Folks
My apologies for the newbie question. I own an RB962, which runs quite well, not with optimal performance because I'm using auto frequency selection, with download rates near 300mbps, though. This is the bandwidth I have with the ISP.
Last week I bought a cAP AC to improve the signal coverage on the 1st floor and replace a TP-Link repeater. Configured CAPsMAN, set both the router and the AP WLAN interfaces to CAP. The 2.4ghz is working ok. I don't even bother about speed because it's used primarily for IoT and home automation.
But I'm struggling with the 5ghz. I've tried many different channel configurations, channel extension, always setting 20MHz, and the max RX I got so far with BTest (standing 2m always from the antenna) was 227mbps, which is ok, but then it dropped to 90mbps. I was utilizing 5765/20/eCee.
Does anyone have any recommendations on channel configuration or anything else I can do in CApsMAN to improve 5ghz?

I appreciate any advise.
Cheers.

Your max TX power will be limited to country regulations (and for channels above 5725MHz for some countries this is 14 dBm, including antenna gain) only, and that is very low.
http://bowdennetworks.co.uk/downloads/5 ... -%20v1.pdf
See "/interface wireless info country-info" for your dBm limitation.

5765MHz/20MHz-eCee is a 80 MHz wide channel (149-153-157-161).

Look in the registration tabel (or in CAPsMAN) to see what interface rate you get ... eg "867Mbps-80MHz/2S/SGI" ... this is dual spatial stream,VHT MCS 09, short guard (https://mcsindex.com/), and is the max possible here.
Stand alone you can expect 360Mbps UDP unidirectional data rate. 400Mbps-40MHz/2S/SGI would give 260Mbps unidirectional UDP data throughput, but the signal density is double.

All depends on your surrounding RF signals. Use "Snooper" in stand alone mode to look for other interfering wifi transmitters.

You won't ever achieve performance (by 2021 standards) with CAPsMAN.
It's not a matter of fine-tuning, it is just by design for now.

We recently experienced quite a lot and discovered that CAPsMAN performance (being local forwarding or CAPsMAN encapsulation) is very different (and much worse) than standalone mode (independant APs configured in their own RouterOS)

Why exactly i can't say, probably cause existing CAPsMAN system has deep implications in drivers (all drivers are proprietary with Mikrotik) with a lot of legacy code, and the AP-controller communication protocol is obviously over-weighted and might has too many inter-dependant hooks even for simple forwarding job.

I agree however that CAPsMAN -can- be fine for setups that do not require extreme throughtput or instant-roaming between APs.
It just doesn't come close to performance you can expect with given chipsets (IPQ4019,QCA9882...)

On the bright side, wifiwave2 initiative looks like a parallel attempt to reconcile with generic SoC drivers, and i really hope a new (and lighter) CAPsMAN mode comes one day, which will operate the same as competition (Unifi/OMADA) :
- remote provisionning
- log collection
- while the APs remain independant, with optimized drivers, for network & forwarding tasks

As in another post I've made, got low 5GHz performance on cAP ACs. I get that there is additional routing through the CAPsMAN system. Inference is that this is a software/driver issue. Any better with v7?

I tried with v7.1.1 and no good.
As I have just two APs, I decided to set them as stand-alone. Now I get almost the same as the cable, when close to the AP.

As in another post I've made, got low 5GHz performance on cAP ACs. I get that there is additional routing through the CAPsMAN system. Inference is that this is a software/driver issue. Any better with v7?

Why are you asking?
The capsman is older technology and older chipsets that version 7 does not and will never address, - they have moved onto audience and capacXL products and newer technology/chipsets to support.
(by newer I mean technology that everyone else has been using for at least 5 years)

Why are you being so rude? If you've got nothing useful to say, don't say anything. Not everyone has your god like knowledge of Router OS...neither does everyone need the latest state of the art equipment like Wi-Fi 6. Millions of installations are still working quite happily on Wi-Fi 4 esp. as the average internet speed in the UK on FTTC is often less than Wi-Fi 4's maximum speeed.

All you needed to say is that CAPsMAN adds an overhead to processing Wi-Fi packets which reduces throughput. This overhead could be acceptable if you're not overly concerned about throughput, i.e. 120Mbps is okay for your needs and the ease of management that CAPsMAN brings outweighs the overload. In the case I'm talking about with just 4 access points, configuring them manually is not a bit problem.

Because CAPsMAN is clearly software, then it was a perfectly reasonable question to ask if v7 improved this. I know it's been an almost complete re-write with a newer Linux kernel. The Wi-Fi chipset in CAPs is clearly capable of reasonable Wi-Fi speeds as I've disabled CAPsMAN on one of the access points and I'm now getting >200Mbps which considering the laptop I'm using is going through a wall to the AP, then that's acceptable.

My advice is blunt, and practical. I see your polluting the forums on different threads with the same questions that 1000 people before you have already gone through and wasted their time and made very little to zero progress. Take it or leave it............ Can lead a donkey to water
(If I was rich I would send everyone who comes to these forums free business class APs from other vendors in order to ensure they have a good experience and dont waste an iota of their time)

I'm getting 400mbit download and 500mbit upload 3m from wap ac or cap ac via capsman in local forward, connected to samsung s21 for example.(using local test server)
80mhz, n/ac, nothing special confing...

Capsman is on v7.2RC3 and caps on v6 long therm.

I'm getting 400mbit download and 500mbit upload 3m from wap ac or cap ac via capsman in local forward, connected to samsung s21 for example.(using local test server)
80mhz, n/ac, nothing special confing...

Capsman is on v7.2RC3 and caps on v6 long therm.

Very interesting seeing that the advertised 2 way speed of a capac is 867 and thus one way the theoretical max possible BEFORE LOSSES is 433.
Not saying your a bold face liar but I dont think your testing setup is credible.

I'm getting 400mbit download and 500mbit upload 3m from wap ac or cap ac via capsman in local forward, connected to samsung s21 for example.(using local test server)
80mhz, n/ac, nothing special confing...

Capsman is on v7.2RC3 and caps on v6 long therm.

Very interesting seeing that the advertised 2 way speed of a capac is 867 and thus one way the theoretical max possible BEFORE LOSSES is 433.
Not saying your a bold face liar but I dont think your testing setup is credible.

A 2 way speed? What are you even talking about I hope you didnt understand that I get this both way same time? I can record you ap, laptop and wifi interface of ap inside winbox doing this bandwidth on video if you don't belive.

Very interesting seeing that the advertised 2 way speed of a capac is 867 and thus one way the theoretical max possible BEFORE LOSSES is 433.

@anav, you've got this wrong. WiFi is time-division duplex, which means traffic can flow only in one direction at any given time. But nowhere is limited to divide uplink and downlink in equal ratio (which is what you're saying). In contrary, since WiFi is using random access (and optionally RTS/CTS), it is possible to use almost 100% of air time for one direction. For TCP transers the return direction only needs around 2% of bandwidth (for ACKs), which might translate into 10% of air time. But that still leaves around 90% of air time for "busier" direction, which may mean around 790Mbps gross throughput.

The "half" rule of thumb comes from calculations of overhead and packet timigs ... and is only rule of thumb, depending on wireless settings it is realisticaly possible to get more than 50% net. @bpwl might be able to tell realistic max net throughput.

Not surprised with those good results. Is somewhat the max that RouterOS can do in a clean area with it's smaller buffer sizes.
1/2 PHY (interface rate) speed is a rule of thumb. One can get somewhat better, but MT hits some other limit on the highest interface rates.
Detailed calculations are available, and A-MPDU aggregation is very important in this. (MPDU (3839) and A-MPDU (262143) are rather small as seen in the beacons for MT)
viewtopic.php?t=174294#p852582 ; viewtopic.php?f=7&t=165698#p817088
But still it gets higher than 1/2 as we see in other devices (or with the wifiwave2 drivers), but I only reached 360Mbps on 866Mbps/2S/SGI connections on MT.

wAP ac (mipsbe, 1 ethernet port ) is a 3S device. So the interface rate can go up to 1300Mbps/3S/SGI It's important to check the "Registration" table when testing.

To Test: would lowering the max # of stations (from 2007) allocate larger buffers ????

I'm getting 400mbit download and 500mbit upload 3m from wap ac or cap ac via capsman in local forward, connected to samsung s21 for example.(using local test server)
80mhz, n/ac, nothing special confing...

"Local forward" seems not to be using the AMSDU=2048 as is assumed to be set by CAPsMAN, or it would not be possible to get that download speed. (Upload is using the buffers in the client device, =others)

"with 3 and you get speeds like 434Mbps for MT and 776Mbps for others.

And now we optimize when the conditions are good , rising the basic rate from 6Mbps to 24Mbps .... becomes 544Mbps and 844Mbps for triple stream"

viewtopic.php?t=165698#p912622
Setting Protection is off gains some air-time and throughput.