Audio Interface - Low Latency Performance Data Base : [Archive]

TAFKAT

11-20-2009, 10:39 PM

http://www.dawbench.com/images/dawbenchdsp-audioint-notebo.jpg

http://www.dawbench.com/images/dawbenchdsp-audioint-deskto.jpg

Hey All,

I haven't done any detailed comparative shootouts between audio interfaces for quite a while, and with the focus on Windows 7 and Snow Leopard I thought it would be a good idea to get some insight on how the various hardware is shaping up on both of the new O.S's.

Firewire interfaces will be of particular interest on Win7 due to the change in the native driver , and we also have to make particular note of what the actual reported latencies are compared to the respective settings for the hardware, as they vary quite dramatically depending on whether extra buffers are utilized - especially on playback.

In the coming weeks I will post the results for all of the hardware I have at my disposal, but would be interested in hearing from any member who can chip in with hardware that I don't have access to.

I currently have RME PCIe/Firewire, MOTU PCI, M-Audio Firewire/USB 2.0 hardware that I can test cross platform.

To maintain consistency with my dev system a preferred spec would be i7-920, Cubase 5.x or Nuendo 4.x , XPSP3 / Win7 , OSX 10.6.1

Of course others are more than welcome to post results from other systems, but it would be good if those posting results outside of the base spec have results for at least 2 bits of audio hardware for comparative analysis.

The test results will be using the Cubendo DSP test session , as there is already a preliminary collated data base to compare to.

:009:

eightyeightkeys

12-07-2009, 02:39 PM

Yes, Vin, this would be very useful since I'm a composer for film and tv and really need to able to play many VSTi's in real time to achieve the desired feel and emotion. Latency, or, having to increase buffer sizes just ruins it for me. I then turn into a programmer rather than a performer.
What were your results for the RME PCIe card ? Was it the AIO ?

TAFKAT

12-07-2009, 07:10 PM

The latest Windows 7 and OSX SL results were based on the AIO.

FWIW: I have not seen any quantifiable difference between the PCI and PCIe HDSP cards , with the driver being unified I have achieved identical performance between both versions.

nikki-k

12-09-2009, 02:10 AM

Hi!
Rig in sig, MR816X via onboard firewire. If I have all that is needed, could run some tests if you want. Supposed to have new W7x64 approved drivers this month I think?

TAFKAT

12-09-2009, 03:13 PM

Hey Nikki,

Thanks for the time and energy , I'll definitely take you up on the offer of testing the Steini MR unit on Win 7 for me , I'll get some added detail to you via PM.

:009:

TAFKAT

03-26-2010, 02:00 AM

Hey All,

Its taken me a while to get this rolling , but I have been doing some qualifying and testing on my current i7 Notebooks the last few days and have documented not only the results but the actual I/O latency for each respective buffer size as they differ greatly between interfaces.

Here are some numbers :

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

FF800 - 2.9992 - Native FW Mode

048 : 39 RXC - Aero : 48 RXC - Basic : I/O - 2.109/3.265

All further results are with Basic Mode :

064 : 54 RXC : I/O - 2.472/3.628

128 : 59 RXC : I/O - 3.923/5.079

256 : 64 RXC : I/O - 6.825/7.982

--------------------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

FF800 - 2.9992 - Legacy FW driver mode:

048 : 56 RXC : I/O - 2.109/3.265

064 : 58 RXC : I/O - 2.472/3.628

128 : 64 RXC : I/O - 3.923/5.079

256 : 68 RXC : I/O - 6.825/7.982

--------------------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

M Audio Profire 610 - 2.0.5.10.10.0.582 - Legacy FW driver mode:

064 : 63 RXC : I/O - 3.152/3.152

128 : 65 RXC : I/O - 4.603/4.603

256 : 68 RXC : I/O - 7.506/7.506

-------------------------------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

Focusrite Saffire Pro 40 - 2.0 - Native FW driver mode - as specified by Focusrite :

064 : 23 RXC : I/O - 1.905/3.900

128 : 45 RXC : I/O - 3.356/5.351

256 : 60 RXC : I/O - 6.259/8.254

-----------------------------------------------------------------------

Some quick notes - The FF800 actually performed better with the Win7 FW driver set to legacy over native TI at the lowest latency , despite the fact that RME have stated the current driver does not require the controller to be set to legacy.

The M-Audio Profire performed well under legacy as well, didn't test it under native as yet - M-Audio do not make any reference which is the preferred mode. I have been pleasantly surprised with the M-Audio Profire line across both XP and Win7 , holding their own against the RME interfaces that have been the benchmark of FW Audio interfaces for low latency performance over the years.

The Focusrite Saffire Pro 40 is a mixed bag, great build quality, great preamps, but as an interface, well lets just say its not on my preferred or recommended list. There are a few modes in the Focusrite buffer control , which alters how large the added playback buffer is , I had it set to Low to be able to get to the 064 lowest setting. Standard is Medium which then only allows a 96K low buffer, but all playback buffers are further padded by a fair bit i.e @ 128 samples Output was upped to 7.347 over 5.531 @ 256 it was 10.249 over 8.254 : Input buffers remain the same as the Low setting. Using the default Medium setting the results at 128/256 were much the same despite the added buffers.

As with a lot of other manufacturers who use OEM firewire controllers , the Focusrite performed miserable when pushed to any type of decent loading , drop outs starting at very minimal ASIO loadings of around 50% - despite have larger added buffers on the output, overall the drivers are below average, IMO !

I'll get to some testing of PCI/PCIe cards on the desktop systems next..

efernan

03-26-2010, 05:14 AM

Hey Vin, as you said, newer Saffires sucks at low latencies (the older line was average, also with padded buffers). The same happens with various Presonus models. They all use the same DICE II Firewire controller if I'm not wrong.

I always had great results with M-Audio Firewire (and USB) interfaces. With both the older Firewire (yeah, original name) and the newer ProFire lines.

MOTU is another brand that has good Firewire drivers, starting from 828 Mk II (the first model was crappy). I don't have any of these now, but I'll try to get one for the benchmark.

Cheers!

Beat Productions

03-26-2010, 02:35 PM

Vin:
Thanks for the tests! I've found that the latencies are marginally better on the Saffire Pro40 with the new 2.0 drivers (W7 64bit).

TAFKAT

03-29-2010, 01:45 AM

Hey E ,

Yeh , I am not a huge fan of the box mover mentality audio manufacturers who use these OEM controllers, but its not only the lower end, I had a run around with a Prism Orpheus a while back as well , I posted a thread about it here as it was doing my head in , its a hoot.., check it out Here (http://www.cubendo.com/showthread.php?449-Prism-Sound-Orpheus-Test-Drive)

I'll keep posting numbers for different audio interfaces as I get in front of them.

@ BP,

I did use the Version 2.0 driver, I made a note of the driver version number after the interface , but its may not be clear. I 'll amend that when I get a chance.

Beat Productions

04-01-2010, 02:02 PM

Vin:
I ran a couple of tests, I limited it to 128 (196) because I need relatively low latency. My basic setup uses a RME HDSP Digiface as my main interface with a Saffire Pro40 and Presonus Firestudio 2626 connected via adat. I find this is an extremely eloquent solution both for latency (RME still has the best drivers), and RME Totalmix, IMO the best software mixer there is and TOTALLY INVALUABLE. However, I was pleased to see that the Saffire Pro40 performed equal to the RME HDSP at 192 samples (on my overclocked I7 920).

Cubase 5.1.2

Core i7 920 3.20 GHZ:

6GB PC1600 DDR 3

Windows 7 x64 Pro

Focusrite Saffire Pro 40 - 2.0 - TI chipset driver

128 : 80 RXC : I/O - 3.356/5.351

192 = 120 RXC I/O - 4.807/6.803

Cubase 5.1.2

Core i7 920 3.20 GHZ:

6GB PC1600 DDR 3

Windows 7 x64 Pro

RME HDSP Digiface 3.08.2

128 : 120 RXC : I/O - 2.948/3.673

TAFKAT

04-01-2010, 05:04 PM

Hey B,

Thanks for the time and energy , much appreciated.

One thing to note is the actual I/O for the interfaces , as the Playback buffers are a lot higher on the Focusrite .

I'll run some numbers for the Fireface 800/Profire/Focusrite on the development X58 system on Win7 soonish.., and compare the to your results.

BTW: @ 128 - I managed 158 RXC out of HDSP/i7-920/X58 Rig running an HDSPe AIO, also had identical results with a HDSP 9632 - so I would suggest you have some more headroom there especially at 3.20 GHZ.

I have a clients system with a Xeon 3550 @ 3.06 GHZ on the bench at the moment that will be using a FF800, so I'll run some numbers and post those as well.

Peace

TAFKAT

04-03-2010, 12:27 AM

Some more numbers ,

---------------------------------------------
DAWbench DSP RXC :

Nuendo 4.3.0

Xeon 3550 @ 3.06 GHZ:

12GB PC10600 DDR 3

Windows 7 x64 Pro

FF800 - 2.9992 - Legacy FW driver mode:

048 : 149 RXC : I/O - 2.109/3.265

064 : 160 RXC : I/O - 2.472/3.628

128 : 163 RXC : I/O - 3.923/5.079

256 : 165 RXC : I/O - 6.825/7.982

---------------------------------------------------------------------------------

Amazing low latency performance for a Firewire interface under Win7 Legacy, and interestingly not much scaling above 064 samples.

The client could comfortably set and forget this rig at 064 IMO.

@ Beat,

I double checked my results for the HDSPe AIO on not only my dev system which has a i7 920, but also a clients Xeon 3520 , both chips run at 2.66GHZ and are identical performance wise, the results @ 032 were 132 RXC on both systems , right on the dime. Same as the initial results I achieved with the older PCI HDSP 9632 card. Can you please run up your Multiface @ 032 and report back if you get a few minutes , Thanks.

Peace

TAFKAT

05-10-2010, 09:09 PM

More i7 notebook results, this time with RME's EXpresscard solution..

--------------------------------------------------
DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

RME HDSP EXpress/Multiface : Driver 3.083 :

032 : 65 RXC : I/O - 1.451/2.177

064 : 70 RXC : I/O - 2.177/2.902

128 : 73 RXC : I/O - 3.628/4.354

256 : 76 RXC : I/O - 6.531/7.256

------------------------------------------------------

Great low latency results , surpassing any of the FW options easily.

TAFKAT

05-10-2010, 09:39 PM

Presonus's turn..

--------------------------------------------------
DAWbench DSP RXC :

Cubase 5.1.2

Core i7 620 2.66 GHZ:

4GB PC10600 DDR 3

Windows 7 x64 Pro

Presonus Firestudio Mobile : Driver 3.5.2 :

064 : 20 RXC : I/O - 1.905/3.900

128 : 48 RXC : I/O - 3.356/5.351

256 : 65 RXC : I/O - 6.259/8.254

------------------------------------------------------

Hmmmmm, very similar to the Focusrite results, which I am not overly surprised about as they are more then likely using the same questionable 3rd party FW controller..

TAFKAT

05-12-2010, 07:34 PM

Some numbers for the FW Audio interfaces on the i7 development box.

--------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 920 2.66 GHZ:

3GB PC10600 DDR 3

Windows 7 x64 Pro

Presonus Firestudio Mobile : Driver 3.5.2 :

064 : 58 RXC : I/O - 1.905/3.900

128 : 112 RXC : I/O - 3.356/5.351

256 : 137 RXC : I/O - 6.259/8.254

--------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 920 2.66 GHZ:

3GB PC10600 DDR 3

Windows 7 x64 Pro

Focusrite Saffire Pro 40 - 2.0

064 : 57 RXC : I/O - 1.905/3.900

128 : 112 RXC : I/O - 3.356/5.351

256 : 137 RXC : I/O - 6.259/8.254

----------------------------------------------

Quick note : If you notice the actual I/O latency is identical between the 2 units, as are the results.

This is indicating to me its the same 3rd party OEM FW controller and base driver.

-------------------------------------------------------------------

DAWbench DSP RXC :

Cubase 5.1.2

Core i7 920 2.66 GHZ:

3GB PC10600 DDR 3

Windows 7 x64 Pro

M Audio Profire 610 - 2.0.5.10.10.0.582 - Legacy FW driver mode:

064 : 133 RXC : I/O - 3.152/3.152

128 : 144 RXC : I/O - 4.603/4.603

256 : 152 RXC : I/O - 7.506/7.506

-----------------------------------------------------------------

The M-Audio Profire unit again returns a great result across the board, the difference compared to the 2 previous interface cannot be emphasised enough.

--------------------------------------------------------------------

DAWbench DSP RXC :

Cuabse 5.1.2

i7 920 @ 2.66 GHZ:

3GB PC10600 DDR 3

Windows 7 x64 Pro

FF800 - 2.9992 - Legacy FW driver mode:

048 : 141 RXC : I/O - 2.109/3.265

064 : 148 RXC : I/O - 2.472/3.628

128 : 151 RXC : I/O - 3.923/5.079

256 : 154 RXC : I/O - 6.825/7.982

------------------------------------------------------------------

RME Fireface again topping the list for low latency performance

Anyone who questions that an audio interface is only as good as its driver needs to have a good read of the above.

I'll create some nice graphs shortly so its easier to digest.

FWIW: I was going to do some OSX comparative for all of the interfaces, but after seeing the actual reported results for the set buffers , I concluded they are a farce, so I'll stick to Win7 for now.

i.e : Presonus driver on Win7 / OSX

Windows 7 x64 Pro / OSX 10.62

Presonus Firestudio Mobile : Driver 3.5.2 :

032 : I/O - Win7 - N/A : OSX 10.6.2 - 4.807/4.807

064 : I/O - Win7 - 1.905/3.900 : OSX 10.6.2 - 5.533/5.533

128 : I/O - Win7 - 3.356/5.351 : OSX 10.6.2 - 6.984/6.984

256 : I/O - Win7 - 6.259/8.254 : OSX 10.6.2 - 9.887/9.887

Say No More......

efernan

05-12-2010, 11:24 PM

Hi,

Quick note : If you notice the actual I/O latency is identical between the 2 units, as are the results.

This is indicating to me its the same 3rd party OEM FW controller and base driver.

Yes Vin, that's right. Both interfaces uses the Dice II chip from TC Applied Technologies. This company was started by people from TC Electronic (TC Electronic's interface line also uses the same ASIC chip). You know, TC Electronic's interfaces always were underperforming.

http://www.tcelectronic.com/product_news.asp?AjrDcmntId=336

As you already noticed, this chip is not good enough for low-latency/high-loads. I don't recommend to anyone any Firewire interface that is based on this chip.

Cheers!

efernan

05-12-2010, 11:30 PM

Some numbers for the FW Audio interfaces on the i7 development box.

BTW... Could you post PCI/CPIe card (like RME HDSP) performance/latencies of that same system?

Thanks!

TAFKAT

05-13-2010, 12:06 AM

Here ya go..

---------------------------------------------------------------
DAWbench DSP RXC :

Cuabse 5.1.2

i7 920 @ 2.66 GHZ:

3GB PC10600 DDR 3

Windows 7 x64 Pro

RME HDSP AIO : Driver : 3.083

032 : 130 RXC : I/O - 1.746/2.517

064 : 151 RXC : I/O - 2.472/3.243

128 : 158 RXC : I/O - 3.923/4.694

256 : 163 RXC : I/O - 6.825/7.596

-----------------------------------------------------------

One interesting thing of note : The latencies on the Express card/Multiface combo are slighter lower :

2407WFPHC

05-19-2010, 04:42 PM

Any chance of getting some SSL card comparisons (MADI Extreme and the MX4 if possible)?

TAFKAT

05-19-2010, 05:24 PM

Wow, thats freaky..

I have an SSL MX4 / Alpha Link on its way as we speak.

The Win7 64 Bit drivers are still a little green , but should be final shortly I am told.

I'll post as soon as I have something solid.

2407WFPHC

05-19-2010, 07:16 PM

Sounds good! XP-32 comparisons are appreciated as well.

TAFKAT

05-19-2010, 08:49 PM

Hmmmm,

I'll have to dig out the old XPSP3 install ..., I'll give it a run up for you if I get some time.. :-)

2407WFPHC

05-19-2010, 09:24 PM

Don't go out of your way for me. I'm sure the scaling will be similar, I just mainly wanted an RME HDSP vs SSL comparison (SSL only goes to 64 Samples plus a 64 sample safety buffer!). Thanks!

Beat Productions

05-20-2010, 10:41 AM

I'd be interested how the Steinberg MR816 X stacks up against all the other firewire units tested.

nikki-k

05-20-2010, 04:01 PM

I'd be interested how the Steinberg MR816 X stacks up against all the other firewire units tested.

I am as well.
Unless Vin has one or someone else has at least one of the other pieces, would it be possible to produce comparative results? While two people can have same/similar systems, unless they are (close to) identical in setup/software, wouldn't there be a decent amount of room for variation? When I ran DPC, initially I got 100-200. After altering one registry entry, it dropped to 15-25. I figure there are other variables I do not even know about, and possibly some that may unknowingly hurt performance.

2407WFPHC

05-20-2010, 04:53 PM

When I ran DPC, initially I got 100-200. After altering one registry entry, it dropped to 15-25. I figure there are other variables I do not even know about, and possibly some that may unknowingly hurt performance.

Which registry setting was that? IRQ8 Priority? I can never get W7 as low as XP - XP easily runs in the single-digits to teens with the usual tweaks where W7 is in the 100-150 area on the same hardware. I know others claim anything below 100 is pointless - but I suspect otherwise.

I have yet to get Vista or W7 to match XP in terms of low latency scaling (64 or 32 samples) using identical hardware and software.

Thanks.

TAFKAT

05-20-2010, 06:08 PM

I am as well.
Unless Vin has one or someone else has at least one of the other pieces, would it be possible to produce comparative results? While two people can have same/similar systems, unless they are (close to) identical in setup/software, wouldn't there be a decent amount of room for variation? When I ran DPC, initially I got 100-200. After altering one registry entry, it dropped to 15-25. I figure there are other variables I do not even know about, and possibly some that may unknowingly hurt performance.

Hey Nikki,

Your results would be close enough to par to my dev/test system I suspect, so can you please post the Yamah results.

Just use the layout from one of my earlier pots as a template , listing the actual I/O latency reported by Cubase at each buffer setting.

Re dropping the DPC that low on Win7, I know anther way to get it that low, but I don't do it. There is a powermode setting in the BIOS called HPET - High Precision Event Timer, that if disabled will give you those low DPC results in Vista/Win7, so I am guessing that is what the reg hack is related to. I have looked into it and it directly relates to synchronising media streams , it was not supported under XP, so it was ignored, thats why there was such a low DPC reading , but it is an integral part of Vista/Win7. I have heard mixed reports from users toying with it.

My advise is that the DPC of 100-200 is not going to significantly impact low latency performance , my desktops sit around 90-100 , the notebooks a bit lower, go figure , and without knowing exactly what disabling HPET could impact in regards to audio, it would be wiser to leave it on.

TAFKAT

05-20-2010, 06:10 PM

I have yet to get Vista or W7 to match XP in terms of low latency scaling (64 or 32 samples) using identical hardware and software..

Thats interesting..,

Win7 64 has proven to be on par, if not a little better for me on the current chips right down to 032 samples.

2407WFPHC

05-20-2010, 06:36 PM

I don't know why I can't get similar performance, but XP always edges out W7 in a dual-boot setup on the same system. The difference is not drastic - W7 generally requires the next highest buffer to remain clean compared to XP on C2D's on 965 and P45 chipsets. The DPC's are really the only thing that was drastically different (2-15 on XP, 100-150 in W7 - both heavily optimized in the old XP stripped down "offline" fashion).

I really think my heavily tweaked XP installs just kick that much ass to be honest! I've tried this at least 6 times now (fresh installs of W7 each time) and XP always wins :eusa_think: I'll absolutely look into disabling the HPET on W7 - sounds like it might be a winner!

TAFKAT

05-20-2010, 06:44 PM

I hear ya..

I still am amazed when I boot into my optimised XP installs , just how responsive and fast they are, but the times they are a changing.. :-)

I think that the difference maybe that Win7 is taking advantage of the Core i7 inherent lower latencies better than on Core 2 , to be honest I never ran Win7 on Core 2.

nikki-k

05-21-2010, 03:45 AM

Ahhh! Yes- I believe it was HPET that I disabled via Registry. I compared the two states, and off worked fine for me, so I figured I would leave until something happens to change things. FInding a MIDI I/O was a trick and a half, and once that got resolved...

I will try to do numbers soon. Health has been horrible, no music making for over two months now. I do not have Cubase on my laptop due to the stupid dongle. I know I would snap that sucker off, and while I do typically use a USB extension, I simply do not want to deal with that. If Reaper had scoring/TAB and VariAudio-like features....

TAFKAT

05-23-2010, 02:33 AM

TAFKAT

05-25-2010, 11:27 PM

Added some graphs to the original post..

TAFKAT

06-26-2011, 11:52 PM

Time to push away the tumble weeds , brush off the cobwebs and blow the dust off this thread.. :-)

I have been doing a whole stack of testing lately with a bunch of interfaces across PCIe, FW and USB 2.0 using both DAWbench DSP and DAWbench VI to give a better cross reference.

Here are the results for the first batch of audio interfaces on the new round of testing.

http://www.dawbench.com/images/dawbench-audioint-06-11.jpg

Explanation of how the LLP- Low Latency Performance Rating is derived.

The results for the DAWbench DSP RXC across the latencies of 032 thru to 256 ( which has been the M.O for the last 5 years ) are added and the total is then % wise gauged against the result for the RME HDSPe AIO baseline card. The same is then calculated for the DAWbench VI CV/ NCV tests for 032-512.

Those 3 % results are then added and divided by 3 to give an average % .

I thought it important for the I/O and RTL figures to be an influencing factor on the rating as some cards have a lot lower overall latency than others, so the average % results is then multiplied by the last % result for the RTL.

How the RTL % is calculated is I combine the total of the RTL's across the specific available buffer settings for the cards ( all cards past the base reference RME were calculated using 064-512 except the M-Audio FastTrack which had a range of 128-512. You will notice 2 values listed for RTL for the RME HDSPe AIO , first being for 032-512, second being 064-512 ) , and then calculate the % variable against the baseline. You will see that some cards are actually better than the RME AIO card in regards to I/O /RTL i.e RME Babyface and the AVID MboxPro 3.

I think that is a fair appraisal using the collated data, and it gives deserved credit and advantage to those cards that do have lower individual In/Out and Round Trip Latencies.

Some cards do not report the correct I/O values to the DAW Hosts , only reporting the nominal value for the respective buffer sizes. In those instances I will make a note and default the RTL rating to .95 which is below the average of the other cards RTL % rating of around .97 , not perfect but a reasonable compromise.

A few side notes - the baseline reference is the RME HDSPe AIO card using the 3.08.5 driver. You will notice that even the new version 3.24 driver is also tested and gauged against that baseline result as the performance has dropped. If a later RME HDSPe driver levels or betters the 3.08.5 driver, then it will become the new baseline, unless of course another card has better overall performance and becomes the new baseline.

The performance of the M-Audio Fast Track Ultra was actually quite good for a USB2 interface at the available latencies, but its rating suffered due to the lowest available buffer being only 128. The results for the RME Babyface was very impressive for a USB 2 interface , edging out the FW - Profire 610 in the rating due to its lower overall I/O and RTL. The Result for the Focusrite unit using the dreaded Dice II FW controller , speaks for itself !

More results and analysis coming in the next few weeks

Stay Tuned.

Bman

06-27-2011, 12:51 AM

Fanstastic Vin. Interesting reading. I'm thinking about going multichannel for surround work soon so how do you think something like the multiface II would compare? Same/similar driver?

Brett

TAFKAT

06-27-2011, 02:51 AM

Hey Brett,

Same driver , slightly lower I/O & RTL due to the AD/DA's being a little tighter , performance would be on par to the AIO.

The Multiface II is only 96K and lacks the SteadyClock feature, not a deal breaker but something to be aware of.

Bman

06-27-2011, 03:04 AM

Nice - Thanks Vin

Animus

06-27-2011, 02:12 PM

Good work as usual Vin!

TAFKAT

07-14-2011, 10:12 AM

O.K, here are the final results for the current series of testing.

http://www.dawbench.com/images/dawbench-audioint-06-11-3.jpg

Some notes on the latest interfaces tested.

The MOTU 424 PCI / 24 I/O delivered exceptional performance right down to 032 samples , and despite the overall numbers being below that of the RME reference card, if you take a closer look the I / O and RTL are measurably lower as the MOTU uses very minimal padding on the buffers. By taking that into account using the RTL % rating of 1.17 , it drew level with the reference RME HDSPe.

The MOTU's have always drawn mixed reports on Windows and I know first hand how flaky the cards can be having 3 different revs of 424 card across the bench in the last few weeks, and its a crap shoot whether the respective revs will work on current systems , but the drivers themselves have always been good with the PCI/PCIe cards, ( as long as you don't use external clocking- LOL )

The ESI U46XL was the USB audio interface that I had hinted at a while back that surprised me and the results speak for themselves. Yes the reported latencies are nominal and it has the oddity of having the highest buffer setting of 256 ( actually one step higher to 288 ) , so I had to do some juggling with calculating % ratings , but seeing that this whole exercise is about LLP , I gave ESI some rope with the interface not having a 512 setting. Great little interface , especially for it being USB2.

Staying with ESI, the Maya 44e is a mixed bag, respectable performance although beaten by the U46 , but the card has an issue where the driver has to be reset in Cubendo every time you open a project with any buffer setting below 512 ??? !!

PITA - Still waiting to hear back from ESI who have initially said they can not reproduce it.

The Steinberg MR816 delivered decent overall performance, latency settings available all the way down to 032 samples, mind you that's with the help of some extra padding on the output buffers, which resulted in higher RTL values and subsequent lower comparative RTL %. The 032 setting is more Window dressing IMO.

SSL Nucleus - the results don't tell the whole story as the unit is so inconsistent across multiple systems and even respective USB ports. The results are pretty decent but it took some major arm wrestling to settle the system to achieve those results. The driver panel still looks like the vanilla OEM breadboard and the odd non standard buffer sizes are a sore point if you want to try and use it with Protools 9 - which it is advertised as being compatible with - as PT9 simply will not accept the odd buffers.

The Mackie Onyx actually performed O.K for a Mixer/FW combo, DAWbench DSP were quite good, but it collapsed under the VI testing - the OEM controller definately feels different to some of the Dice variations, so not sure what Mackie are using there.

The Midas Venice was an interesting one, basically a mid size live desk with a Dice FW hacked into the back - I/O - RTL were identical and performance was close enough to identical to the Focusrite Sapphire , so its obviously using the same ( poor performing ) controller.

AVID Mbox 3 is identical to the M-Audio FTU range with slightly better I/O and RTL , not too hard to see that AVID are using the same under bonnet components across the AVID Mbox 3 Pro / Mbox 3 / Profire / FTU ranges.

AVID HD Native card under ASIO , well to say it was an interesting experience is an understatement. Input buffers are double the usual, Output are reporting Nominal , overall performance isn't too bad right down to 032 samples, but there were some major hoops to activate the ASIO driver in Cubase . i.e : needing to toggle off the ASIO driver back to the Generic and then back again to simply wake the driver on every session load .

Its obvious to me that the card is specifically geared to Protools , the ASIO driver really being only for convenience for some compatibility with other DAW hosts , so its not really a player in the current comparative.

Thats it until the next round..

Peace

TAFKAT

08-23-2011, 03:19 AM

Hey All,

Here are the latest results with some added interfaces, a new baseline and amended ratings.

http://www.dawbench.com/images/dawbench-llp-08-11.jpg

http://www.dawbench.com/images/dawbench-llp-08-11-2.jpg
Notes -

First off we have a new Baseline Interface - RME HDSPe AES : ADI8-QS - results across the 3 respective benchmarks were identical to the previous baseline AIO card , proving that the performance of the driver is consistent across the HDSP/HDSPe line. The I/O and RTL was significantly lower due to the tighter AD/DA. A few points, the reported I/O values in Cubase were sans the AD/DA , ( common amongst interfaces when using AD/DA connected via standard AES ), I calculated the added latencies with my RTL Utility which reported an extra 28 samples across the board - 14 samples respectively for the AD/DA ( RME spec is listed as 12 ). The RTL Values are including the AD/DA and all other interface RTL ratios have been amended.

MOTU 424 : 24 I/O - I discovered was not reporting the AD/DA , so the correct RTL was calculated using the Utility and values amended.

LynxTwo C - I was hoping to test a current AES16e and Aurora combo but the loan units didn't eventuate so I dusted off an old PCI Lynx TWO C card to at least get a heads up on the performance of the current driver. Double Playback Buffering has been disabled ( still have no idea why Lynx thought it needed implementing ). The card performed extremely well only being bettered by the 2 PCIe RME units , its a shame that these PCI legacy cards are not suitable for the current Intel chipsets with the non native PCI, hopefully Lynx will eventually update the Lynx TWO line to PCIe as well.

RME Fireface UFX - This was a unit that I know a lot of end users are interested in seeing the results for and I was keen to get it back on the bench after my preliminary experience with a very early unit was less than stellar under FW. In the meantime there has been numerous firmware updates as well as drivers , but I actually used an earlier driver sans the MultiClient option to remain consistent with the earlier Fireface 800 testing. I was pleasantly surprised that the unit under FW had improved measurably to the point of it actually bettering the FF800 results. Thats quite an achievement IMO considering the unit no longer has a dedicated FW controller. USB2 performance was also exceptional , however slightly below the FW results , which may surprise a few people considering the greater focus has been on the USB2 aspect. All in all, a big thumbs up to the RME devs on ironing out the early quirks with the FW on this unit , an exceptional performer on both FW and USB2.

ESI U46 XL - The initial good impression I noted on the preliminary testing changed significantly when I discovered that ESI by reporting the nominal values only, were actually masking large safety buffers as well as double buffering on playback. The calculated RTL via the utility showed significantly higher values than most other interfaces at the respective latencies. RTL % and LLP ratings have been amended and have dropped accordingly. There are also issues with the unit not initialising correctly on current H67/P67 chipsets.

Echo AudioFire 12 - As I noted earlier , this unit has large safety buffers as well as double buffering on playback to the point that the nominal values reported in the Echo Control Panel are almost comical. Performance was solid enough but the high I/O and RTL need to be taken into account when weighing up the figures. The excessively high RTL dropped the over all LLP rating substantially, as will be the case with any interface employing large safety and double buffering.

I have also removed some interfaces from the results pending further testing.

I am currently further investigating some of the listed interfaces which I suspect may also not be reporting the AD/DA. Once I have confirmed those results I'll amend the LLP ratings accordingly. Interfaces under investigation - AVID/M-Audio/Focusrite/Midas.

Peace

Beat Productions

08-23-2011, 12:04 PM

Vin:
Thanks so much for this. I have been watching the new Fireface UFX as well. Pardon my ignorance, but what does "MC off" mean on the RME interface I'm using a Digiface/PCI setup)?

TAFKAT

08-23-2011, 09:47 PM

Hey,

MC Off is the new MultiClient Mixing Option - Off.

Sorry Mate, I did explain it on other active threads I have covering this latest testing , I found that the 3.24 was not performing as well as the earlier 3.085 driver , and with some help from MC at RME we tracked it down to the new MultiClient Mixing Option.

The performance variable isn't huge but measurable when pushed to the limits, also the behaviour of the driver at that break point is distinctly different , its quite severe with the new option enabled.

Peace

Bman

08-23-2011, 11:36 PM

Fantastic Vin. Watching with interest. As you know I have a Lynx L22 PCI card - good to see the driver is performing comparably well

its a shame that these PCI legacy cards are not suitable for the current Intel chipsets with the non native PCI, hopefully Lynx will eventually update the Lynx TWO line to PCIe as well.

As you know, I'm looking at moving to a RME multiface and so the above comment caught my eye. Which chipsets have non-native PCI?

Cheers - Brett

TAFKAT

08-24-2011, 06:08 AM

As you know, I'm looking at moving to a RME multiface and so the above comment caught my eye. Which chipsets have non-native PCI?

Cheers - Brett

Hey Brett,

Anything 6 Series or beyond, so H67/P67/Z68 ( socket 1155 ) and the coming X79 ( Socket 2011 )

Bman

08-24-2011, 06:27 AM

Hey Brett,

Anything 6 Series or beyond, so H67/P67/Z68 ( socket 1155 ) and the coming X79 ( Socket 2011 )

Cheers Vin. Good to know. picking up a second hand RME pci card rather than pcie would definately be a mistake (seen a couple on ebay recently)

Beat Productions

10-07-2011, 02:39 PM

Vin:
Have you tested any of the MOTU "hybrid" interfaces in USB mode?

TAFKAT

10-08-2011, 07:20 PM

Hey B.P,

I haven't had the opportunity as yet.

I have tested an 828 MkIII under FW recently , performance was O.K but it is employing double buffers on playback, which will effect RTL and real time playability.

I suspect the Hybrid interfaces are using a different variation of the ArchWave ( Bridgeco) OEM controllers that have both USB/FW.

If/When I get a chance to get in front of one, I'll post the results as per usual.

Peace

TAFKAT

10-23-2011, 01:45 AM

Hey All,

I haven't had a lot of time for testing lately but I thought it would be good to post the latest batch of results and amendments that I have collated since the last posting.

http://www.dawbench.com/images/dawbench-audioint-22-10-11.jpg

Interfaces Added :

MOTU - 828 Mk3 - FW : Respectable performance across the board, only caveat being that it uses double buffering on playback which effects its RTL.

Presonus Studio Live 16.4.2: Another unit using the now familiar Dice OEM controller/driver solution , performance being on par with other interfaces using the identical controller /driver.

Focusrite Saffire 6 USB : Worst performing and highest I/O and RTL of any interface tested , needless to say this did not rate well. * Testing and results by Scan U.K *

Amended Results:

AVID Mbox Pro 3 , Focusrite Saffire 56 , Midas Venice F - all units were not reporting AD/DA , RTL has been calculated using the utility, RTL and LLP Ratings have been adjusted accordingly.

http://www.gearslutz.com/board/images/smilies/nu/009.gif

TAFKAT

01-08-2012, 05:42 PM

Hey All,

Here's the final list of interfaces tested in 2011 , with some added detail of the new interfaces added/amended to the list and a summary.

http://www.dawbench.com/images/dawbench-llp-01-12-2.jpg

http://www.dawbench.com/images/dawbench-llp-01-12-1.jpg

New Interfaces added.

NI Komplete Audio 6 : Respectable performance , very solid drivers, higher than normal RTL due to added safety buffers. * Testing and results by Scan U.K *

Presonus Audio Box 44VSL : Good I/O , RTL results, 032 latency setting is unusable being more window dressing than anything else , performance close enough to equal to their FW interfaces.

Steinberg UR28M : High I/O and RTL , one of the worst performing interfaces tested so far. It did at least manage a result/rating.

Amended Results.

Echo AudioFire 12 : I went back and retested this interface with the latest 5.7 version driver on the request of a few members reading in ( chart has a typo and still shows driver version 5.5 ) . Performance was identical to 5.5 , but this time I managed to get results using the latest build of the in house RTL utility. There had already been some confusion with the reported I/O and RTL, as they would report differently in Cubase to Reaper/Sonar , so I wanted to get an accurate reading using the RTL utility to get a result independent from the DAW hosts. Heres where the funs starts as the unit reported huge added safety buffers per respective setting which would increase as the settings went up - ie. added 256 samples from 064 to 128 setting, 512 samples 128 to 256 , I couldn't get the final figure as the RTL utility instantly crashed trying to get a reading for the 512 setting.

The inconsistency that this unit displays in reported and actual RTL's is quite disturbing and I am way past spending any more time on it considering the product line is DOA here locally , literally , I had the only working unit from the old distros' and there will be no further units available from that distro at least. I won't be revisiting the Echo FW units.

Pending Results.

Focusrite Scarlett 2i2 : This one is an interesting case , we did manage to get testing and results completed but the respective settings are so far off the standard , - i.e 44/88/132/264/528 , that is was difficult to correlate against the baseline. The upper 3 settings I could let slide, but the 44/88 settings were the tipping point.

Mind you I/O and RTL are thru the roof and have absolutely no correlation to the listed settings - i.e Reported I/O @ 44 - 3.356/4.376 : Calculated RTL 11.1 ms , all the way to Reported I/O @ 528 - 25.351/37.370 : Calculated RTL 66.00 ms , yep, you read that right.

Needless to say this would not not rate well even if we could get a final rating.

M-Audio C400/C600: This unit tested very close to the results for the Fastrack Ultra 8R, but had a strange quirk where the performance dipped at 512. I have reported the issue to the devs and am awaiting resolution before posting final results.

Tested but No Rating Achieved :

Lynx Aurora 8 - FW : Unfortunately this unit proved extremely unstable and unpredictable that testing was abandoned , still hoping to test the unit with the AES16 PCIe card.

EMU 0404 - USB2 : This unit was extremely inconsistent , lower results at higher buffer settings than the previous, vastly different results at respective buffer settings on consecutive runs , etc.

That closes off the testing for the last year , and its been interesting to say the least. A quick summary on my part would be that there are only very few manufacturers who really focus on their drivers to deliver good Low Latency Performance , and its obvious from the charts who they are.

The shift to USB2 that has been gaining momentum is not going to guarantee better performance/compatibility , especially for those manufacturers again relying on 3rd party OEM controllers/drivers. There are some stand outs at both ends of the spectrum listed in the charts , I am not going to go into further detail or commentary as the results speak for themselves.

I will be writing a more detailed report for the main DAWbench website when I get some clear air.

Peace

MattiasNYC

01-08-2012, 08:38 PM

Awesome work Vin!

248

paulwr

01-09-2012, 12:26 AM

Such a valuable reference. Thanks, Vin!

Paul

Beat Productions

01-13-2012, 10:31 AM

Yeah, thanks for all the work, it's greatly appreciated.
First we're spoon fed mp3s, now USB2 interfaces, does it seem like we're going backwards?

TAFKAT

01-15-2012, 05:56 PM

Quick Update.

3 years back I had a rather frustrating experience with a Prism Orpheus that I outlined in a thread at the Cubendo Forum , in short the unit was extremely inconsistent , had a convoluted buffer setting protocol , large safety buffers and mediocre performance , so not exactly impressive in respect to LLP .

With the focus on the 3rd party OEM Firewire controllers being used in some of the current crop of interfaces , this unit has my added interest because it doesn't use the more widely used Dice chip, instead using the BridgeCo/ArchWave OEM controller/driver , which would allow me to do some direct head to head against the 2 available OEM controllers.

This unit is a strange bird IMO - Prism are renown for the extreme high end converters , and this unit of course has the geneology of the higher end product line but is priced substantially lower, but not what anyone would call low mind you at around 5K, and it is an audio interface first and foremost. The full AD/DA potential of this unit is reserved to the audio interface as using it as a stand alone AD/DA it has only 1 x ADAT I/O which locks it to 48K, and no AES/MADI option which would allow it to use its 192 capability.

With the potential of the unit being so dependent on the audio interface component , I would expect that Prism would ensure that the audio interface controller and driver would deliver performance that would be worthy of such a high end professional unit.

Since my last experience which was on XP and Driver/Control version 1.05 , there has been a few upgrades to the current version of 1.07, not exactly a lot in the last 3 years but it was worth investigating and I was lucky enough to have a unit available for me to run it thru its paces again. FWIW - The client has been running his Orpheus units via ADAT the last 3 years connected to an RME HDSPe RAYDAT , as the units were simply unworkable for him as stand alone interfaces.

First off , the units control panel is identical the last time I tested the unit, there are 2 parameters that need to be adjusted to set the latency - Buffer (us) / Latency (ms) , the buffer time setting can be manually adjusted in increments of 250 us which in turn alters the available latency settings as multiples . Standard setting is 1500 us, which would then give a range of latency values from 3 ms to 12 ms. The 2 settings give a lot of flexibility to the available latency values over all other interfaces I have tested, but also an added level of complexity and there will be latency settings that will be available across multiple buffer time settings as well.

Some quick points on the latency values available on the units - the settings are listed in ms ( nominal ) , so I was able to use .75 / 1.5 / 3.0 / 6.0 / 12.0 ms for the standard test values of 032 / 064 / 128 / 256 / 512 , however the actual sample buffers as calculated by the RTL Utility were actually 032 / 066 / 132 / 264 / 528 . I tested both the 064/128 settings across 2 available buffer time settings 250/500 for 064 , 500/1000 for the 128 setting, and found in both instances that the lower buffer time setting for the respective latency values performed substantially better , in the range or 15-30%. I then used only the lowest buffer time for each respective latency value.

I hope that makes sense.

O.K, with that out of the way , how did it perform.

http://www.dawbench.com/images/dawbench-llp-orpheus.jpg

Performance is typical of interfaces using 3rd party OEM controllers , large safety buffers , average to below average performance at lower latencies in regards to Virtual Instruments , but there is an interesting quirk in the RXC results. Performance on that particular test was quite good all the way down to 032 samples , but the performance achieved at 064 samples could not match the 032 result - I reran the tests 3 times at each respective buffer to be 100% sure. Add to that if we factored in the 064 result using the larger buffer time which dropped the number of RXC's 30% to 89 , and the 032 results stands as even more of an oddity.

So in closing , the experience this time around was definitely more consistent than my first run on the unit , but the overall performance of the unit as an audio interface is average at best , and with no option to use the full potential of the AD/DA sans the interface, those looking at the unit as a primary interface will need to take that into account.

Of course YMMV.

Now stating the obvious before I get a wave of irate Orpheus users on here taking issue, this report is purely on the Low Latency Performance of the FW Controller/Driver , it is not in reference to the feature set , quality of the AD/DA , so please use the numerous other threads to continue that debate.

P.S : I have updated the previous charts from last week to now include the Orpheus

Peace

TAFKAT

01-16-2012, 11:11 PM

Quick heads up,

The Round Trip Latency measuring utility that I have helped develop is now available for public download.

RTL Utility by Oblique Audio : Here (http://www.oblique-audio.com/free/rtlutility)

Animus

01-16-2012, 11:41 PM

ROCKINROG

01-27-2012, 06:09 PM

Cool. Thanks Vin.

Beat Productions

02-02-2012, 11:41 AM

TAFKAT

02-13-2012, 05:06 PM

Vin:
Again, thanks SO MUCH for all your testing!
I'm thinking of getting A M-Audio Profire 2626, do you know if it uses the same drivers as the Profire 610 you tested? The 610 seems to have one of the better latencies of the firewire units you've tested.

Hey B,

Yes the 2626 shares the same driver and performance to the 610 .

It was one of the better units , it actually uses the Dice controller that a lot of the others are using, difference being AVID/M-Audio develop their own drivers instead of using the bundled OEM.

Having said that be careful with 2626 units on laptops as they are a lot more finicky than even the 610 , which has a smaller version of the controller that is more compatible with most FW chipsets , the 2626 will not woprk on my current Audio notebooks for example, whereas the 610 works prefectly.

TAFKAT

02-13-2012, 05:06 PM

Quick Update,

I managed to finally run a Lynx AES16e / Aurora combo thru its paces.

The earlier Lynx 2 card performed very well so I was interested in getting the latest PCIe offering on the bench to see if the performance could be maintained , especially with the ultra tight AD/DA latency of the Aurora.

http://www.dawbench.com/images/dawbench-llp-lynx-aes.jpg

Overall performance of the PCIe unit did not match the older Lynx 2 card unfortunately , I also experienced some very odd loss of buffer sync issues at 256 samples with the DAWbench VI test sessions which resulted in random garbled playback and metallic like artefacts on release tails, etc. This was more often than not triggered by loading a previous maxed out session from the 128 buffer size and trying to incrementally continue to load the session. I had to to start the 256 session from scratch each time. Where this could be a problem in Real World is if you had a heavily loaded session and wanted to bump up a buffer size to get some more headroom, you would in fact not be able to play the previous session without artefacts.

Overall I/O and RTL latency is actually lower than the RME AES/ADI8 reference unit , so potentially could have tipped the base reference interface if it had performed equally to the older Lynx 2 card, but unfortunately it didn't measure up to its older sibling. The results look quite good, but there are a few quirks that left me with the feeling that its not as solid as it could be.

I am currently in contact with Lynx support and am working thru a few angles.

I'll give you guys a heads up if anything fruitful eventuates.

* I have added the Lynx to the most recent database list *

Peace

Beat Productions

02-14-2012, 10:37 AM

Hey B,

Yes the 2626 shares the same driver and performance to the 610 .

It was one of the better units , it actually uses the Dice controller that a lot of the others are using, difference being AVID/M-Audio develop their own drivers instead of using the bundled OEM.

Having said that be careful with 2626 units on laptops as they are a lot more finicky than even the 610 , which has a smaller version of the controller that is more compatible with most FW chipsets , the 2626 will not woprk on my current Audio notebooks for example, whereas the 610 works prefectly.

Tafkat:
Thanks for this, my laptop firewire has a TI chipset (HP nc6320) so I should be OK.
I've been experimenting with using Cubase as a live mixing (and recording) tool so I need a low latency device. I'm using a MOTU 424 (via a 1 unit Magma PCMCIA), 2 2408 MK2s and 5 ADA8000s at the moment at a buffer of 128 and it works quite well. I would like to have a firewire device for smaller (24 channel) gigs and also for backup and ease of setup. I had a Focusrite Pro40 and, although I really liked it, the latency wasn't the best so I'm thinking the Profire 2626 would be the best choice as it can do 26 channels.