minutes, they finally achieved a decent average speed, but the overall experience in realistic scenarios was not satisfactory.

Hopefully, new drivers will fix this at some point.

[rkrisi]

I also did some tests on the RB4011 5 GHz performance this weekend (using ROS 6.47). It looks like it does not like to transmit data. Using iperf3 over the wifi-connection, the RB4011 was able to receive about 300 MBit/s from a 2x2 client (Intel AC7260), but it only managed to send 80 Mbit/s back to the client. Both under almost perfect radio conditions. I also checked the same client’s performance on several last-generation 802.11ac wave 1 APs from different manufacturers, and always got better performance.

Connecting two RB4011, standing on the same table, on 5 GHz, it looked that the one in AP mode was really slow in increasing TX rate, while the one acting as station seemed to behave about normal. Running Ping Speed over the 5 GHz link for a few minutes, they finally achieved a decent average speed, but the overall experience in realistic scenarios was not satisfactory.

Hopefully, new drivers will fix this at some point.

I’m still on 6.46.6 and I hoped 6.47 will come with some improvements! Anyway thanks for your detailed tests. However I’m still not able to reach even 300Mbps on 5GHz with RB4011.
Which is a shame because a friend of mine bought an hAP ac2 almost the same time when I bought the RB4011 and he gets much better speeds with that one (around 450Mbps down).
I have read that hap ac2 had similar problems before and updates slowly fixed it. I don’t know if this is the case with the RB4011 also. Looking at the specs sheets, it seems that besides both router is arm based, nothing else matches between the 2 routers (completely different wifi chips, cpu, etc…). So I just can hope that this will improve over time.
One thing I did not test is moving the router in the house (which is not that easy for me), but I will test it soon. I will be happy if I can get around 300-400Mbps speed…

Also I can somehow confirm you. I only have one RB4011 and sometimes I see when starting a speedtest it start at around 50Mbps and slowly increases for around 210Mbps. This happens 1-2 times out of 10. In other cases the speedtest jumps to around 250Mbps and slowly decreases until 210Mbps…
Once I was able to get around 335Mbps speed, but I don’t know why and how… I just tested it randomly (nothing changed) and it jumped to 450Mbps and avaraged around 335Mbps down. However right after starting a new test (same position) I was not able to get more than 250Mbps… So it really fluctuates in speed and I don’t know why…

[WeWiNet]

Setup:

• Near range (2meter) between client and AP
• iperf3 on Android / MacOS
• iperf3 on Rasp4 with Ethernet connected to RB4011
• RB4011 4x4 ROS 6.46.6, reset to default, 20/40/80MHz channel and Wifi security added (to avoid also anyone connecting to the unit)
• Add also the country in Wifi interface to ensure I do testing which actually means something.
• No other ROS changes done.
  Any other setup than above might give wrong results… !!!

Average throughput over 30 seconds:

• 520Mbps TCP DL with 3x3 client
• 400Mbps TCP DL with 2x2 client

If you do that test and get lower numbers, then your Wifi environment is crowded (mine is not really in 5Ghz) or your unit is defective…
I guess its rather first one for you…
PS: DFS channels might play or not a role in your country. I often do test in the non-DFS channels but those are the most crowded ones
in urban environment problematic.

[rkrisi]

Setup:

• Near range (2meter) between client and AP
• iperf3 on Android / MacOS
• iperf3 on Rasp4 with Ethernet connected to RB4011
• RB4011 4x4 ROS 6.46.6, reset to default, 20/40/80MHz channel and Wifi security added (to avoid also anyone connecting to the unit)
• Add also the country in Wifi interface to ensure I do testing which actually means something.
• No other ROS changes done.
  Any other setup than above might give wrong results… !!!

Average throughput over 30 seconds:

• 520Mbps TCP DL with 3x3 client
• 400Mbps TCP DL with 2x2 client

If you do that test and get lower numbers, then your Wifi environment is crowded (mine is not really in 5Ghz) or your unit is defective…
I guess its rather first one for you…
PS: DFS channels might play or not a role in your country. I often do test in the non-DFS channels but those are the most crowded ones
in urban environment problematic.

Did similar tests today. Only I did not reset the config to default on RB4011, but everything is similar. iPerf3 server on Raspberry 4, 2x2 client close to router, I got around 250Mbps speed…

[WeWiNet]

I implicitly assumed this, but to be sure, I use 80 MHz capable clients.
You can check in the Wireless “registration” tab that they do receive and sent with 80 Mhz
and 2 spatial stream.

[rkrisi]

I implicitly assumed this, but to be sure, I use 80 MHz capable clients.
You can check in the Wireless “registration” tab that they do receive and sent with 80 Mhz
and 2 spatial stream.

Same clients here (I get around 800Mbps Rx rate 80Mhz/2SP/SGI)

[WeWiNet]

Exactly. AP RX rates I get >1000Mbps with MacPro 3x3 but on AP TX I rarely go over the 800Mbps rates…
So client UL is a but better than DL.
On the 2x2 client I get 800mbps rates on both Rx and Tx.

PS: I keep AMPDU priority as default (only “1” checked). I see bad performance when enabling the others.
I really keep all “Wifi” settings to default.

[bpwl]

PS: I keep AMPDU priority as default (only “1” checked). I see bad performance when enabling the others.

A surprising performance hit, bad performance with A-MPDU enabled. These are needed for high performance throughput, by avoiding a lot of overhead.
If most is default (no extra added mangle rules to set the priority) then all is priority 0 , and only the priority zero setting for A-MPDU is used, and that is enabled by default.
Mikrotik default configurations never set the priority.
Aggregation is good for throughput (reduced overhead), its bad for latency and jitter. So it should not be set for priority 6 and 7 (eg. voiceIP)

Expected performance without A-MPDU set: http://forum.mikrotik.com/t/how-to-get-rid-of-slow-wifi-can-it-be-done/139640/1

[rkrisi]

Exactly. AP RX rates I get >1000Mbps with MacPro 3x3 but on AP TX I rarely go over the 800Mbps rates…
So client UL is a but better than DL.
On the 2x2 client I get 800mbps rates on both Rx and Tx.

PS: I keep AMPDU priority as default (only “1” checked). I see bad performance when enabling the others.
I really keep all “Wifi” settings to default.

I did not change AMPDU priorities and I have 0 checked.
So you say you only have “1” checked?

Anyway it seems that the placement of the router is also wrong. Did iperf tests again from different positions, and some place I was able to get around 350Mbps average. So I will try to place it somewhere else first.
What I don’t know is that other (online) speedtest tools give really fluctuating performance (which I can sometimes see on iperf also). I know that this is an online service, etc but my Raspberry Pi does speedtest tests on a selected server periodically and it always gives me 500Mbps (through cable) which is my ISP maximum. Same server on a wireless client, max 210Mbps. iPerf sometimes gives 350Mbps if I can place the wireless device correctly…

[rkrisi]

Well, another observation through longer time… It seems that only speedtest is slow(er), using P2P connection (like torrent client) is much faster, even on my personal computer 15m away from the router.
Speedtest (does not matter which - ookla, fast.com, etc… or even local iperf) usually gives 180-190Mbps here, torrent client can reach up to 36-37MB/s which is almost 300Mbps… Still far away from 500Mbps which I want to achieve, but this is much better. Though testing it like this is much harder, so the question is, why speedtest is slower than other connections?
I don’t care if the speedtest is not as fast as the actual connection is, but I might try again different configs, channels, etc and would be great if I could easily test it with a proper tool…

[WeWiNet]

This is a post about Wifi problem, is your question still Wifi related?

If so you CAN NOT test Wifi connection with Ookla Speedtest. Well you can, but your results do not say anything
about the Wifi, but about the complete system.
So the only way to do this is test Wifi connection between the AP and the client ONLY, and enusre othe elements used
to test support at least max. Wifi speed (or 1Gbps).
Such a test will show you your Wifi performance and then you can start playing around (a bit) with Mikrotik Wifi parameters.
Use Iperf3 is simplest and most reliable solution if you do not have professional Wifi test equipment.

Giving advice based on results obtained via “speedtest” is in my sense point less…

If you talk about ethernet speed, then torrent do use many connections in parallel, which is often better to “load the system” up to its maximum,
versus “single stream” testing. Even compared to ookla “multi” setting, the torrent will have many more connections.
This could be reason you see highest speed with torrent (and they all use UDP, which is simpler/faster than TCP which is used by Speedtest…!!!)

[anav]

I see lots of reference to theoretical speeds, 867, 1300, etc… Remember its marketing hype as those are two way added speeds.
The most ludicrous is the term ooh aahh 1750 which is adding both 2.4 and 5ghz two way streams 450+1300.
Or 867+300 = ~1200 on the packaging.

Consider 1300 for example, its two way tx and thus when we are talking uploading or downloading one should be discussing 650-prop losses as the starting point.
Using the MTUNA certified 1/3 rule any turkey looking to get beyond 433 speed, needs to be bitten by a llama and endure his/her flatulence!! Sure under ideal conditions of no interference and LOS on 5ghz… but most have competing elements in the home of all sorts from house wiring, type of lighting, appliances etc (worse in apartments) and walls of unknown characteristics. Further as has been noted a mix of wifi devices can seriously impinge upon ideal performance. Obviously 2.4ghz due to progopation will travel further and be more useful over a longer range than 5ghz. In fact 5ghz in a house of any size is useless as a single entity and one needs multiples regardless of vendor. Even a single 2.4 may not be suitable for a larger home.

The frustration is that I can setup a TPlink eap245 plugNplay in the place of a capac, and without any fuss at least equal and surpass the cap performance without any tuning (to be fair it has more chains too). Mind you the capac with routerOS has many other pluses and capabilities that the TP link lacks. The key is for MT to ensure that the wifi is either EASY to configure (comes with a stock setting equivalent to other vendors plugNplay) and what I was hoping for most, was the ability to tailor those settings (one size fits all) to size to my location. However even an expert and generous man bpwl, with great knowledge and patience is still befuddled by the lack of information available. So concur, that MT is not doing a service to its customers at the moment on the wifi side. It seems they have made a half-hearted effort to provide some features but not all and no instructions… and then gave up on putting any more resources into wifi (well other than into products to sell).
I have stated this before … they should hire bpwl on a contract and let him work with the programmers and come up with a useful guide as well as any changes need to the code to ensure that users can optimize their devices.

[rkrisi]

This is a post about Wifi problem, is your question still Wifi related?

If so you CAN NOT test Wifi connection with Ookla Speedtest. Well you can, but your results do not say anything
about the Wifi, but about the complete system.
So the only way to do this is test Wifi connection between the AP and the client ONLY, and enusre othe elements used
to test support at least max. Wifi speed (or 1Gbps).
Such a test will show you your Wifi performance and then you can start playing around (a bit) with Mikrotik Wifi parameters.
Use Iperf3 is simplest and most reliable solution if you do not have professional Wifi test equipment.

Giving advice based on results obtained via “speedtest” is in my sense point less…

If you talk about ethernet speed, then torrent do use many connections in parallel, which is often better to “load the system” up to its maximum,
versus “single stream” testing. Even compared to ookla “multi” setting, the torrent will have many more connections.
This could be reason you see highest speed with torrent (and they all use UDP, which is simpler/faster than TCP which is used by Speedtest…!!!)

Yes this is still a question about Wi-Fi.
Yes, I know how TCP/UDP works… and I know that you can’t trust speedtest on this, but iperf seems to perform worse than it can actually handle. iperf gives around 230 Mbps (local server - the server is connected wired to the router - so no 2 clients using Wi-Fi extensively), which is the same as speedtest…
Wired client can easily use 500 Mbps even on ookla speedtest (which is my max isp speed).
But I slowly accept that I can’t get better results with this Wi-Fi chip… also what I can’t understand is that using a COMPLETELY empty channel 100-116 gives worse results than it does on 52-64 which always has 1-2 AP on it as well…

[rkrisi]

I see lots of reference to theoretical speeds, 867, 1300, etc… Remember its marketing hype as those are two way added speeds.
The most ludicrous is the term ooh aahh 1750 which is adding both 2.4 and 5ghz two way streams 450+1300.
Or 867+300 = ~1200 on the packaging.

Consider 1300 for example, its two way tx and thus when we are talking uploading or downloading one should be discussing 650-prop losses as the starting point.
Using the MTUNA certified 1/3 rule any turkey looking to get beyond 433 speed, needs to be bitten by a llama and endure his/her flatulence!! Sure under ideal conditions of no interference and LOS on 5ghz… but most have competing elements in the home of all sorts from house wiring, type of lighting, appliances etc (worse in apartments) and walls of unknown characteristics. Further as has been noted a mix of wifi devices can seriously impinge upon ideal performance. Obviously 2.4ghz due to progopation will travel further and be more useful over a longer range than 5ghz. In fact 5ghz in a house of any size is useless as a single entity and one needs multiples regardless of vendor. Even a single 2.4 may not be suitable for a larger home.

The frustration is that I can setup a TPlink eap245 plugNplay in the place of a capac, and without any fuss at least equal and surpass the cap performance without any tuning (to be fair it has more chains too). Mind you the capac with routerOS has many other pluses and capabilities that the TP link lacks. The key is for MT to ensure that the wifi is either EASY to configure (comes with a stock setting equivalent to other vendors plugNplay) and what I was hoping for most, was the ability to tailor those settings (one size fits all) to size to my location. However even an expert and generous man bpwl, with great knowledge and patience is still befuddled by the lack of information available. So concur, that MT is not doing a service to its customers at the moment on the wifi side.
I have stated this before … they should hire bpwl on a contract and let him work with the programmers and come up with a useful guide as well as any changes need to the code to ensure that users can optimize their devices.

Yes I know and I never assumed that I could get even 1Gbps Wi-Fi speed with this device… Though 400-500 Mbps should be there on 5Ghz I think.