DAWbench - Reference Benchmarks :.


Audio Interface - Low Latency Performance :.


There are a wide and varied range of audio interfaces available, all with their strengths in regards to specifications and features. The current market is crowded with new interfaces coming online cramming more and more features at pricing even more affordable than anytime in the past , which on the surface seems like a win/win for those looking at dipping their toes into the pond, and for many it is just that.

However for those that place higher demands on their systems, the cracks start appearing very quickly when these interfaces are driven to lower latency, and as that becomes more and more important , a lot of the gloss of the extra features pales when the performance of the drivers do not hold up to the demands of the low latency working environments that many of us require. Now while a lot of these interfaces will perform well for the majority in less critical working environments , the performance variable can become quite dramatic at the lower latencies, and that is what I am specifically going to be focusing on.



Hardware Buffer Setting v Actual Latency :.    

All audio interfaces have their respective hardware control panels where adjustments to buffer /latency settings can be made.

The buffer settings in most case have little correlation to what the actual latency achieved due to various factors including added sample buffering for both the Input and Output streams and also added sample delays associated to the AD/DA conversions.

Its also worth noting that simply having the latency value available, doesn't guaranteed that the interface will work reliably at that latency.

  The actual latency can be measured in numerous ways, one method is to use the Centrance ASIO Latency Test Utility which will test round trip latency for both Digital and Analog Streams ( Analog taking into account the AD/DA ) , the other is by using a DAW application that accurately reports the actual Digital Input / Output latency reported by the ASIO driver.

In this instance Cubase 5.x, but that is not infallible as mixing and matching say a PCIe card with an AD/DA from another manufacturer can result in the AD/DA delay not being reported to the ASIO driver.

To further complicated things, some mixers with added FW interfaces that are becoming more popular do not even report the actual latency to the host DAW, only the theoretical value for respective buffer settings, which could cause synchronization errors.

To say that there is a wide and varied range of reported latencies across the interfaces is an understatement, and some developers have also chosen to have double buffers on playback to help loosen the reigns so to speak.

Reference Test Session / Systems :.    

The Reference test session is the DAWbench DSP Universal - RXC.

This test has a large data base of collated results over the years that we can directly correlate the results to, is freely available for numerous DAW applications, and even tho the results listed will be for Cubendo 5x the individual results for the respective interfaces will be relative to any application that utilizes the ASIO driver efficiently.


All interfaces have been tested on both Core i7 based desktop and notebook reference systems which have been specifically configured for low latency audio application.

Base reference hardware is the RME HDSP line of audio interfaces that have proven to be the top performers on both platforms , all other hardware will be graded comparatively against the base hardware results


Desktop System Detail:
Intel i7 920 Quadcore/ 2.66 GHZ/
Intel X58 / 6 GB DDR3-PC12800.

Notebook System Detail:
Intel i7 620 Dualcore / 2.66 GHZ/
Intel HM55 / 4 GB DDR3-PC10600.

O.S Detail:
Windows 7 x64 Pro


Round 1 : Desktop System :.

The RME HDSPe card is the only PCIe interface in the current list of interfaces and its not overly surprising that it would be the better performer overall. It is also specifically why it has been chosen as the baseline comparative for the other interfaces that are FW400 - the RME Fireface 800 can also run under FW800 but in this instance it was running under FW400.

The RME HDSPe card delivers impressive performance right down to 032 samples , the drivers have always been at the top of the heap in regards to stability and ultra low latency performance over the years, and continue to do so.

With the remaining 4 interfaces all being Firewire, what is clearly evident is that not all FW interface controllers and drivers are equal.

The RME Fireface 800 has always been one of the most respected and proven in regards to low latency performance , as evidenced by the results from 064-256 samples , there are small added buffers at respective latencies over the HDSPe, but overall the performance is quite impressive and very close to PCIe


Next in line is the M-Audio Profire 610. M-Audio over the years have had a mixed bag with driver performance , but the Profire 610 proved itself an extremely worthy opponent to the RME's FW low latency performance crown .

Despite not performing quite as well , I was pleasantly surprised and duly impressed at the performance , so hats off to AVID/M-Audio.

Now we come to the Focusrite and Presonus , and as evidenced by the results , these interfaces are not even on the same playing field as the previous, actually to be more precise, not even in the same ball park.

The results for the 2 interfaces are near identical and on further investigation I discovered they share the exact same under bonnet OEM Firewire controller and base drivers , which Focusrite and Presonus have licensed from TC Electronics - Dice II. It is well known that the Dice II has been less than stellar in the performance department, but I think the results may still surprise some.


LLP - Low Latency Performance at 064 samples was terrible , with the session clearly breaking up at around 40-50% ASIO load , CPU loadings were even less , so it was severely hobbling the potential resource overhead of the system.

Comparative performance against the M-Audio Profire 610 and RME FF800 was in the range of 39-43% at 064 thru to 75-78% at 128 to around 88% at 256.

While the performance starts leveling at 256 for those working at the lower latencies the variable is quite substantial, and would be even more so on current higher clocked Quad and Hexacore systems.

The LLP for the Dice II powered interfaces highlights clearly just how important the dedicated FW controllers are and just how detrimental they can be to the overall performance.



Round Two : Notebook :.    

Results for the notebook showed a very similar pattern to the desktop system but did have a few variables in the results that are worth noting.

The RME HDSPe Express Card connected to a Multiface II in this instance, again clearly outperformed the FW interfaces as expected , but the results for the FW interfaces were not as consistent as on the desktop.


Interestingly the M-Audio Profire 610 and the RME Fireface 800 swapped roles in this round, with the RME slightly trailing the the M-Audio unit at 064-128.

The Presonus and Focusrite results were not as in sync as they were on the desktop , the Focurite being better at 064 , but then trailing the Presonus at the higher latencies.


The results for both interfaces was again substantially down on the other FW units - comparative performance at 064 was around 35% , at 128 around 70% and 256 it leveled out to 90-95%.

Well at least we can conclude that the Dice II powered units have remained consistent, unfortunately its in the wrong direction.

Conclusion :        

Its quite clear from the testing results that for those that require ultra low latencies , the PCI/PCIe alternatives are always a good option .

Sadly there aren't a lot of professional PCI/PCie cards currently available on the market , developers opting more and more for FW/USB 2 as they can service a far wider market on both desktop and mobile , which is understandable in such a niche market such as ours. I can't help feeling that for some of the developers the focus has definitely shifted away from performance , which I find quite concerning.

LLP- Low Latency Performance is extremely important for users who require those lower latencies in their work flow, prime example being those that compose with virtual instruments and also guitar amp simulators where the round trip latency also comes into play , more so than those who are doing simple tracking of bands and mixing, for example.



However its increasingly common for audio interface units to be marketed , reviewed and sold on the bells and whistles , more so than the actual performance, which for some can lead to a rude awakening when the interface delivers poor performance at the preferred working latencies.

I have read countless reviews over the years in the mainstream media of audio interfaces and many barely scratch the surface on the LLP of the units. Some may throw a cursory glance at winding the latency down to its lowest available setting and running some low to moderately loaded sessions and see if they can playback without artifacts, but that is not telling the whole story.

IMO , too many in the industry don't have an interest past selling a box , they will simply point to the lowest available latency setting in a control panel as evidence that the interfaces are capable of working at that level , and its increasingly frustrating to find those settings are nothing more than Window dressing.


In the next round of testing I'll include some PCIe cards from different ends of the spectrum

One being the AVID PTHD Native Solution that is aimed primarily at Protools HD obviously , but it will be very interesting how the ASIO driver performs, and the second being an ESI PCIe card, ESI in the past have been known to deliver exceptional LLP drivers , so it will be good to to get some more variables into the mix.

I will also be including the new Mbox Pro 3 that for the first time AVID has offered sans PT software, specifically marketing it as a stand alone audio interface in its own right.

You can follow the ongoing testing on the thread at the DAWbench Forum Here

Vin Curigliano
AAVIM Technology
June 05 2011

Part I | II |III

© AAVIMT 2005-2011