In order to reduce latency, does it make sense to monitor from the soundcard separately?
 
I'm running Sonar X2 on an old Core Duo laptop. Using a Lexicon Omega for I/O, and I'm getting about a 20ms round-trip latency. Trying to play lead guitar of do anything remotely timing sensitive is problematic. Using ASIO, and have the latency slider down to the next-to-the-last setting (last one produces unstable audio).
 
There is a sound card built into the laptop. Not happy with the cabling options (no way to do headphones and attached monitor speakers without an outboard switch).
 
So the question is, during recording situations, would the apparent latency be decreased if I used the built-in sound card for output instead of routing back out the USB to the Lexicon? 

The built in card won't be fast enough or clean enough. You can monitor your input directly through the Omega (I think), which would give you near zero latency with audio inputs like electric guitar, albeit dry (without any effects). Also, you might be able to reduce total latency by turning off plugins in your project.  Certain "look-ahead" plugs can cause project-wide latency. Good luck!

Although the onboard audio chipset is connected physically to the motherboard, it is connected logically via a driver that is almost always much slower than ASIO.  As Larry says you can get to effectively zero latency with your interface if timing is the problem--the signal you hear never enters the computer but just loops back electronically as an analog signal to your output speakers or headphones.
 
Incidentally twenty milliseconds is much faster than your brain can receive and process the sound and then send motor impulses to your hands. That physiologically limited reaction time is longer than 200 msec. Musicians do not coordinate their playing by mentally processing each note and adjusting the next note on the basis of the last note they heard. That is physiologically impossible. They adjust the speed to an internal brain metronome commonly called finding a groove. Some people can tolerate a pretty long latency in the feedback from what they have just played without getting lost, and that can improve with experience. 

slartabartfast 
Incidentally twenty milliseconds is much faster than your brain can receive and process the sound and then send motor impulses to your hands. That physiologically limited reaction time is longer than 200 msec. Musicians do not coordinate their playing by mentally processing each note and adjusting the next note on the basis of the last note they heard. That is physiologically impossible. They adjust the speed to an internal brain metronome commonly called finding a groove. Some people can tolerate a pretty long latency in the feedback from what they have just played without getting lost, and that can improve with experience. 

I haven't looked at the science in as much depth as that, but I should have mentioned this in my first post: In the real world, 20 milliseconds would translate to standing about 20 feet away from your guitar amplifier. That's about how long the delay (latency) would be. Many stage settings do place the guitarist that far from the amp, so if 20 msec is really your total roundtrip latency, it isn't perfect, but it should be usable.
 
FYI, if your PC has a USB 2 port, you can get a faster interface starting around a hundred bucks.

Thanks to you both for quality feedback.
 
The drive says 20ms, but it sounds longer. 
 
That being said, I picked up a Digittech RP355 on the cheap this weekend, and I'll see if I can split the signal, getting some effects outside of the recording input chain and clean signal going in (wich me luck).

I posted on a thread not too long ago that since disabling the High Definition Audio Controller in System Devices in the Device Manager, my latency with Windows 10 and 2 Studio Capture interfaces has gotten down to 96 samples in ASIO! That is about 2msec!!!!! At the most, I have bumped it up to 128 samples which is around 5msec. And that's running 32 to 64 tracks of realtime midi VSTis and most any FX plugins plus playing guitar thru Amplitube and an open vocal mic ---live!
The key is to disable the internal sound card and use ONLY the high grade audio interface. With those two adjustments you should have a world of difference.

Larry Jones
I haven't looked at the science in as much depth as that, but I should have mentioned this in my first post: In the real world, 20 milliseconds would translate to standing about 20 feet away from your guitar amplifier. That's about how long the delay (latency) would be. Many stage settings do place the guitarist that far from the amp

If I would ever be that far from my amp, I'd be relying on stage monitors to get the sound to me sooner. When playing fast music, the notes are often only 100ms long - an error of +/- 10% or 20% relative to that would be very audible.
 
In this case, I'd definitely be looking into ways to reduce plugin latency (eg. if there's a convolution speaker simulation in use, ensure it's a zero-latency model).

guyshomenet
The drive says 20ms, but it sounds longer

It may well be longer. Sonar gets its latency figure from what the driver is reporting. The trouble is that many drivers report inaccurately in the first place, some wildly inaccurately, and many interfaces have a built-in "safety buffer" that doesn't get reported or slow internal processing that doesn't get reported by the driver either.

Some interface manufacturers, RME for example, explain exactly what the minimum possible latency for their hardware is, most manufacturers don't. So the only way to be sure of what the numbers actually are is to do a loop-back test to find out.

Centrance's free latency test utility is very useful for measuring actual round-trip latency - http://centrance.com/products/ltu/

Thanks. Interesting, but it had no effect on perceived latency on my humble little studio. Since the integrated sound card and the Lexicon are the only sources, I'm still stuck.

Investigating signal splitting now ... seems like the only good option.