Surprised to see those differences in speed, if “local-forwarding” is enabled.
The control(management) of the wlan interfaces is different, but the data-flow processing is identical. But we can’t see the CAPsMAN setting of the WLAN interface.
The local settings remain the same in WinBox, but those are not used when the interface is managed by CAPsMAN.
The only way I know to check WLAN settings with CAPsMAN is via the transmitted beacons.
Checked them and those are identical (analysed with Wireshark). They contain many parameters, but not ALL possible settings
e.g. if WMM and more important A-MPDU aggregation is enabled or not. Those can be set in standalone but I could not find it in CAPsMAN.
Without A-MPDU then performance would be a disaster. cfr: http://forum.mikrotik.com/t/help-mikrotik-capsman-wireless-download-speed-max-200mb-but-pcs-mobiles-link-speed-is-866mbps/142677/1
With “auto” frequency selected , nothing is predictable. CAPsMAN and the next local configured run could use a different channel. This makes results unpredictable.
Only compare when using the same channel.
If “local-forwarding” is not used in CAPsMAN, i do expect different results. The queue’s for CAP can be different than WLAN. The data must be transferred between WLAN and CAP interface. Some comments found on “OpenFlow”, but no real information, if thats always used, still used, or obsolete. At least there is no encryption overhead. Would “priority” be used for WMM or not? And of course how does A-MPDU aggregation work?
EDIT. Probably not OpenFlow but a CAPWAP implementation, as I see UDP port 5246 and 5247 in FW connections, UDP ports as in Huawei forum, and in Cisco documentation.. https://www.rfc-editor.org/rfc/rfc5415.html. CAPWAP data does do fragmentation/defragmentation.