This one comes from something I've read, which was written 2 or 3 years ago I think, and may or may not be relevant anymore. The author himself admits that this was more of an issue in the days of USB 1.

In short, he's saying that a Firewire device supposedly operate on a peer-to-peer architecture whereas USB operate on a master-slave system, meaning that there is an additional load on your CPU when using a USB device.

This part, I could and did verify using Google no problem.

But in terms of how this translates into actual difference in performance, is the theoretical advantage really worth considering or are we talking about something of the magnitude of being able to stream 4965 tracks vs 4963 tracks? Which would be totally irrelevant...

When I was buying my interface 6 years ago... (has it been that long?)  there were USB and Firewire interfaces.  USB 2 was new. USB 1 was slow. Firewire was faster than both. 

At the time since Firewire was faster of the two types available,  I went with it.  That is no longer the case as the new USB ports are now faster than FW and they don't have chipset issues. 

Firewire has additional advantages that most DAW users don't need to consider that are valuable if you are streaming multiple streams of video (with synchronized audio) into switchers and trying to switch without black glitches.

USB devices may be able to stream chunks of data on parity with firewire but it is rare that you see a USB I/O implemented that has the low latency and high count of synchronized tracks that firewire devices are provided with.

The few examples of highly efficient usb audio I/O drivers achieve the status by tweaking the drivers. Take a look at the RME fireface ufx.

If you are under 16 channels and you don't need the fastest latency you'll see less differences.

for the most part firewire and USB2 should be equally fast.  but USB does put a higher strain on the CPU than firewire does and, consequently, will likely have a higher latency than a similarly built firewire device with comparable drivers.

On my MOTU Ultralite Hybrid, I have used both USB and Firewire (the hybrid allows either USB2 or Firewire to be used).  using the ultralite on firewire always has less popping/clicking/stuttering problems than USB at the same latency settings.

Thanks guys. 

Beagle - a couple of audio interfaces I've considered have both options, which I thought would be a plus. Thanks for confirming. That Ultralite Hybrid does look nice.

I actually love the ultralite!  it only has 2 mic pres, but that's fine with me because it has enough to get me "by" if I need them, but most of my mic pres are external in the studio.  it has 6 line inputs for me to use with my external mic pres.

the USB / firewire flexibility gives me the ability to use it with USB on a mobile system I have which does not have a TI chipset firewire (which is almost imperative for smooth firewire operation).

mike_mccue


Firewire has additional advantages that most DAW users don't need to consider that are valuable if you are streaming multiple streams of video (with synchronized audio) into switchers and trying to switch without black glitches.

USB devices may be able to stream chunks of data on parity with firewire but it is rare that you see a USB I/O implemented that has the low latency and high count of synchronized tracks that firewire devices are provided with.

The few examples of highly efficient usb audio I/O drivers achieve the status by tweaking the drivers. Take a look at the RME fireface ufx.

If you are under 16 channels and you don't need the fastest latency you'll see less differences.

This ^^^
 
Again it's more of a concern for a solution developer to confront these differences, if a USB device performs within your requirments for it then it is of no concern to the user whether it is implemented one way or the other.
 
If low latency was that important a consderation I'd say forget either and use something that works off the PCIe bus directly and preferably something that has drivers that take advantage of that configuration fully.  Probably again something RME.

I'd like to have an affordable 24 in/12 out solution; all in one box with XLR line level connectors on the back and full size meters, per channel, on the front and I'd like it to run real smooth at 32 samples.

I haven't got a choice for that with either Firewire or USB... so I've settled for 16x16 on firewire that does it at 64 samples all day long and fits that form factor.

I used to use that set up on a store bought laptop as well as a custom config DAW but it's gotten kinda difficult to find fire wire on a laptop and that laptop died so...


I miss that laptop... it is the reason I didn't go PCI when I got the firewire rig.

I can't say that my current USB audio interface failed to meet my expectations. Of course, I only record one track at a time. 

That being said, I rarely need to raise the buffers higher than 32, I very rarely get clicks and pops - unless I'm trying that in Pro Tools. ;)

That being said, I'll have to collaborate on a few larger projects where squeezing a bit more performance out of my rig may be more than welcome, so I'm looking at every little things which can make a difference.

The buffer size you use isn't the deciding factor although it's part of it when going for low latency.
 
The measured round trip is the crucial figure.  You maybe able to run at 32 samples with no problem but the round trip figure will sometimes be higher than someone elses card running 64 samples.  My old E-Mu for example would only go down to 96 samples but I could get similar latency figures as I do with my current USB device running at 48 samples buffer.
 
The only way to get the true picture is to measure the point the signal goes into your interface to the time it comes out the other end of your chain.
 
Not every one needs that low latency which can sometimes be at the expense of audio stability and increased CPU usage, for example if you generally use external hardware monitoring it won't bother you at all.
 
A low enough figure so you can track a guitar say through whatever amp sim you use to get your tone or play a keyboard through your favourite synth without the latency interfering with your performance because of the time lag is usually low enough for most people, although many people treat it like an Olympic sport and want to have the best figures for 'show and tell'.
 
Again that figure will depend on your sensitivity to the time delay.  For me anything between 6-9 ms is OK anything below that is excellent and anything above that and it starts to throw me.