Specs state a 2 W "base" difference:
RB5009UG+S+IN Max power consumption without attachments 14 W
RB5009UPr+S+IN Max power consumption without attachments 16 W

It doesn't sound as a lot, even if it can be read as "an increase of almost 15%" it remains 2 W.

How would the power adapter behave?
The first comes with a 24V1.5A power supply suitable for the stated max consumption of 25 W, the second comes with a 48v2A (PoE) power supply that can evidently bear the (unstated) 25/27? W max and has some 70 W available for other devices (but some excess heat production, if any, is moved to where the mains outlet is and in that the PoE power supply is more flexible in placement).

The math is strange, in the second, if max power is 150W and 130 of them can be used for PoE out only 20 W remain (in theory) for the device itself, which is coherent with the 20W of "reserved power" stated for cases of not enough power for the PoE devices and their shutdown based on priority.

Yes, the 2W won't make a difference in practice, I am more perplexed by the power supply.

The non-Poe one, 24V 1.5A makes 36 Watts, so the 14-25W represent respectively 39% and 70% of the max power the PSU delivers, which should be adequate.

The Poe one, 48V 2A makes 96 Watts, so the 16-27W represent respectively 17% and 28% which are pretty much low values.

It is clear that Mikrotik has chosen a compromise in the second (enough to power a few PoE devices, but not capable to deliver the whole 150W the device can bear) probably because a beefier PSU would be much less efficient with even lower pencentages of load.

On the other hand, generally speaking, it should be preferrable to use higher voltage (which means automatically less current) as dispersion and generated heat should be less.

Probably, but it will have to be seen, a 48V 0.8 or 1 A would be a better choice in the long run if you don't use the PoE option.

Yes, though personally I prefer the "other" style for PSU's with the "brick" connected to mains through a cable as it gives more freedom to place it where you want instead of having it sticking out of the mains outlet.
Mikrotik used to have a POW48 with 48V 0.8A that would be IMHO ideal (I think it still can be found online in some shops), the new one is 48POW 1.5A which is already too much.
But there is no need for it to be necessarily Mikrotik, it's a simple (passive) power supply.

When one uses power adapter only to power router (i.e. no PoE-out), then it's better to use PA which outputs lower voltage (but still equal or higher than lowest acceptable input voltage). The reason: router's electronic parts require pretty low voltages (probably anything between 1.8V and 5V), so there is one (or multipke, even multi-stage) DC-DC downconverters. Efficiency of typical DC-DC downconverters is inversely proportional to difference between input and output voltage, so down-converter to e.g. 3.3V will operate more efficiently if input voltage is 18V than if input voltage is 48V.

The (in)efficiency is overwhelmed by cable loss when using PoE with lower voltages (over long cable sections), so when using same PA for PoE-out, it's better to use higher voltages (if it' selectable at all, 802.3 af/at requires 48+V power adapter).

Yes, in theory you are right, but it depends on a number of factors, load characteristics, amperes needed, the way the DC-DC converter has been designed, the frequency it operates and what not, typically, in practice, in a properly designed converter the difference in efficiency may be in a few % over the proper range, like (say) from 94% peak to 88% or so.
The fact that the good Mikrotik engineers gave prevalence/priority to the higher voltage may (or may not) be connected.
Anyway using the 24V 1.5A power supply that comes with the "normal" RB5009 is a very good idea.

Just tested...
RB5009 PoE version
Idle, nothing connected (except power .
24V > 4,5W
48V > 6,9W

Very good, so we have a real life data point, thanks .

So, it is (approximately) the 2W difference stated in specs between the non-Poe and the PoE one?
Or is it a coincidence?
I mean, also the non-PoE version can be powered with 24-57V, so the 2W difference could be due to the non-PoE version being tested @24V?

The way I read post by @andkar ... it's the same device (PoE version) powered either by 24V or 48V. So the difference is efficiency in downconverters ... and possibly some 802.3af/at specific PoE circuitry which may get switched off when device is powered with 24V (and can thus not provide 802.3af/at PoE). This possibility would be ruled out (or confirmed) if same test was done using the non-PoE version.

And I doubt that MT engineers opted for higher supply voltage considering internal reasons, I strongly believe it's due to PoE out (either passive or even 82.3af/at which requires 48V or higher).

But you're right about range in efficiency of downconverters, most really do operate between around 85% and 97%. Now whether the difference is insignificant or not can be pretty subjective, but if there aren't any other constraints regarding supply voltage, I'd say this consideration can (and should) be relevant as to PA selection.

@VMX
Until this is confirmed by further tests, either:
1) the PoE and not PoE versions are essentially the same BUT the power declared in the specs is for the first relative to 48V (or so) and for the second relative to 24V (and it is the 2W difference)
or:
2) the PoE and not PoE versions are different internally and the PoE version does consume 2W more at the same 24V

If a) the total difference is 2W
If b) the total difference could be 4W

In both cases, I don't think it would make much of a difference from the point of view of thermal management, 4W is a lot in percentage, but it remains 4W in absolute terms.

Out of curiosity, why the CSS610-8P and not, say, the RB260GSP? Were you planning to power half of the devices with the RB5009's and the other half with a smaller switch, and now you need a bigger switch that can feed 8 PoE devices?

The RB260GSP has passive only (low voltage) PoE out, not 802.3af/at capabilities.

Only to keep things as together as possible, gotspring's misadventure with a RB5009 is posted here:
viewtopic.php?p=1061574
though it is not (yet) clear whether the power supply or the RB5009 unit he tested can be defective.

Even if Mikrotik's way to deal with PoE specifications appears more often then not "vague", I don't think that when they declare 150W for the router and 96W for the power supply a setup with only 40W+25W=65W can fail, if all the devices involved function correctly, it would be - more than deceiving advertising - a fraud.

It seems that the declared 802.3at support by Mikrotik is not fully-fully compliant with the standard, but at least in theory the 13W or so access points are well within the 15.4W of the 802.3af specs, so it cannot be an issue with negotiation between 802.3af and 802.3at power requirements.

Since the RB5009 model that supports PoE is a relatively new device there are not many reports on the field, let's wait and see if there are new similar reports of the standard setup being so much underpowered.

Mikrotik looked at the support output and said nothing about defect. They specifically stated BECAUSE OF THE POWER SUPPLY I WAS USING, I should get 20watts per port.

Well, if they told you that they are contradicting themselves or they were not clear, the 20 W is an upper limit for each port, no matter how many amps your power supply provides, you could have had a much beefier power supply and still not be able to get more than 20W from a port.

The specifications for that router say that EACH port can deliver:
Max out per port output (input 30-57 V) 420 mA

At 48V, that means that each port can deliver 20.16 W max (which is not really-really 802.3at compliant, as that should be 25.5 W, and even at 57V 0.42A mean 23.94W ), anyway these around 20W at 48V available are more like a "beefed up" 802.3af than a real 802.3at.

The total amount of deliverable power being 130 W it is clearly not a limit when using a 96W psu.

The power supply you have (again by Mikrotik specifications) can provide 2A at 48V, 96W, so you have to take out the amount the router itself might use, that still according to Mikrotik, is max 16W, what remains, 80 W is available as PoE power on all the ports.

Mathematically, that can be around 11W on 7 ports, around 13W on 6 ports or around 15W on 5 ports, or 20W on 4 ports.

Even if the declared 16W for the router is optimistic, and it is more like 20 or 25 W, you still should have no less than 70W available.

But you reportedly used only 3 ports, attaching to each of them 13W devices, so in total 3x13W=39W when the deliverable power on those three ports should have been 3x20W=60W and the power supply you used can deliver a residual 70 or 80W.

So, according to specs, the discussion could be if you could power 5 or 6 of those cambium AP's, not only 3.

The explanation of that configuration not working can be any of:
1) malfunctioning power supply (giving out much less than 96W)
2) malfunctioning RB5009 (giving out on each of the three used ports less than 65% of the power it can deliver by specs)
3) the cambium devices needing much more than 13W, actually more than 20W
4) some "queer" compatibility issue with the PoE negotiation between the devices (the 802.3af allows 12.95W, maybe what the cambium devices ask for is 802.3at and maybe the MIkrotik does not properly understand the request)
5) something else in cabling, connectors? (sounds very improbable)

If you exclude all of the above, what remains is a huge difference between the declared specs and the actual field capabilities of the RB5009.

And the difference is so huge that it cannot be due to a "rounding error" or a "tested in different conditions", if confirmed, it represent what in non-PC terms would be called "a blatant lie" in the specifications.