1. cAP ax 5GHz AX performance is sometimes, but not consistently, better then cAP ac 5GHz AC
2. cAP ax 2GHz AX performance is typically worse than cAP ac 2GHz N
3. cAP ax signal strength is generally better (as seen via Registration signal)
4. Orientation of the cAP makes a difference which doesn't align with the idea that the radiation pattern is roughly spherical
   See viewtopic.php?t=131467#p645555

• RB951G-2HnD, RbcAPGi-5aD2nD (cAP ac), cAPGi-5HaxD2HaxD (cAP ax)
• RouterOS 7.15.2 on all devices
• cAP ac has wifi-qcom-ac, cAP ax has wifi-qcom package
• Client is Dell XPS 13 (9300) with Rivet Killer AX1650s (Intel Corporation Ice Lake-LP PCH CNVi WiFi) which supports WiFi-6 (AX) running Linux Mint 21.3.
• Client is approx 7.5m from cAPs with direct line of sight i.e. no walls etc
• cAPs connected using the same port and ethernet cable, with PoE injector
  Each cAP is connected, tested and then swapped with the other cAP to avoid channel interference
• Test file is 2GB stored on a file server connected via ethernet to the RB951G

• cap-wifi1 cAP ax 5Ghz AX 20/40/80Mhz 5680/ax/eCee
• cap-wifi2 cAP ax 2Ghz AX 20/40Mhz 2412/ax/Ce
• cap-wifi3 cAP ac 2Ghz N 20/40Mhz 2412/n/Ce
• cap-wifi4 cAP ac 5Ghz AC 20/40/80Mhz 5680/ac/eCee

Radio           Channel         Speed       Time        Signal
cAP ax 5Ghz	5680/ax/eCee    67.0MB/s    00:00:28    -57
cAP ax 2Ghz	2412/ax/Ce      2.5MB/s     00:12:50    -54
cAP ac 2Ghz	2412/n/Ce       5.5MB/s     00:05:47    -62
cAP ac 5Ghz	5680/ac/eCee    37.9MB/s    00:00:50    -74

Radio           Channel         Speed       Time        Signal
cAP ax 5Ghz	5680/ax/eCee    32.8MB/s    00:00:58    -62
cAP ax 2Ghz	2412/ax/Ce      5.0MB/s     00:06:21    -51
cAP ac 2Ghz	2412/n/Ce       4.2MB/s     00:07:47    -68
cAP ac 5Ghz	5680/ac/eCee    51.2MB/s    00:00:36    -76

Radio           Channel         Speed       Time        Signal
cAP ax 5Ghz	5680/ax/eCee    62.3MB/s    00:00:30    -60
cAP ax 2Ghz	2412/ax/Ce      2.7MB/s     00:11:55    -55
cAP ac 2Ghz	2412/n/Ce       3.1MB/s     00:10:18    -69
cAP ac 5Ghz	5680/ac/eCee    40.2MB/s    00:00:47    -75

/interface wifi channel
add band=2ghz-n disabled=no frequency=2412-2462 name=cap_ch2n skip-dfs-channels=10min-cac
add band=5ghz-ac disabled=no name=cap_ch5ac skip-dfs-channels=10min-cac
add band=2ghz-ax disabled=no frequency=2412-2462 name=cap_ch2ax skip-dfs-channels=10min-cac
add band=5ghz-ax disabled=no name=cap_ch5ax skip-dfs-channels=10min-cac
/interface wifi security
add authentication-types=wpa2-psk,wpa3-psk disabled=no name=cap_sec1
/interface wifi
add channel.band=5ghz-ax .frequency=5680 .width=20/40/80mhz configuration.country="New Zealand" .mode=ap .ssid=capax5 disabled=no name=cap-wifi1 radio-mac=D4:01:C3:58:88:8B security=cap_sec1
add channel.band=2ghz-n .frequency=2412 .width=20/40mhz configuration.country="New Zealand" .mode=ap .ssid=capax2 disabled=no name=cap-wifi2 radio-mac=D4:01:C3:58:88:8C security=cap_sec1
/interface wifi datapath
add bridge=bridge disabled=no name=cap_dp1
/interface wifi
add channel.band=2ghz-n .frequency=2412 .width=20/40mhz configuration.mode=ap .ssid=capac2 datapath=cap_dp1 name=cap-wifi3 radio-mac=2C:C8:1B:43:6E:D0 security=cap_sec1
add channel.band=5ghz-ac .frequency=5680 .skip-dfs-channels=10min-cac .width=20/40/80mhz configuration.country="New Zealand" .mode=ap .ssid=capac5 datapath=cap_dp1 name=cap-wifi4 radio-mac=2C:C8:1B:43:6E:D1 security=cap_sec1
/interface wifi capsman
set enabled=yes interfaces=bridge package-path="" require-peer-certificate=no upgrade-policy=none

• Client is approx 7.5m from cAPs with direct line of sight i.e. no walls etc

....
Any suggestions for the "better" antennas?

(in any case the WAF will jump from the current -37 to -142.5 if you apply external antennas to a CAP Ax).

Nice pictures, thanks.

Still I will try to re-word the question.

Any suggestion on which specific antennas one can buy that have been tested as being better than the built-in ones? (and that if internal will fit inside a Cap Ax)

The notion that the emission is roughly spherical is (IMHO) "wrong".

It is (should be) omnidirectional as a toroid in the same plane of the device, and likely the performance when front-facing is due to either the "thickness" of the toroid or to reflections (or *whatever").

[*]Client is Dell XPS 13 (9300) with Rivet Killer AX1650s (Intel Corporation Ice Lake-LP PCH CNVi WiFi) which supports WiFi-6 (AX) running Linux Mint 21.3.

I don't get it.

OK, now I understand. OP maxes out at 65MB/s while tothefull reaches 95MB/s at same signal strength. Hmm, iwlwifi Linux is some driver. I don't trust it to reach full speeds for certain hardware. I had a Killer wifi AC card originally in my Dell 9360 XPS. I exchanged it with a AX200 several years ago.