I have seen and mentioned this before. When running TDD as you increase channel size the individual SMs do not see a linear bandwidth increase. Meaning at 40mhz you can get close to 88% of the estimated bandwidth (ePMP Capacity Planning Tool) on DL and 91% on UL. Moving to 80mhz and I usually see at best 73% DL but still greater than 92% UL.
Now from the AP side if I run link tests to multiple 4625SMs I can get the AP DL to put out numbers much closer to the ones Cambium lists on the Planning tool. My guess is that the TDD overhead on the SMs limit individual performance at higher channel sizes, but the AP has the processing power to push more as seen when doing Link Tests to multiple SMs at the same time.
ePMP 4k - 80mhz performance bottleneck - Products / ePMP - Cambium Community
Is this a limitation of the processor on the 4625/4525 SM being not able to really take full advantage of the larger channels above 40mhz? Meaning that the larger channel sizes will mainly show their benefit in more capacity for the AP and not necessarily in higher speeds for the SM?
It seems that might be the case as I did a Dual SM 20s link test using two CPEs with similar stats (MCS11 DL/MCS9UL) and I got a combined total of 777DL and 139UL. Using the planning tool with 2 SM it showed I should get a total AP capacity of 765DL and 151UL. That means when doing dual SMs I was getting 101% of max DL and 92% of max UL.
4625SM better at MCS9 vs MCS11? - Products / ePMP - Cambium Community
This is also interesting. This performance ‘increase’ only seems to affect single SM link tests. If I do a dual SM link the AP gets more performance at MCS11 than it does at MCS9 but single SM link tests are better at MCS9 vs MCS11.