i ran some lab tests between 2 GPS AP's(in TDD, TDD-ptp, and ePTP modes), with the following results:
(for 40mhz channels): i reset the statistics after each reconnection, to make sure i had 100% MCS15 packets.
ePTP:
142/154mb (DOWNlink/UPlink) run from AP(master) side.
Aggregate will be something between those, as the system can not do 100% RX or RX, so each side represents at most some 80% of total-aggregate troughput here.
Ping starts at 1ms, but scales up as troughput grows.
------------------------------------------------------------------------------------------
TDD-ptp 2.5ms:
104/108 (aggregate 213Mbps) (50/50 split)
162/42 (aggregate 205Mbps) (75/25 split)
ping is 7-8ms, and very stable until you saturate the link.
-------------------------------------------------------------------------------------------
TDD-ptp 5ms:
114/116M (230M) 50/50
183/58 (242M) 75/25
ping is the good old 17-18ms until the world ends.
------------------------------------------------------------------------------------------
As you can see, TDD-ptp with 2.5ms frames cuts the latency in half, has comparable troughput, if not somewhat higher than ePTP, has more consistent performance under load,and allows for QOS.
So i would really consider the possibility of running TDD-2,5 over ePTP if you can afford the (moderatelly) increased latency, and that difference gets even lower when you load ePTP links and their latency increases, over TDD-2.5 which doesn't increase latency, at least until you saturate that link, but in that case, you have bigger problems than latency.