I have an old M-audio delta 1010 and run it at 1.5ms latency (24/44k) ... I maybe doing a mobile rig with laptop and wonder if there is any USB sound cards that run that fast. I dont need 10 in and outs...more like 4 at best - cause Im just playing live thru it.
however, I can tell a difference in the 'feel' when you get above 5ms...givin there is already embedded latency.  I have heard for years that USB is just as good...or have heard it has never had the latency even close to PCI ...even the old stuff is still faster. I just dont know, i just dont have expereince in that. Howver, I would like to see as I am also looking at android and stuff slowly coming out for midi support/etc. (its just not at the level pc is...and I like having nice sounds in a compact laptop) and I like recording while we play.  FYI - i did do a search on the forums for latency and didnt pull up much

RME Fireface UCX on USB has the lowest latency I know of (USB or PCI), around 3ms total round trip at 48 samples/44.1KHz. I don't know how you're getting 1.5ms but it doesn't sound like a realistic figure. I know the 1010 is a good low latency performer but even so. Is this round trip or one way? Reported by the driver or measured? And at what settings?

Hi Ben,
 
The Delta 1010 yields 5ms total round-trip latency at a 64-sample ASIO buffer size 44.1k
The best USB audio interfaces are right on par.

• The RME Fireface UFX yields 4.3ms total round-trip latency at a 48-sample ASIO buffer size 44.1k
• The RME Fireface UFX yields 5.1ms total round-trip latency at a 64-samples ASIO buffer size 44.1k

Many USB audio interfaces use a large hidden safety-buffer.
If that's the case, it'll have high round-trip latency.
 
If you stick with RME or MOTU, you'll have low round-trip latency (at or below that 5ms mark).

FWIW, The RME Fireface UCX uses the same innards as the UFX.
The UCX yields 4.3ms total round-trip latency at a 48-sample ASIO buffer size 44.1k
That's excellent performance...
 

i believe it was 64... thats what it says on the Sonar profiler. Now that may have been at 96k for 1.5ms...round trip I just dont know...but did see a jump in cpu however,i do have a fx-9590 running at full throttle so it can handle it... ok, i may go and see if I can look at RME or motu. I hope there is better development for usb 3 and faster-smoother-high resolution,etc. I would like to go as cheap with tascam or alesis or whatever but I know sometimes you pay for what you get...I just need LOW LATENCY round trip...

If lowest possible round-trip latency is paramount, stick with RME or MOTU.
 
Round-trip latency is the sum of the following:

• ASIO input buffer
• ASIO output buffer
• The driver's safety-buffer (often hidden - this is the X-factor and not all units are equal)
• A/D D/A converter latency

At 44.1k, 64-samples equates to 1.5ms
Thus, at a 64-sample ASIO buffer size 44.1k... you've got 3ms for just the ASIO input and output buffers.
Add to that the safety-buffer and the A/D D/A latency... and you have the total round-trip latency.
 
Increasing the sample-rate does lower the round-trip latency... at the expensive of higher CPU use.
Doubling the sample-rate will cut the ASIO input and output buffers in half (latency wise).
You still have the hidden safety-buffer and the latency from the A/D D/A converters, so the figure is typically higher than half.
 

Jim, I don't mean to question your thorough understanding of these things but IIRC the UC and the UFX have the same converters and the UCX has newer AD/DA leading to slightly better figures.

EDIT:
Actually it's more complicated than that but likely the different AD in the UCX doesn't contribute to lower latency. To be fair, I didn't actually measure this number myself, I got it from an article with a long list of tested interfaces that I can't find now, so no use debating the 1ms difference :)

No worries!
You're right... the difference is minimal at best.  

Latency is the enemy.... -S
 

Couldn't we take a load off our processors using cell phone powered vst's over USB? more ineractive also...copyright me.