RME Firewire/USB Audio Interfaces and Latency

I saw Jim Roseberry praise the Fireface UC for it's great round trip latency.  Do the Fireface 400 and Fireface 800 also have similar technology that permits very low round trip latency for tracking, etc?

Thanks; my big issue is tracking a guitar or bass to a click rack or drum track.  When I play the finished product back, the guitar or bass always ends up noticeably lagging the click or drums.  As I understand it, this is a function of the round-trip latency.  If I have things right, this is what I'm trying to improve.

I have 8.5 on Windows 7 Pro, with a MOTU Ultralite.

Hi drmathprog

The RME stuff is awesome.

Please note though that the RME  FW unit  [ like all interfaces that use the FW protocol ]   "double-buffers"  - from memory it uses a 64 sample extra buffer - so when you select  48 samples or 64 samples, in reality it is running at  48  or  64  + 64 samples

The last time I tested the RTL of a FW RME FF, the absolute best  RTL  you could get  [ using Centrance ]  was around  ~5.5ms.

By all reports, the   " UC "  verison  is very close to if not the same as the FW version in terms of RTL - for USB, this is astounding !

If lowest possible latency is what you want - go PCI or PCIe with a reputable audio card maker - I suggest Echoaudio or Lynx.

In my case, I run a Layla 3G  @  32 samples in a i7 fully x64 DAW.

Being PCI, it employs no type of " extra buffering" in addition to the normal signal flow.

At 32 samples, using the ASIO 8.2 driver in Sonar 8.5.1 X64  in a Windows 7 X64 DAW, it has a RTL of ~3.67ms and its rock solid at that setting whilst running massively loaded projects  [ ie: CPU loads in the  %80 region ]

It might not sound like much better than FW or USB, but thats  over  %30  less RTL  than the best FW or USB you can find - at such low / extreeme settings, thats a big difference.

In reality, a RTL of  ~3.5ms means I can track a  Guitar sim and the "lag" is so close to  "not there"  that I really do think I am plugged in to a real amp - its that transparent / non-existant.

Once you have a DAW that is stable and reliable at such low latency and you can run at 32 samples at all times regardless of  [ most ]   loads, you will  forget you are using a "computer".

If any sort of mobility with lots of in/outs is an issue - go FW or USB.

If you want to go mobile and large numbers of in/outs are not an issue, grab an ExpressCard interface like and  Indigo DJx, Indigo IOx - no extra buffering and PCI / PCIe speed and efficiency in a mobile setup  [ but make sure your laptop has an Express Card socket ]

Tom

mysteryman


Reported latency is not RTL.
If you want to know the real RTL, use the Centrance tool.

Pete Leoni

 
Not sure who you were replying to, but I'll assume it was to info tom.  If that's the case, then the number he reported was actually the RTL and not the one way latency.  Remember that at high sample rates, the latency number halves.  IOW, latency is twice as fast at 88.2KHz than it is at 44.1KHz.  Add to that the fact that he was using an ultra low buffer size and you'll get those numbers.  Also, as of Sonar 8 (and above), you can view the correct RTL in the Audio Options window when using ASIO drivers.  No need to use CEntrance anymore :-)
 
 
@ Tom G,
 
Just wanted to clarify that the 64 sample 'hidden buffer' that the FireFace units have only applies to the playback side.  The recording side does not have a hidden buffer.  Another thing is that the FireFace only goes down to 48 samples, so it's not fair to compare this to an audio interface that goes down to 32 samples.  If the FireFace could go down that low, it would actually give me ~3.05 ms of RTL at 32 samples, which would actually be a little lower than what you get with the Layla ;-)  I get a RTL of 6.1 ms with buffers if I set it to 64 samples at 44.1KHz.  If my calculations are correct, that means that the Layla's RTL at the same latency and sample rate settings should be ~7.34 ms, is this correct?
 
Either way, both interfaces are capable of giving enough low latency performance that this now becomes nit picking.  I doubt that anyone would tell the difference between the two at such low latencies.
 
 
Take care!