I've been using the RME FireFace 400 via Firewire for several years now and it's great.  Reliable, great drivers, low-latency. Everything I could ask for. Decent enough clean pres, but I also have a couple of more colorful preamps when I need something like that.  (Warm Audio Tone Beast and UAD LA-610 Mk II).
 
While what I have works great and is paid for, I'm considering going for the RME UFX+ and using it via USB3.  On paper it looks like a pretty nice unit and a big step up all around compared to what I have now.  More and better pres, better converters, USB3 for now, Thunderbolt for the next generation DAW I may move to in 2 or 3 years.  Lots of things tempting me. 
 
I'm looking for a bit of real-world experience here on how they compare though. Will the USB3 connection meet or exceed the latency I get with the FireFace 400? I comfortably run at 64 samples in Studio 1 on everything with the FF400, so I'd definitely need that or better, but the overall improvements across the board interests me.  Any other pros/cons I should consider, besides the outlay of cash for this unit?

The biggest difference will be the on board fx. You get an eq and compressor on every input and output. You also get reverb and delay. Recording onto a USB stick. Control with a tethered iPad if you want it. Class compliance so it works with Linux Ubuntu studio.

Auto level for four tracks so you don't have to keep your eyes on gains.

Thanks for the feedback.  Those features are part of what catches my attention.  I think having that built in to the unit might make things nice for monitoring some reverb and delay while recording.  I hadn't heard about the auto level part, so that sounds kind of nice too.  I'll have to do a little reading on how that works.
 
As long as the USB3 round-trip latency is the same or better than my FF400 gets with FireWire, I may have to think this over.  Definitely tempts me, but I don't have to decide right away.

https://www.youtube.com/watch?v=lTHmki4VyfE
 
https://www.youtube.com/watch?v=NVa3AyUNGuI
 
In this one, one of the spectral analyzers is pre-eq and one is post eq
https://www.youtube.com/watch?v=VYNE9ppu9bY
 
The auto-level automatically reduces the gain until you have some pre-configured amount of headroom. If a sound comes down the line that is louder than the configured amount, it increments down the gain so that it would not.
 
Now if something happens on set break, like someone unplugs a guitar, the level might get pushed down more than you want by a pop, but there is a floor that it will not go below so you never have a track that is unusable.
 
 Don't forget you can use the youtube controls to increase the quality so you can see what I'm clicking on as well as increase the playback speed so you can get through them faster.

Thanks for posting those... you're too late. LOL 
I did a little more searching for videos (didn't see those, but they help too).  I saw enough last night that I did a little bit of looking around and found a pretty good deal on one and ordered it.  Looks like I'm going to be able to do some really wild stuff if needed.  I probably won't get nearly as involved, but those videos gave me a few ideas on how to have some fun with it when I get it in.
 
I think what sold me on it was a video where the guy showed a pretty full Reaper project with the buffers set at 32 via USB and it was working well.  I knew the overall preamp and converter quality would be a step up, and the Total Mix stuff would be useful, but it reassured me that I wouldn't lose anything on the latency (and might actually gain even better performance).

I don't push the performance super hard, myself. I like to use soft synths, so my latency is mostly driven by Sonar and plugins, not by the RME. With no plugins, the RME can handle incredibly low latency. Plugin related latency really isn't the RME's fault.
 
I'm glad you found some videos that sold you on it.
 
Some helpful links...
http://www.rme-audio.de/en/
 
RME User Forum
https://www.forum.rme-audio.de/
 
Don't miss out on DigiCheck... super useful.
http://www.rme-audio.de/en/downloads/driver/digicheck.php
 

I use Studio One, and they added in the capability to use the low buffer setting only on tracks that you are monitoring, like soft synths or maybe a vocal track you are recording while monitoring with verb and delay, while playing back the tracks that you have already recorded at a higher setting.  Can't remember what it's called, dropout protection or something like that.  I set that to high and it basically syncs up your playback at 1024 while the live monitored tracks are 64 buffers.  Now that I have that feature I can't imagine getting by without it. :)
 
I'll check out those links as well.  I do have the main company links, but I don't think I ever joined the forum.  Been a long time since i even lurked there, but that's basically because their stuff just works and you don't have to do tons of research to get it to work.
 
Thanks, and can't wait to get this thing fired up now!

Several DAW applications are doing the same type of "Hybrid buffering" (Logic has done it for ages, ProTools, StudioOne, etc)... where are large processing buffer is used to tracks that are just playing back... and a small buffer is used for tracks that are being monitored in realtime.
Makes sense... as it makes much better use of the CPU.

I made a similar switch about 6 months ago. I think you will like it! Although my UFX II just works with USB 2.0
 
I also use studio one 3. Lots and lots of tracks, and the soundcard sounds great. It's slightly but not dramatically superior in the sonics, but I could tell a difference and all works well so far.

Thanks emeralsoul.  Good to hear the practical experience is good.  I'm sure it is going to be killer.
 
And Jim, I had heard that Logic had a similar feature.  My favorite DAW builder only sells PCs, so I've stuck with the PC options over the years. :)  No intention of switching, but I will be in touch in a few years if and when I start to feel the need to upgrade to a Thunderbolt setup.