You need to provide more data about your setup so that it's possible to correlate the data in relation to the subject matter.

1. computer (CPU, RAM, HD/SSD,Mainboard/Chipset).
2. besides ASIO buffersize also the sample rate
3. besides ASIO buffersize also the RTL of the driver
4. number of audio channels, which are transmitted through computer interface (e.g. USB, PCIe)

The ASIO buffersizes at different sample rates
- single speed, 44.1/48 kHz, 256
- double speed, 88.2/96 kHz, 512
- quad speed, 176.4/192 kHz, 1024
are de facto the same, if you double or quadruple the data throughput, then inevitably the buffers must also become larger.

But with increasing sample rate, the load on the overall system also increases, because each component (hardware/software) has to process more data in total. Also from each VST/VSTi in the DAW.

On the other hand, the latency between computer and recording interface (the RTL, round trip latency over USB/FW/TB/PCI/PCIe) and also the converter latency decreases with increasing sample rate.

So if you compare the Motu interface with the AIO, you really need to ensure that the setup is equal / compareable
- same project, same load
- same sample rate / ASIO buffersizes

And then you still have to consider whether both interfaces have the same number of channels to transmit and if the chosen ASIO buffersizes for the different products (in this case RME and MOTU) have the same RTL.

I have often seen that the RTL of RME interfaces with the same ASIO buffersize settings was up to half lower compared to other vendors products.
This is good for the user, but means for the computer more load, higher near-realtime requirements, but then also to fetch the data in time, for this the computer has less time.

In such a case it can be that badly written drivers on your computer can make themselves more disturbingly noticeable and occupy CPU cores too long, so that the audio processes cannot be processed in time. In this case you have to measure with LatencyMon, which drivers these are and see if you can minimize these latencies by driver up/downgrade or even change a hardware component, so that the CPU cores are available faster for audio processes. So minimize DPC latencies.

The goal of optimizing the computer (BIOS/OS Settings, HW/driver) is to minimize internal latencies (those DPCs) because: the more efficient your computer works (good drivers = low DPC latencies) the smaller you can choose ASIO buffersizes or in other words, the more threshold you have until if comes to a drop of audio.

Or to put it differently, if your Motu has less channels or the driver was more comfortable, then the load on your computer was not that high. With RME you may need higher ASIO buffersizes, but it may be, as I said, that the RTL is significantly lower with RME drivers compared to other interfaces.

That's the reason why you can't really contribute to the matter at this point with the little info you provide, necessary details are missing and still then you have to correlate and rate all this information.

First you would have to find out if the two setups are so easily comparable:  Number of channels, how high is the RTL with the same ASIO buffersize, ...

And then it's still hard to say how your system (about which there is no usable info so far) performs and whether it would perhaps also get along well with a USB-based solution.

It may also be that you would need to use a dedicated USB card to decouple the recording interface from the rest of the audio infrastructure.

Maybe there are general latency / DPC problems on your system that you need to get a handle on first (-> LatencyMon). Maybe then the whole discussion .... is unnecessary, it could be that a driver here CPUs take too long.
This is not that uncommon and can change any time after driver updates.
If it runs well once, it doesn't mean that it will continue like this forever, if you get other versions installed again and again due to automatic driver updates.
It is fundamentally different whether the drivers all come from Microsoft or you install a certain version of the manufacturer yourself for some drivers, whereby you have control over whether and when you upgrade.

> Minimal track count going in and out.
Again, all channels of a recording interface are always being transferred always regardless how many you use in the application.

Still no information about RTL between the two interfaces, maybe the MOTU uses more safety buffers or is not written that efficient, so that the PC has more time to process audio.

> could a UCX change the situation or should I sell and go back to Motu?
Nobody will be able to tell you this upfront, you will need to try that out things like this are much system dependend, board design, BIOS, chipset and what not.