Advantages of Thunderbolt/USB3 audio interfaces

[UltraJv]

Ive had my RME HDSP9632 down to 1ms while also browsing the net. Windows 7 untweaked.

[penguinfromdeep]

Apple buffering is the best, look even in those videos they are using apple

[golemus]

From SoundOnSound:

For many people, USB and Firewire audio interfaces are far more convenient alternatives, since you can quickly swap them between any desktop and laptop machine, but when your goal is really low latency be aware that some USB and Firewire models include small extra buffers to help smooth playback performance, which increases overall latency slightly. This may or may not be declared to your sequencer, which is why in my interface reviews I check the 'real world' latency to see if it's higher than expected.

http://www.soundonsound.com/sos/dec09/a ... 1209_2.htm

[penguinfromdeep]

Coreaudio F*cking rocks!

[hibidy]

Oh no, did we really go there? I mean, really?

[penguinfromdeep]

a lil' truth never hurt anybody ...

[aMUSEd]

i am 100% sure that you CAN NOT get below 32 buffer. IF PCIe cards dont have it
(thats what TB goes thru is PCIe) then no.

at 32 buffer you are pretty much at whats called "real system time"
to get any lower you would need a dedicated OS and non interupts
(another words 100% proprietary system) and then it would not be much less latency. besides who needs less than 64 or 32 buffer? and why?

First of all I think it is better to talk in terms of ms rather than buffer size because that is what really matters, how long it takes from pressing enter to the audio to be hearable. In 44100 sampling rate 32 buffer corresponds to 0.73ms of latency.

I'm actually running everything at 32 samples on my Mac Mini and so far have had no problems. In Logic it quotes me a "roundtrip latency" of 14.5 ms though (at 44100) but I'm not sure if that is the same thing? There is no lower setting though.

[hibidy]

a lil' truth never hurt anybody ...

[hibidy]

i am 100% sure that you CAN NOT get below 32 buffer. IF PCIe cards dont have it
(thats what TB goes thru is PCIe) then no.

at 32 buffer you are pretty much at whats called "real system time"
to get any lower you would need a dedicated OS and non interupts
(another words 100% proprietary system) and then it would not be much less latency. besides who needs less than 64 or 32 buffer? and why?

First of all I think it is better to talk in terms of ms rather than buffer size because that is what really matters, how long it takes from pressing enter to the audio to be hearable. In 44100 sampling rate 32 buffer corresponds to 0.73ms of latency.

I'm actually running everything at 32 samples on my Mac Mini and so far have had no problems. In Logic it quotes me a "roundtrip latency" of 14.5 ms though (at 44100) but I'm not sure if that is the same thing? There is no lower setting though.

That doesn't make any sense. I found appale to be about exactly the same as my windows machine. 44.1 and 64 buffer gives me 6.2ms RT and that is regardless of which machine I'm using (same sound cards)

[camsr]

A good excuse for low latency is guitar cab feedback. For a 44100 sample rate and 32 sample buffer with 0.73ms of input and output delay, this yields 1.46ms of system latency at least. That time in the air is about 1.65 meters or more. I guess if most DACs have buffers in them for interpolation or whatever they do, that would increase the latency more. Wherever that final number is from sound into the interface and sound out of the speaker + distance from speaker, is how far you would be from a real cabinent to achieve this feedback. This makes using software based amps kind of impossible for this thing. If you could get the lowest latency possible, you would still have an impossible feedback zone, no matter how close the guitar is to the cab.

[justin3am]

i am 100% sure that you CAN NOT get below 32 buffer. IF PCIe cards dont have it
(thats what TB goes thru is PCIe) then no.

at 32 buffer you are pretty much at whats called "real system time"
to get any lower you would need a dedicated OS and non interupts
(another words 100% proprietary system) and then it would not be much less latency. besides who needs less than 64 or 32 buffer? and why?

First of all I think it is better to talk in terms of ms rather than buffer size because that is what really matters, how long it takes from pressing enter to the audio to be hearable. In 44100 sampling rate 32 buffer corresponds to 0.73ms of latency.

I'm actually running everything at 32 samples on my Mac Mini and so far have had no problems. In Logic it quotes me a "roundtrip latency" of 14.5 ms though (at 44100) but I'm not sure if that is the same thing? There is no lower setting though.

That doesn't make any sense. I found appale to be about exactly the same as my windows machine. 44.1 and 64 buffer gives me 6.2ms RT and that is regardless of which machine I'm using (same sound cards)

Yeah, that sounds weird. I usually work with 128 sample buffers at 44.1 and my round trip latency isn't that much. With a 64 sample buffer round trip latency is about 6ms with Live and Pro Tools (using a single buffer in Pro Tools) in OSX 10.6.8.

aMUSEd are you using more than 1 buffer in Logic?

[risome]

Coreaudio F*cking rocks!

+1 to that bro

[ermi]

With a 64 sample buffer round trip latency is about 6ms with Live..

Just out of curiosity, is that reported or measured? Which soundcard are you using?

I'm asking because here with 64 sample buffer at 44.1 the reported overall latency in Live is 4.51ms and measured it's 4.3ms or 190 samples (recorded out+in in Reaper with reported latency compensation disabled). Using M-Audio 2496 PCI in Windows. I wonder if Coreaudio adds some more latency, as jcschild said they are handled differently.

(edited for accuracy)

[jcschild]

RTL is another subject but equally as important. whilst not part of this conversation (TB vs) is a valid question.
RTL has to do with numerous things most importantly drivers and built in buffers. AD/DA adds to it as well

huge thread about it as GS
http://www.gearslutz.com/board/music-co ... -base.html

as usual Vins #s equal what we have found.
much about it in there..

RME/Lynx IS the standard to which all other interfaces are held against.
there are no good PCI/PCIe interfaces below that budget.
Echo sort of but they have funky buffering as well. (and midi issues)

RMEs FW/USB are also the standard.. FYI i can run all day long @ 32 buffer with Lynx/RME PCI/PCIe and 64 buffer with FW/USB..

[jcschild]

i am 100% sure that you CAN NOT get below 32 buffer. IF PCIe cards dont have it
(thats what TB goes thru is PCIe) then no.

at 32 buffer you are pretty much at whats called "real system time"
to get any lower you would need a dedicated OS and non interupts
(another words 100% proprietary system) and then it would not be much less latency. besides who needs less than 64 or 32 buffer? and why?

First of all I think it is better to talk in terms of ms rather than buffer size because that is what really matters, how long it takes from pressing enter to the audio to be hearable. In 44100 sampling rate 32 buffer corresponds to 0.73ms of latency.

I'm actually running everything at 32 samples on my Mac Mini and so far have had no problems. In Logic it quotes me a "roundtrip latency" of 14.5 ms though (at 44100) but I'm not sure if that is the same thing? There is no lower setting though.

Logic is notorious for absurd built in buffering.. do a logic test and change buffering and it does nothing...

14.5 ms RTL is crazy high.. 4ms is more like it..