music,
I think it's more about what the round trip latency time is - for a given sample buffer size. Let's assume that the more tracks/vsts you have, it's up to your PC to supply the grunt - but it seems that not all audio cards are created equal when setting latency to, say, 64 samples.
I've never used a USB audio interface before - I've always had PCI. It seems that USB interfaces generally have higher latencies for, say, 64 sample settings - than a PCI card would have. This is where a better driver will give you lower latencies for the same "sample buffer" settings.
 
Does that make sense?

As I consider what I/O would be an improvement for me (conversion, mostly, latency and features in that order), your comments, John, make me miss my Aardvark Q10 a fabulous budget PCI interface with great preamps 10X10 plus you could stack more of them!, and never did I have an issue, till XP SP3 and Aardvark's dissolution. That was a choice leading to Antelope and I wish they'd come out with something more "budget" friendly.
Oh well.....
I always freeze VST tracks, so I don't seem to notice that kind of build-up.
Michael

If you have a well-configured current generation machine, you can run pretty substantial loads at low round-trip latency.  I leave my MOTU Ultralite AVB set at the smallest buffer sizes.
Rarely do I need to increase it (only on the most dense of projects).

Good information! Thanks Guys

Jim Roseberry
If you have a well-configured current generation machine, you can run pretty substantial loads at low round-trip latency.  I leave my MOTU Ultralite AVB set at the smallest buffer sizes.
Rarely do I need to increase it (only on the most dense of projects).

Jim - you're really helping to steer me to the Ultralite AVB. It seems to be pretty similar to the RME Fireface UC, that I've also been looking at - but with the Fireface UC being 5 yrs old (or so), I'm thinking the newer Ultralite AVB might be a little better "future proofed".
 
I agree with Michael - if it weren't so expensive for a "good" (esp drivers etc) audio i/f - I wouldn't worry so much about "future proofing" ;-)

Previously, I was deciding between the Behringer Air XR18 and the Tascam US 16x08 but since following this tread, I'm leaning towards the Motu Ultralite AVB.
 
I'm thinking the Motu is simply a better buy (compared to the Babyface Pro) as it has more input on the unit alone, can be expanded via ADAT (same with Babyface Pro).. but the AVB feature simply sets the unit above the Babyface.
 
Anyone having the same thought ?

Jim Roseberry
You might (also) want to checkout MOTU's new Ultralite AVB.
Low round-trip latency, solid drivers, and noise-floor of -112dB.

Quick question on "noise floor" as I have seen this comment made a few times (as a differentiator). When I got my interface, this article by Focusrite caught my eye, which basically says that many manufacturers quote the chip itself, and not how it performs in the system (which can be up to 12dB different).
 
Isn't this pretty much splitting hairs anyway, since in a digital realm this is so much lower than analog (i.e. the digital floor is so low it is almost "immaterial" as a differentiator)?

 

Quick question on "noise floor" as I have seen this comment made a few times (as a differentiator). When I got my interface, this article by Focusrite caught my eye, which basically says that many manufacturers quote the chip itself, and not how it performs in the system (which can be up to 12dB different).
 
Isn't this pretty much splitting hairs anyway, since in a digital realm this is so much lower than analog (i.e. the digital floor is so low it is almost "immaterial" as a differentiator)?

Yes, many budget manufacturers just quote the maximum for the AD/DA chip(s).  I have seen as bad as one product claiming 124dB on the AD but it really only reached about 102dB due to an inferior PSU and analog circuitry.  
 
Depending on the application, a 100dB+ noise floor may be overkill, especially in a noisy recording environment.  It really depends on how much headroom you want to leave in the recording.  For live orchestral recordings, I don't want the peaks to reach anywhere near full scale due to stray spikes of 12dB or more beyond what I had planned for.  So a good noise floor on the AD is essential if the input needs to remain low.  
 
The "passive" noise properties of good analog circuitry has the capability of reaching about 130dB, not much more.  So the full 144dB of 24bit audio is impossible with today's technology.  

I also think of noise floor in terms of softer/quieter tracks/parts like classical guitar, whispered vocals, etc.
If I get any noise, an expander will help, but I try to avoid that.
Michael

What I imagine of this card is that the cables get out in all directions. No way to make it tidy. But the sound is the most important part.
And why USB3 or USB3.1, because there is much more voltage and ampere in it. You can get more steady power.