1 (edited by Timur 2010-01-04 10:35:00)

Topic: Latency comparison: HDSPe vs FF400 vs FFUC

Because this questions keeps coming up here is a comparison of the HDSPe (ExpressCard + MF2) vs Fireface 400 vs Fireface UC on both Windows and OS X.

The following list shows the number of samples used by hardware (AD/DA) + safety-buffer, but *without* audio buffers (which obviously vary according to what you setup yourself).

Input OS X vs. Windows (in samples)

HDSPe: 69 vs. 32
FF 400: 109 vs. 45
FF UC: 69 vs. 55

Output OS X vs. Windows (in samples)

HDSPe: 103 vs. 64
FF 400: 96 vs. 96
FF UC: 70 vs. 63

These are the latencies your interface begins with *without* audio buffers being added. So on Windows a Fireface UC at 44.1 kHz with a buffer setting of 64 samples would sum up to 63 + 64 = 134 samples = 2.88 ms output latency.

Windows buffer sizes are lower to begin with but OS X allows to set audio buffers as low as 1 sample (if the application supports it), while the minimum on Windows is 32 for PCI(e) and 48 for FW and USB (in practice rather 64).

Re: Latency comparison: HDSPe vs FF400 vs FFUC

That is very interesting...it would seem that the FF UC gives the best latencies...however, what about CPU load at these latencies?

I would be interested to know if the FF UC didn't load the CPU anymore than a HDSPe/Multiface setup?
(...especially with using softsynths, Kontakt et al, in Logic Pro.)

Thanks.

3

Re: Latency comparison: HDSPe vs FF400 vs FFUC

Both FW and USB rely on Microsoft's basic drivers for these services, while PCI/PCIe works 'directly'. That's why they will always have the edge.

Regards
Matthias Carstens
RME

Re: Latency comparison: HDSPe vs FF400 vs FFUC

Actually I was wondering about OSX and comparative "CPU load" while using these low latencies...

Re: Latency comparison: HDSPe vs FF400 vs FFUC

well?

Re: Latency comparison: HDSPe vs FF400 vs FFUC

If you are comparing Cubase/Nuendo in W7 vs OSX then forget about it!  Cubase and Nuendo scale much beter on the Windows side so a direct comparison is utterly pointless and not a direct result of your audio interface.  You are basically comparing the quality of the software porting and OSX's poor multi-core scaling at low latencies.

Cubendo.com and DawBench.com have lots of interesting Windows vs OSX comparisons if you want to research further.  Basically, a single CPU Quad Core Windows machine will beat a Dual CPU 8 Core Intel Mac Pro at the same latencies in Cubase according to the tests fryingpan

Natve = PC & RME for me - hands down

cool

MADIface-XT+ARC / 3x HDSP MADI / ADI648
2x SSL Alphalink MADI AX
2x Multiface / 2x Digiface /2x ADI8

Re: Latency comparison: HDSPe vs FF400 vs FFUC

Yeah, I concur, the Nehalem Octo Mac Pro's are rubbish at low latencies...even in SL/Logic Pro 9...I replaced an iMac with one, to no great impact other than loss of massive wads of cash... Sure, I could run shitloads more tracks at massive latencies: but so what, I can freeze tracks if I really need more.

But yeah I'm curious about OSx/Logic not Cubase, sorry didn't mention that...

But, anyway, I don't think I really made myself clear: essentially what I'm wondering is what kind of variance of CPU load there is between these different interfaces when using the same buffer setting?
(A difficult test, I know, but, if you have all three you could make a few general observations...)

8 (edited by Randyman... 2010-08-13 03:39:42)

Re: Latency comparison: HDSPe vs FF400 vs FFUC

Check DawBench.com.  You should find something loosely pertaining to OSX and different performance with different interfaces, but I doubt the FFUC and Babyface have made their way into the benches as of yet (but you should get an idea of PCI/PCIe vs FW)

And PS - While I'm not a fan of Apple Hardware regardless, it's the software (OSX) that falls on its face at low latencies, not the hardware, as Bootcamp will run Windows just as good as it does on a real PC! wink

cool

MADIface-XT+ARC / 3x HDSP MADI / ADI648
2x SSL Alphalink MADI AX
2x Multiface / 2x Digiface /2x ADI8

Re: Latency comparison: HDSPe vs FF400 vs FFUC

I can do a FF vs UC comparison once I find time, but the HDSPe isn't with me anymore.

People getting low latency dropouts should try running Coolbook (www.coolbook.se) with CPU load at 9 or 10. That will kepp Speedstep from clocking down your CPU.

Re: Latency comparison: HDSPe vs FF400 vs FFUC

I just remembered that I already did CPU comparisons of all three interfaces on OS X when the UC was fresh:

On my Macbook Pro 2.8 with Ableton Live I get the following OS X Kernel CPU loads for a heavy project that utilizes *all* I/Os:

HDSPe + MF2 (ExpressCard): 1%
Fireface 400 (Firewire): 1.5%
Fireface UC (USB): 2.5%

Ableton Live's own load with the Fireface 400 is higher than with the UC though so in the end they both use about the same (with the FF UC performing slightly better and at lower effective latencies).

Re: Latency comparison: HDSPe vs FF400 vs FFUC

Ok, that's kinda cool...Although Logic might be a bit different, I'd say it'd still be weighted fairly similarly?

I think it's convinced me to sell my Mac Pro rather than buy a HDSPe system for it and get a new iMac with a FirefaceUC instead.

Cheers,

12 (edited by imonium 2010-08-14 19:17:23)

Re: Latency comparison: HDSPe vs FF400 vs FFUC

( @Randyman... It's nuts to think that Apple with all their integration, being in full control with both the OS and the hardware, would fall behind PC's when it comes to low latency...bit of a joke really. Even Logic is disappointing with anything other than it's own Plugin's at low latencies...but I'll stop now, sorry, we'll say no more. )

Re: Latency comparison: HDSPe vs FF400 vs FFUC

I did some low latency comparisons with Logic 8 back then. Without "I/O Safety Buffers" OFF all three devices performed about the same.

With "I/O Safety Buffers" ON (=audio buffer size is doubled and 1 CPU core is reserved for audio device I/O) the UC could be used with one buffer size smaller than the other two. It's not much of a difference at these sizes though.