Most low cost interfaces does not have this capability. So you can’t use ASIO direct monitoring with those. For them the check is grayed out in the Cubase or Nuendo ASIO device settings.
So in this case you get the monitoring through the DAW, with added latency, or you could use the interface analog mixer (typically a push button or a mix knob on low cost non dsp interfaces), or use an external analog mixer, or a digital mixer with low latency.
The later two solutions are better because they do allow to add a reverb to the direct monitor input. Some performers or songs cannot tolerate a dry monitoring.
Monitoring through the DAW is ok if your latency is very low and for recording non critical instruments / performers. This is a simple solution because there is no complication regarding effects on the monitored input. You get the effects you put on the DAW mixer and the cue monitor studio mix is exactly the same as the control room mix.
But as soon as the latency rise, for example because you have many tracks and effects and you need to rise the buffer setting a lot, monitoring through the DAW is not a comfortable situation for the performer anymore, or is a no go.
I’ve seen some performers that do not want or cannot play with such a setup. Specially for very fast guitar tracks.
So the pro solution for recording is to have an efficient and automated form of direct monitoring. ASIO direct monitoring with a DSP mixer inside the interface is quite a decent solution for that, keeping the i/o latency in the millisecond range, almost as good as analog direct monitoring, and allowing to add a reverb / EQ / compressor on the monitored input through the interface DSP mixer.
ASIO direct monitoring complete interface support (with DAW mix levels and pan positions copied to the DSP mixer) mark the difference between amateur and professional i/o hardware.
In the old days of analog mixing, armed track input monitoring was perfectly managed by studio consoles without any latency. Specially on larger consoles like SSL or NEVE where you had the option to send the tape output or the group output or mix the two together independently for each record armed track.
Since years 1995-2000, we started to commonly use digital mixing, with added latency. Because of that digital audio recording add a lot of complexity to get a good and easy to use direct monitoring solution. Only manufacturers that do have a good understanding of what audio recording is, do have the right hardware here even for large setups with more than ten or hundred tracks.
In those old years, i remember we had input monitoring latency problems with some performers, even with very high end hardware like SSL AXIOM digital mixing consoles who were using exclusively large multi DSP boards (running at 48 KHz).
For small setups with only a few tracks and effects, DAW monitoring with a small buffer count setting can be ok. But for larger recording setups, ASIO direct monitoring is certainly the answer.
This could change in the future if interfaces could integrate a high computing power DSP at a reasonable price, and all the computer audio computing was transferred to the interface computing. Actually this is the case only for very expensive hardware.
Thunderbolt, with PCIe support giving access to the host DMA system (direct memory access) as well as more powerful computers, could help as well to lower the latency even for high track counts without using interface DSP processing. This could reduce the need to use costly DSP hardware for mid range professional recording setups.
I’m not sure that actual Thunderbolt interfaces are using this PCIe embedded mode. So there is eventually some room for progress here. I would be interested to know if RME do have a driver using this mode.