School and Cactus, that's helpful information; thanks a lot! I'll try to have a go at that Centrance RTL test in the next couple of days. I'm mainly a synth and piano guy, so everything is smooth sailing for now. I do plan on tracking guitar with amp simulation in the near future, however, so I do want to make sure my latency is as low as possible. The Focusrite 2i2 does have a zero latency monitoring feature, so I'll have to explore that a little further when/if it becomes necessary. Thanks for the help!

I can say with 100% certainty, that ASIO "measured" latency is not totally accurate. I've never seen one latency measurement by ASIO drivers come within 30 samples of 100% accuracy. The offset will help here, and I strongly suggest using that to improve transient material. Even the best clocks in the world won't fix this, and you could pay thousands of dollars for a clock solution that is totally undone by having an ASIO driver that can't correctly measure latency.
 
I believe this is because the ASIO driver only measures to a limit, without taking any specific interfaces in mind, such as FireWire or USB, or the external hardware unit's D/A-A/D latency. I suggest transmitting a metronome out one channel and recording it with another onto a new track. Repeat this again, so you can compare the transmitted metronome with the next recorded track of that metronome after going through your interface. Count the sampled difference (in samples) between the two tracks and adjust in the Offset of the ASIO driver. You can get the accuracy to exactly one sample, regardless of the latency.

A few thoughts/corrections:
 
- Cactus' link to Centrance is the way to go to measure your true RTL. Subtract what SONAR reports from what CEntrance measures, and that's your Manual Offset. But it only corrects recording (i.e. Input) latency after the fact and has no bearing on real-time latency.
- Buffer size affects outbound latency, and outbound latency will affect soft synths.
- Latency also includes A/D/A conversion and input bus latency, and this is generally the largest part of what the driver doesn't report. This typically adds about .5ms each way, and is why manual offsets are often on the order of 40-50 samples, though higher is common (especially for USB2 interfaces), and lower is possible if the driver programmer accounts for it.
- Soft synths are also subject to MIDI "latency", including the keyboard scan time, MIDI transmission time, and SONAR's time to echo live MIDI input to the synth. This can be quite variable, depending on your MIDI interface/driver performance. Typically this is adding another 4-6ms.
- Ignoring MIDI latency for the moment, most users agree that a true ASIO RTL under 10ms feels pretty decent for playing soft synths or monitoring analog inputs in headphones. It's quite possible to get used to more latency, but I think most would start to be bothered by anything over about 15ms. 37ms would be unbearable for most.
- Most decent interfaces can perform well with an ASIO buffer of 128 samples, giving a true RTL of 6-8ms at 44.1-48kHz, depending on the the driver and what bus is used to connect to the DAW. And so long as the Deferred Procedure Call (DPC) latency of the PC is reasonable (preferably consistently under 200 microseconds), performance should be smooth and click-free for the tracking phase of a typical project.

Nicely articulated.
 
I have been through all of this stuff in detail over the years, and have found good balance with the settings I am currently using, which are noted in my signature.
 
I don't want to push the envelope - I just want the letters I write to make it to the mail box without a lot of grief, as it were.
 
Bob Bone