Best is 'only n' and only all clients are duo chain too with 'only n'.

Now, if we have some clients in this same network with single chain antenna, what is best option:

All single clients in the furthermore duo chain network same polorization? Either H-pol or V-pol.

Or balance H-pol and V-pol between these clients. So 1st H-pol, 2nd V-pol, 3rd H-pol, 4rd V-pol etc. etc.

Would it make any difference at all? For the radio? For the cpu?

Yet again, best is ALL duo chain antennas. But due budget and/or distance at times it can't be avoided to work with single chain antennas too.
Or, when you are updating your old single chain antenna (that can do 'mimo!) into a duo chain network. You don't want to throw away all your old single chain antenna's yet.....
(And sometimes you need to hook up a client with a 27dBi to get reasonable signal. 27dBi mesh is not so expensive, 27dBi mimo is hard to find and huge in size and very expensive......)

Any input into this matter?

dual chain at AP, single chain at client(s) no problem at all.

Dual chain AP (sector antenna Raytra)
and
Single polarized clients - Vertical
Dual polarized clients - SXT

no problems

I wasn't saying there was a problem.
I am merely wondering if it would make a difference in CPU cycle usage when in a mixed AP network I would put all single chain CPE's in the same polorization or have them distributed over the both chains.

This is actually only interesting to know when your AP network is becoming saturated and you want to have your network to work as economic as possible.

Lets AP with 50 clients. AP radio works with both chains.
30 clients also duo chain, so they use both chains.
20 clients single chain. They can be all Horizonal polorized or Vertical or mixed.

I can imagine the CPU has to work harder if he has to swich between the chains to receive and send data in a mixed network or if all in the same chain, the data is presented to that one chain (for the single units) each time again....

Its sort of the same if you have to work with both 'a' and 'a/n' clients. The fact 'a' clients are around not only deprives the network from the possible high troughput compared to a 'n' only network. It actually even slower than a full 'a' network. Mind you, when we have 40 or 50+ clients that all can have 3 to 6 Mb traffic demand.

I've had severe problems on a network where an omnitik had to server 49 clienst, amongst them still 6 r 7 'a' clients. It gave problems for the whole network. Only by replacing the 'a' clients for 'n' clients suddenly the performance of the whole network improved.
Same routerboard (rb433AH also server to backhaul its traffic to another remote unit) could not even process high data throughputs on the backhaul link when the AP radio had a mixed network.

(When the AP radio was switched off, suddenly the backhaul quadrupled its throughput. Later, when all 'a' units were replaced by 'n' duo chain radios the backhaul also started to perform 4 times better. So my conclusion was that the fact the radio has to switch between communication in 'a' and 'n' mode, also on one chain only, and on duo chains, cost so much cpu power the board simply couldn't handle it.)

So this merely theoratical discussion is; Is there any extra labor needed for the cpu if its radio has to communicate only on both chains all times, or for some CPE's has to talk and listen to them in 'single chain' mode? And would it make it even worse when these single chain units are distributed over both chains.. or would it be better to at least have them all in the same chain?

If there would be a difference it would specially become more interesting if you are planning to setup a triple, or even quadruple chain AP network where you'll have single and duo chain clients connection to one or two of the 4 available chains....
Its al about stretching the radio to its limits.... with only 10, 20 or even 30 clients a rb433AH or higher is powerfull enough to handle any mix. (I have one rb800 with 55 duo chain clients and still 4 single chain clients. But this network runs fine.... probably can handle a couple of more units...)

The MIMO-Stuff is done by the Wlan card so it should not impact main cpu.

Using one AP V and one AP H you might increase frequency capacity at the tower.
The H/V Separation helps with tighter channel spacing.

The MIMO-Stuff is done by the Wlan card so it should not impact main cpu.

Is that really true? If so, how to explain that my backhaul radio of the same rb433AH could handle more througput when the AP radio is switched off? (Bandwidth test done from router outside the actuall radio link members). If the radiocard would handle the mimo stuff than switching one card off would have no impact on the other. (When the traffic the bh card has to handle is still the same = bw test fm other board)
Just by making my full client network all 'n' instead of the mixed I had before also improved the backhaul radio..... So imho the cpu of the board must be involved in the work the actual radio is performing. If not I might as well use slower boards.... al they have to do is passing the traffic ....

Using one AP V and one AP H you might increase frequency capacity at the tower.

Yes true. But not very realistic in everday setup. Existing ('a') networks are little by little upgraded to 'n' networks. That goes slow due restrictions in resources. So I start with having the new incoming clients getting a 'n' radio. Than, as soon as the money is available, I put a 'a/n' able AP as replacement for the old 'a' AP. (That is left for backup, but switched off). Now little by little boards from existing 'a' clients are wherever possible replaced by 'n' type boards in the same (RIC) antenna's to keep the cost of the replacement under control.
So here is the 'mixed' network now. For months the new 'n' AP has to deal with a whole bunch, but slowly degreasing, amount of 'a' units working all in one polorization, and a slowly incraesing amount of 'n' type of CPE's that still work in one chain. On some of my new networks I am also faced with the fact that although the new AP is H+V polorized, not all CPE's can be made dual chain. (At 15km distance I need a BIG antenna on the client's premises to get a sufficiant enough signal on a dual chain CPE. That is too costly for me, thus the client. So I can only serve him when I put an 'n' single chain (Groove) on a 27 or 30dBi antenna.
So now I have the choice; should I make no.1 client V-pol and than the 2nd H-pol and the 3rd V-pol again etc. or should I put them all in the V-pol (or H-pol) chain? (There are also some duo pol client attached to same AP using both chains)
Basically the issue here is; is there a need to balance out the usage difference over the chains of the AP or not?

The H/V Separation helps with tighter channel spacing.

True. But not always the most practical thing to do.

The MIMO-Stuff is done by the Wlan card so it should not impact main cpu.

Is that really true? If so, how to explain that my backhaul radio of the same rb433AH could handle more througput when the AP radio is switched off? (Bandwidth test done from router outside the actuall radio link members). If the radiocard would handle the mimo stuff than switching one card off would have no impact on the other. (When the traffic the bh card has to handle is still the same = bw test fm other board)
Just by making my full client network all 'n' instead of the mixed I had before also improved the backhaul radio..... So imho the cpu of the board must be involved in the work the actual radio is performing. If not I might as well use slower boards.... al they have to do is passing the traffic ....

Of course enabling a second card on the same board influences the first cards behavior.
We've done this in the past but never do this again. None of the systems using more than one card runs good in our network.
This is a matter of interference and I guess RB600 is overloaded with handling nv2/nstreme on 2 pci slots.
We use one board (411ah) per Antenna/Sector and get much better results now.

Using one AP V and one AP H you might increase frequency capacity at the tower.

Yes true. But not very realistic in everday setup. Existing ('a') networks are little by little upgraded to 'n' networks. That goes slow due restrictions in resources. So I start with having the new incoming clients getting a 'n' radio. Than, as soon as the money is available, I put a 'a/n' able AP as replacement for the old 'a' AP. (That is left for backup, but switched off). Now little by little boards from existing 'a' clients are wherever possible replaced by 'n' type boards in the same (RIC) antenna's to keep the cost of the replacement under control.
So here is the 'mixed' network now. For months the new 'n' AP has to deal with a whole bunch, but slowly degreasing, amount of 'a' units working all in one polorization, and a slowly incraesing amount of 'n' type of CPE's that still work in one chain. On some of my new networks I am also faced with the fact that although the new AP is H+V polorized, not all CPE's can be made dual chain. (At 15km distance I need a BIG antenna on the client's premises to get a sufficiant enough signal on a dual chain CPE. That is too costly for me, thus the client. So I can only serve him when I put an 'n' single chain (Groove) on a 27 or 30dBi antenna.
So now I have the choice; should I make no.1 client V-pol and than the 2nd H-pol and the 3rd V-pol again etc. or should I put them all in the V-pol (or H-pol) chain? (There are also some duo pol client attached to same AP using both chains)
Basically the issue here is; is there a need to balance out the usage difference over the chains of the AP or not?

The H/V Separation helps with tighter channel spacing.

True. But not always the most practical thing to do.

When there is a single pol V network you can add Sectors with dual pol Antennas, enable only H, add new CPEs in H (first chain of SXT) and leave the older CPEs on V. You can switch to dual pol when there is enough spectrum.