Macs, interfaces, and latency

[AnotherScott]

PIccking up from an old thread at

 

https://forums.musicplayer.com/ubbthreads.php/topics/2445263/1

 

(which I did not revive here because of what would have been the red herring of the subject line)

 

there was a lot of discussion there about the benefits of using an interface on a Mac (and how that compared to the Windows environment).

 

Something that seemed to get a little confused there is whether some comments were referring to adding a MIDI interface or adding an Audio interface (and of course, some interfaces perform both functions).

 

I am finally getting around to doing some experimentation of my own here, on a Macbook Air (1.6 gHz i5, 4 mb RAM, and of course, SSD).

 

First, to be clear, I always need to use a MIDI interface, as opposed to connecting my keyboard vis USB, because the board I am using is standard MIDI and not USB. So that's a given here.

 

The question, then, is about the need/desirability of using an Audio interface (as opposed to the headphone out). Obviously, if it's a long run, the headphone jack is not sufficient, but compared to traditional keyboard setups, that's not a limitation of the headphone output per se... even if I was running a line out directly from a keyboard, unless it had a balanced out, I would need a direct box regardless.

 

My biggest question here is about latency. Ashville Guru said that an interface greatly reduces latency, and in thinking about it, I'm not certain he was talking about that being a benefit of an audio interface, or merely one of a MIDI interface, which as I said, I would use regardless.

 

I have connected my keyboard with a little Edirol 1x1 MIDI interface (basically, little more than a cable with an in-line piece the size of a fancy cigarette lighter). I have plugged headphones into the output of the Air. I am triggering the software from an older Casio PX piano that has its own speakers. I have adjusted the settings on the Mac to the point where I sense no latency, i.e. no "echo" between the immediate-feeling native keyboard piano sound I hear coming from the Casio's speaker and the sound of the simultaneously triggered computer-genereated piano sound coming through my headphones. So at least so far, since I have no latency issue or hums/buzzes, I see no reason to consider an audio interface.

 

Besides passing along my experience there, I wanted to ask this question to clarify my understanding: Does an audio (not MIDI) interface have the potential to improve latency? It wouldn't matter to me now, as I'm effectively experiencing no noticeable latency anyway, but this "low" setting on the computer for no perceptible latency does increase CPU usage, so I imagine latency could become an issue if I make more significant demands on the system. I have not yet attempted to task its polyphony, or trigger more than one sound at a time, or load any plug-in that does more that straight sample playback. If I do run into issues in attempting to do any of those things, I'm not sure whether it can be addressed via an audio interface or a more sophisticated MIDI interface (or both), or, for that matter, if either of those things are nearly as beneficial as simply using a more capable computer.

 

For example, if I do run into a situation where I want to improve the latency performance of my system, which of these solutions provides more improvement?

 

--- adding an audio and/or better MIDI interface to the 1.6 gHz i5, 4 mb Air I'm using, or

 

--- upgrading to a new Air, which can have a 2.0 gHz i7 and 8 gb RAM (a lot more expensive at $1350, but I could get a bunch of it back my selling my old Air, and have the benefit of a faster computer for everything else I do, so this may not be unreasonable)

 

(I realize a new Mini, which can have a 2.6 gHz i7 and 16 gb RAM at $1500, would be even better than that improved Air, but I can't subsidize the cost of that with the sale of my existing Air, as I still need a notebook, so that's a bit more difficult to rationalize. That's apart from the separate discussion about the relative suitability of a notebook vs. a Mini vs. a notebook-in-a-rack for gigging, in terms of ergonomics, aesthetics, accessibility, imperviousness to drops, convenience in setup and in transport, safety from theft, etc., that has been somewhat discussed elsewhere.)

 

[burningbusch]

With Presonus Studio One you can see input and output latency. Here are what the following devices registered with a setting of DEVICE BLOCK SIZE (buffer) of 128 samples (64-bit):

[font:Courier New]

Digidesign TDM (PCI): In 5.8 ms Out 2.9 ms

Built-in Audio: In 5.8 ms Out 4.4 ms

Konnekt 24D (FW): In 8.8ms Out 5.9 ms

Venom Synth (USB): In 10.93 ms Out 8.03 ms

Kronos Synth (USB): In 7.68 ms Out 4.68 ms

Edirol M16DX Mixer (USB): In 14.90 ms Out 6.98 ms[/font]

 

The Digidesign TDM card is the only PCI card in the mix which is undoubtedly why it does so well. I suspect the main reason the Edirol does so poorly is that it provides a lot more I/O (18 in/2 out) although the Digidesign interface provides 8 in/8 out and the TC Electronic Konnekt 24D also has numerous I/O. I believe that the main concern when running MIDI in and audio out is the output latency. FW doesn't do any better than USB. Given my interfaces, in a portable setting the Kronos would be my best bet.

 

Companies like RME and Apogee claim low latency. I don't have those available.

 

Busch.

 

 

[burningbusch]

You also might look at something like this:

 

http://www.praisetracks.com/files/images/LSP-1.jpg

 

or the stereo version (currently for the same price: $80)

 

http://www.dsan.com/sc/images/LSP2LG.jpg

 

Info here: http://www.dsan.com/sc/shop/item.asp?itemid=76

 

Review here: http://www.praisetracks.com/tech/permalink/dsan-lsp1-laptop-soundport/

 

Busch.

[AnotherScott]

Very cool boxes, busch, thanks. Looks like a nice piece to use with an iPad as well.

[AnotherScott]

[font:Courier New]

Digidesign TDM (PCI): In 5.8 ms Out 2.9 ms

Built-in Audio: In 5.8 ms Out 4.4 ms

Konnekt 24D (FW): In 8.8ms Out 5.9 ms

Venom Synth (USB): In 10.93 ms Out 8.03 ms

Kronos Synth (USB): In 7.68 ms Out 4.68 ms

Edirol M16DX Mixer (USB): In 14.90 ms Out 6.98 ms[/font]

...

I believe that the main concern when running MIDI in and audio out is the output latency.

Elsewhere, you mentioned how resampling some of your favorite VST sounds into your Kronos allowed you to play them with less latency. With Built-In Audio's Out latency at just 4.4 ms, I'm surprised it would be noticeable. Or is that not the appropriate figure here for some reason? And is there a way to measure how much faster the Kronos is? (I'm assuming that the Kronos figures above refer to using it in a VST environment, that those figures are not applicable to its use in generating its own sounds internally... otherwise the Kronos would appear to actually be slower than the computer based VST!)

[mate stubb]

I believe he was able to speed up the samples in Kronos because he could trim some blank audio from the front of the samples, not because of latency per se.

[ADino]

I wanted to ask this question to clarify my understanding: Does an audio (not MIDI) interface have the potential to improve latency?

 

Scott, I use an RME FF800 on my mac setup. Def a bit expensive as compared to the cool little boxes above. The only time Ive ever noticed any latency at all is when I have multiple large plugs open AND am up to around 25 tracks down. Im talking EW, Ivory, Symphobia etc. Cant say anything about usb, as Ive never used it, and probably never will. Ive used usb in sessions in other studios and it seems ok, but using the rme, latency has never been an issue.

[burningbusch]

Elsewhere, you mentioned how resampling some of your favorite VST sounds into your Kronos allowed you to play them with less latency. With Built-In Audio's Out latency at just 4.4 ms, I'm surprised it would be noticeable. Or is that not the appropriate figure here for some reason? And is there a way to measure how much faster the Kronos is? (I'm assuming that the Kronos figures above refer to using it in a VST environment, that those figures are not applicable to its use in generating its own sounds internally... otherwise the Kronos would appear to actually be slower than the computer based VST!)

 

Output latency is a part of the overall roundtrip but likely a significant part. To measure the difference between the Kronos and a computer-based synth you would need to send MIDI out the keyboard to a soft synth, send the output of soft synth to an analog out and and re-record that output back into the DAW. You would be comparing this to the key-pressed audio out of the Kronos. I don't use the built-in audio on my Mac Pro.

 

Also note that 128 buffer size is possible sometimes but not always by any means. I have some soft synths that require higher buffer settings and this is on a Mac Pro 2.8 Ghz 8 core. Also some projects require larger buffer settings.

 

Regardless, when playing the Kronos and then switching to a software-based piano the latency is apparent to me.

 

With the iPad I found the following latencies:

 

With Camera connection Kit (USB from Kronos; analog out iPad)

iPad Piano (ultra low latency setting) +5.89 ms

GarageBand Piano: +7.2 ms

 

Alesis Dock (MIDI from Kronos to MIDI in on Dock, analog out of dock)

iPad Piano (ultra latency setting) : +14.85ms

Garageband Piano: +16.93ms

 

Busch.

[Ashville.Guru]

My biggest question here is about latency. Ashville Guru said that an interface greatly reduces latency, and in thinking about it, I'm not certain he was talking about that being a benefit of an audio interface, or merely one of a MIDI interface

Audio, and audio only. Reducing audio latency in a computer system poses a real technical challenge, unlike midi latency.

 

Does an audio (not MIDI) interface have the potential to improve latency?

Unequivocal YES. Notice it or not, latency can be objectively and scientifically measured. And the measurements have been made for various interfaces (DawBench links in the other thread).

 

The results show HUGE variation among dedicated, non-consumer interfaces. And most of these will certainly perform better than the consumer-grade built-in interface in your Mac. Ergo, a good interface will certainly decrease latency.

 

It wouldn't matter to me now, as I'm effectively experiencing no noticeable latency anyway, but this "low" setting on the computer for no perceptible latency does increase CPU usage, so I imagine latency could become an issue if I make more significant demands on the system.

That's not how it works. Latency for a given buffer setting is fixed, more or less. Tasking the audio load will not cause greater latency - it would simply cause clicks and pops (dropouts). To remove the dropouts, you'd be forced to use a higher buffer setting, in turn causing higher latency.

 

Again, a good interface will not only (i) reduce latency at a given buffer setting, but also (ii) allow you to play at a lower buffer setting without pops and clicks.

 

OTOH, CPU power can only affect safe buffer settings, not latency. So for a given interface (builtin or otherwise), more powerful CPU-> safe to play at lower buffer settings -> lower latency.

 

For example, if I do run into a situation where I want to improve the latency performance of my system, which of these solutions provides more improvement?

 

--- adding an audio and/or better MIDI interface to the 1.6 gHz i5, 4 mb Air I'm using, or

 

--- upgrading to a new Air, which can have a 2.0 gHz i7 and 8 gb RAM (a lot more expensive at $1350, but I could get a bunch of it back my selling my old Air, and have the benefit of a faster computer for everything else I do, so this may not be unreasonable)

Both choices will reduce latency. Ideally, you should get the latest and greatest i7 quad-core processor, and use a kick-ass interface like the RME Fireface.... .

 

I don't think there's any reliable source of info out there that can tell you exactly which option will yield lower latency. Also - there are interfaces, and there are interfaces. There are i7 processors, and there are i7 processors. Good and bad. The specifics matter.

 

But does all this matter, when you can't even notice the latency? Let's hope that it doesn't. One thing's for sure: some of us notice latency better. Kanker says he can notice a millisecond. Others can only notice the improvement when they move to a lower-latency system.

 

Here's a scary thought - nobody knows if latency can subtly throw your timing off, without you even noticing it...!

Think about it - in the history of musical instruments, latency is a very recent artefact. It hasn't been around long enough for either the science or the anecdotal wisdom to mature.

 

Happy new year, everyone...!

 

 

[AnotherScott]

It wouldn't matter to me now, as I'm effectively experiencing no noticeable latency anyway, but this "low" setting on the computer for no perceptible latency does increase CPU usage, so I imagine latency could become an issue if I make more significant demands on the system.

That's not how it works. Latency for a given buffer setting is fixed, more or less. Tasking the audio load will not cause greater latency - it would simply cause clicks and pops (dropouts). To remove the dropouts, you'd be forced to use a higher buffer setting, in turn causing higher latency.

Yes, I think we're both saying the same thing... that if I make more demands on the system, I could conceivably have to adjust the buffer setting upwards, leading to more latency. Maybe my reference to "CPU usage" was a misnomer though... I thought that the low buffer (low latency) setting increases demands on the CPU (which, in turns, is what leads to the clicks and pops if your demands exceed its capability), but maybe that's not an accurate way to put it? Regardless, we're in agreement... I may be experiencing no noticeable latency now, but if I run more demanding software, I may have to increase the buffer size and therefore the latency, so it is a potential issue in the future.

 

Again, a good interface will not only (i) reduce latency at a given buffer setting, but also (ii) allow you to play at a lower buffer setting without pops and clicks.

Other than the Air's built-in audio jack, I also do have an mBox 2. I don't know how state of the art that is these days, but I guess if latency becomes an issue for me, I would try that next.

 

Here's a scary thought - nobody knows if latency can subtly throw your timing off, without you even noticing it...!

Think about it - in the history of musical instruments, latency is a very recent artefact. It hasn't been around long enough for either the science or the anecdotal wisdom to mature.

My biggest issue, musically, is that if I'm not focussed when playing a gig, I have a tendency to push the tempo. So who knows, maybe a little latency will do me good. ;-)

[Reezekeys]

Does an audio (not MIDI) interface have the potential to improve latency?

Unequivocal YES. Notice it or not, latency can be objectively and scientifically measured. And the measurements have been made for various interfaces (DawBench links in the other thread).

I must be misunderstanding something because the dawbench links you reference seem to be comparing OSX vs Windows using an RME audio interface. It does not seem to be comparing different audio interfaces or external vs. built-in sound. Maybe I saw the wrong link?

 

I am curious as to the reasons why, on OSX, an external interface can get better latency than the internal sound. I think I read the entire thread but unless I missed something, you never said why this is. Not disputing you just curious!

[David Loving]

What software sounds are you using? What adjustments did you make on the Mac Air regarding latency? I use a Presonius Audiobox and am running Kontakt on a Macbook Pro. Kontakt has an option screen where you can set the buffer anywhere you want. I usually set it at 128. It defaults to 512. I have never tried to adjust latency in the Macbook, and have never used the headphone out. I use USB headphones (the kind you use with Rosetta Stone language lessons) to play with the sounds in Kontakt. Plug into a usb port. I didn't know I had a headphone out jack until I looked at the Mac after reading your post! I am curious about what you did in the McAir to adjust latency. I have used a PX330's input jacks to monitor the software - that is a good feature. I don't use any boards that are non-USB. I have an old Kurz PC1x that I have used midi-USB connectors, but don't use it with the computer. Latency never was an issue with Kontakt because of the ability to adjust the buffer. I use the Audiobox on gigs.

[Ashville.Guru]

I must be misunderstanding something because the dawbench links you reference seem to be comparing OSX vs Windows using an RME audio interface. It does not seem to be comparing different audio interfaces or external vs. built-in sound. Maybe I saw the wrong link?

 

Yup. There were two links in the thread, and I was a bit too lazy to be more specific, so here it is: Interface latency comparison.

 

I am curious as to the reasons why, on OSX, an external interface can get better latency than the internal sound. I think I read the entire thread but unless I missed something, you never said why this is. Not disputing you just curious!

As the numbers show, interfaces vary w.r.t. latency performance. Bear in mind that these are professional audio devices, designed to reduce latency. Internal sound cards on laptops are consumer-grade devices, designed to allow the average laptop user to listen to MP3s and watch movies, while still being slim enough to be mounted inside a laptop chassis. One involves a design compromise, the other doesn't.

 

The take-home message from the measurements is simply this: building a low-latency interface is no simple engineering feat. Even the flagship products of reputed pro-audio companies struggle to get it right.

 

Now consider the fact that built-in soundcards are el-cheapo generic devices costing about a buck to produce. For clinching evidence in all gory detail, I defer to BurningBusch's excellent post in the other thread. The rest should be obvious.

 

 

Side rant: @Reezekeys, why the "on OSX" in your question? And for that matter, @AnotherScott, why the "Macs" in the thread title? Neither the question nor the ensuing discussion is specific to Macs. This is about computers and interfaces, plain and simple. There's nothing that warrants one particular computer manufacturer's products to be in a separate category. In the current context, at least...

[AnotherScott]

@AnotherScott, why the "Macs" in the thread title? Neither the question nor the ensuing discussion is specific to Macs.

My question was focussed on the Mac environment. I didn't want to get into discussions of any I/O variables that had no relevance on the Mac, like ASIO, etc.

[David Loving]

1. OP said he uses a Mac and wondered about the best way to handle latency

2. OP can't control the responses

[Reezekeys]

I must be misunderstanding something because the dawbench links you reference seem to be comparing OSX vs Windows using an RME audio interface. It does not seem to be comparing different audio interfaces or external vs. built-in sound. Maybe I saw the wrong link?

 

Yup. There were two links in the thread, and I was a bit too lazy to be more specific, so here it is: Interface latency comparison.

Thanks, I had a feeling I missed something. However, reading this, I note that it's Windows-only, and it also compares only external interfaces. No mention of built-in soundcards.

 

I am curious as to the reasons why, on OSX, an external interface can get better latency than the internal sound. I think I read the entire thread but unless I missed something, you never said why this is. Not disputing you just curious!

As the numbers show, interfaces vary w.r.t. latency performance. Bear in mind that these are professional audio devices, designed to reduce latency. Internal sound cards on laptops are consumer-grade devices, designed to allow the average laptop user to listen to MP3s and watch movies, while still being slim enough to be mounted inside a laptop chassis. One involves a design compromise, the other doesn't.

That does sound reasonable, but again, none of the links you supplied contain information specific to LLP of built-in vs. external interfaces. Accepting that a built-in soundcard would have worse LLP, I would be very interested to know how much worse the performance is. AFIK there are no actual numbers on this. I say this as a very happy camper with my current rig an almost 5-year-old MacBook Pro (2.2Ghz Core 2 Duo, the lower-end configuration at the time). My setup consists of seven VIs, some of them up to 5-part multitimbral, reverb, efx & eqs, hosted in Bidule, with lots of processing involved in routing/mixing. All running at a 128 sample buffer setting, on one core! Now, I'm not playing all seven VIs simultaneously of course but I've done 5 with no issues. That's pretty impressive performance for a 5-year-old laptop with half a processor IMO! I take the headphone out direct to my QSC K8s and 95% of the time it's all good. There are issues relating to grounding and my laptop's power adapter which account for the other 5% but that's a common occurrence with switching power supplies and maybe another thread.

 

This is why I am puzzled when I read about people having issues or questions about getting low latency on laptops that are way newer & more powerful than mine. Especially given that Windows XP and 7 is a better OS for low latency performance than OSX (according to the dawbench links).

 

Now consider the fact that built-in soundcards are el-cheapo generic devices costing about a buck to produce. For clinching evidence in all gory detail, I defer to BurningBusch's excellent post in the other thread. The rest should be obvious.

I believe that's about audio quality, not latency per se. The MacBook Pro actually fares pretty well according to that link. On some of my gigs when I need real loud volume and headroom, I switch my QSCs' input to mic level. Then, I hear more hiss, which is the analog electronics. Of course there's also some "buzzing" from the motherboard's digital machinations leaking into the audio. But I'm also backing down on the QSC's input level a bit, and by that point the band is so loud that you don't hear any of this! On 90% of my gigs I leave the QSCs set to line-level input and their input gain at max. As long as there are no issues with ground loops, it sounds fantastic, IMHO of course.

 

When I record for CD projects, I take the headphone out into a DI box, direct to DAW. I don't use these tracks for the final mix because I like to edit as midi I re-render to audio when I'm done. But when I compare my final tracks to the original scratch track recorded from the headphone out, in terms of hiss & hum, they sound pretty much identical to me.

 

I think the point I'm trying to make is that, at least on a Mac, one should not automatically assume that an external interface is needed for high-quality audio, or "low enough" latency. Of course it depends on the context, and I know some people are very sensitive to latency. Also and this point may not have been made before using a computer only for playing VIs is a "one-way" deal; the dawbench figures are all about round-trip latency. Playing VIs does not involve audio going through an input buffer. This does not mean that the round-trip numbers in those charts can be halved, but I'm guessing that the actual latency would be somewhat lower.

 

Side rant: @Reezekeys, why the "on OSX" in your question? And for that matter, @AnotherScott, why the "Macs" in the thread title? Neither the question nor the ensuing discussion is specific to Macs. This is about computers and interfaces, plain and simple. There's nothing that warrants one particular computer manufacturer's products to be in a separate category. In the current context, at least...

Sorry, I didn't mean to imply that these issues are Mac-specific. The original question had to do with audio on a Mac Air.

[richwhite9]

An off-topic aside for those who follow this stuff: Ubuntu for Android formally announced to be shipping for the next Holiday Season (Galaxy S IV phone is the target of course). Reaper works on Ubuntu. Nicely. Still some more work to be done on that Android kernel but....

 

http://www.engadget.com/2013/01/02/ubuntu-for-smartphones/

 

The next leg up is coming with a Wintel Surface Pro Phone (Atom-based). I like to think my Nokia Lumia Fall 2014 Version will replace my Receptor Rev C and have better latency

 

I didn't want to start a new thread for this little aside but we live in interesting times!

 

[Ashville.Guru]

... it also compares only external interfaces. No mention of built-in soundcards.

 

...

That does sound reasonable, but again, none of the links you supplied contain information specific to LLP of built-in vs. external interfaces. Accepting that a built-in soundcard would have worse LLP, I would be very interested to know how much worse the performance is. AFIK there are no actual numbers on this.

Aha - I was wondering how soon someone would bring this up... .

 

Yes, the DawBench studies do not involve built-in soundcards. Thankfully, when it comes to latency, we are not limited to dissecting info on the web, if we're really passionate about settling the issue.

 

Sometime ago, I had the same question, and did a few measurements myself - check out this post in another thread.

 

My humble TC Desktop Konnekt 6 has about half the latency of my laptop built-in soundcard, at 64 buffers. And I can use a much larger load at 64 with the TC than with the builtin soundcard.

 

Hence my confident assertion that a good external interface (at least as good as my TC Konnekt) will improve latency performance.

 

But I'm sure one can find many nits to pick with this reasoning:

- "yours is a Windows system, everything might be different on a Mac"

- "maybe RTL is statistically uncorrelated with output latency"

- "maybe your Dell Latitude has an uber-crappy soundcard, and maybe a Mac Air soundcard trumps even the RMEs and the Lynx Auroras"

... etc, etc.

 

Yup, I don't have the numbers to dispute such claims. But its ridiculously simple to do the measurements yourself...! All you need is a keyboard with builtin sounds similar to your softsynth (APs for example, or ideally pure sine wave tone), and an external recorder OTHER than your laptop. Set it up so that the one stereo channel is from the softsynth, the other is from the builtin sound of your keyboard. Play a note, record. Use any DAW or Audacity to measure the difference in onset times - presto, now you have the latency differencial with hardware.

 

You can use this to see how much of a difference an external interface would make. I'm betting AnotherScott would see an improvement even with his ancient MBox2 (though USB1.1 makes it a really long shot...!).

 

I've posted this suggestion a few times already, but only BurningBusch has been enthusiastic enough to actually do the measurements! I guess it's easier and more fun to speculate, and dissect info on the net, than actually roll up your sleeves and get to the bottom of things. Especially when you have the luxury of online stores with return policies.

 

This is why I am puzzled when I read about people having issues or questions about getting low latency on laptops that are way newer & more powerful than mine. Especially given that Windows XP and 7 is a better OS for low latency performance than OSX (according to the dawbench links).

 

This is where objective measurements cease to be useful. It's hard to explain, since involves ones personal standards of professionalism, and more importantly, intense passion for one's musical instrument.

 

Can you make good music using the headphone outs of your laptop? Of course you can. Can you make great music using a Roland KC amp as a monitor, instead of a QSC K10? Of course you can...

 

... but try telling that to Mate_Stubb!

 

Yes, the analogy isn't quite perfect; almost anyone can notice the difference between the sound of a KC amp and a QSC - perhaps not so with 6 ms vs 12 ms latency. So what? There are lots of anecdotes on this forum, of people being happy with KC amps until they used a QSC. Same way, BurningBusch was happy using a laptop live, until he played the Kronos. There are other anecdotal posts where players were perfectly happy with their laptop setup, until they moved to a less-latent system.

 

Notice it or not, be bothered by it or not, the fact remains that digital latency is an artefact. And whether I notice it or not, I don't want it anywhere near my musical instrument. My goal is to reduce it as much as possible my budget allows it...! YMMV.

 

- Guru

[Theo Verelst]

When thinking about latency between pressing a key and actually hearing a sound, it can pay to aim for *constant* latency, coming from the keyboard scanning, the midi processing, the OS and music program response times, the DA converter driving and built in filters, and the perceived audio path length between you and your monitors and PA sound sources. A lot ofcorners are being cut in all of these departments, and various tricks are being used all over the place, which often do not honor the musical playing experience of the normal kind.

 

Playing a chord on the piano works good because a musician can easily grasp thesound that comes from his or her playing, even if the tones, resonances and sound waves coming towards him have a complex nature. Some "smart" programmer favoring certain notes being kicked in an output buffer ahead of others, destroying the constant delay feel is probably not a good idea at all.

[zephonic]

Aha - I was wondering how soon someone would bring this up... .

 

quote abbreviated, but I'm responding to all of it

 

AG, I'm the guy who posted the links to the DAWbench back then, and you seem to have to have taken off with them. I posted the links because I believe they present good information, things people should know before they buy a computer and/or audio interface/DAW.

 

However, I do not take those benchmarks (or any, FTM) as the absolute truth, as you (and many others over at GS) seem to do. I think they are excellent indicators of what to expect, but don't forget it's the test results of one guy on one particular system.

 

As for internal soundcards, in my experience there is a large difference between OSX and Windows. I was surprised to learn that my old iMac's internal soundcard peformed nearly as well as the external Saffire LE I used to have. You won't squeeze every last inch of performance out of it, but it is certainly possible to use an iMac's or Macbook's internal soundcard at medium latencies (say 128 or 256) and get the job done.

 

Now, for the most part I agree about the latency thing, but to put things in perspective:

 

http://www.kenfoster.com/Articles/anaction.gif

 

How much latency do you think that is? Exactly, yet it doesn't stop pianists from making great music with it.

 

When A/B-ing, I notice a very pronounced difference between playing my Nord Piano and any VI pianos. The lag with the VI's is immediately apparent. When not A/B-ing I just play them and forget about lag.

 

As an aside, I remain unimpressed by Cubase's MIDI timing. It has gotten much better, but my Fantom's internal sequencer feels much tighter and spunkier, as does Logic. Bouncing MIDI to audio shows timing inconsistencies across the board, but that doesn't stop me from getting on with it.

 

Finally, I think Theo makes the good point that if the delay is at least consistent, musicians will be able to work with it. Shucks, try a pipe organ in a church and you'll know what I mean.

 

[Reezekeys]

Yup, I don't have the numbers to dispute such claims. But its ridiculously simple to do the measurements yourself...! All you need is a keyboard with builtin sounds similar to your softsynth (APs for example, or ideally pure sine wave tone), and an external recorder OTHER than your laptop. Set it up so that the one stereo channel is from the softsynth, the other is from the builtin sound of your keyboard. Play a note, record. Use any DAW or Audacity to measure the difference in onset times - presto, now you have the latency differencial with hardware.

After revisiting the earlier threads and digesting everything, I'm ready to check this out but your methodology won't work for me since I don't use a synth with built-in sounds it's a Roland midi controller. One possible way to do it would be to use its midi output wired to an audio plug; that would produce a "click" which would be something I could record and compare onset time with a percussive sound played on a VI in my laptop. Of course there is some latency between the keypress and the midi output, and I don't know whether the midi-via-USB's latency is better, worse, or the same as what's going out the 5-pin DIN port.

 

And just to make things more fun, the only external interface I can compare my headphone output to right now is an ancient Emagic 2|6! I'm actually confused because I think the driver is PPC-only code so it should not work well at all real-time audio software running in Rosetta emulation is usually a recipe for crappy performance (remember the Spectrasonic "wrappers" for Trilogy & Atmosphere?) but the few times I hooked the Emagic up to my regular Bidule layout it worked quite well IIRC.

[Ashville.Guru]

AG, I'm the guy who posted the links to the DAWbench back then, and you seem to have to have taken off with them. I posted the links because I believe they present good information, things people should know before they buy a computer and/or audio interface/DAW.

Zeph - I'm grateful to you for pointing me in the direction of the DawBench studies (I've also mentioned it somewhere). I've been following the discussions on GS. Also, I really liked your ' Sum of Parts' blog post, particularly the 'soothing and calming work environment' analogy.

 

However, I do not take those benchmarks (or any, FTM) as the absolute truth, as you (and many others over at GS) seem to do.

Can you point out exactly why you think I take DawBench as the "absolute truth"? Of course, the whole thing is one guy's results on one system. I'd appreciate it if you could point out some specific unreasonable conclusion I've jumped to from the DawBench numbers.

 

To clarify, in the context of this thread, this is all that I conclude from that study - there's a huge variation among external interfaces, when it comes to latency performance. Nothing more, nothing less.

 

How does this connect to the OP's question? Based on my measurements, and Busch's specs, it's extremely likely that the latency performance of a generic builtin soundcard is not much better than a bottom-quartile external interface. Ergo, there are external interfaces that ought to provide a measurable decrease in latency.

 

Besides, I've provided a simple test by which I can be proved wrong. Heck - I'm willing to stick my neck out and make a specific prediction - anything on this chart with DawBench rating 7 or above will beat the internal soundcard of a Mac Air. Trust me, I'd actually be happy if someone proved that I don't need to invest in an RME Fireface...! :grin:

 

Now, who's ready to put their money where their mouth is...?

 

As for internal soundcards, in my experience there is a large difference between OSX and Windows.

Can you elaborate on how exactly you measured the difference between OSX and Windows? 'Nearly as well' can mean anything; I've posted the specific numbers for my Dell builtin Vs. TC Konnekt. For example, how different were your measurements from the numbers I've posted? Of course, there's also the 'load processing power/buffer' part which I haven't posted.

 

I was surprised to learn that my old iMac's internal soundcard peformed nearly as well as the external Saffire LE I used to have.

I can't say I find it too surprising that any builtin card has a latency performance nearly as good as that of a bottom-quartile external interface.

 

Regarding the rest of your post, while I appreciate the point you're making, I don't want the subjective side of the debate to take over - for now! Maybe that deserves a separate thread in itself. The OP has asked a question that can certainly be objectively and unequivocally settled, without debate (unlike DP/clone sound/action etc!). At least let's provide him with some numbers - "like external interface 'X' will give you 'Y' ms less latency, and 'Z% more load-processing power' over your Mac Air's builtin". He already knows that he won't notice the difference, so what's the point of debating?

 

- Guru

[richwhite9]

This is actually a good Keyboard article IMHO.

 

Are i7 Intel Boards sufficient? I won't be satisfied until we see outboard Thunderbolt and USB 3.0 units for state-of-the-art comparisons

[zephonic]

Also, I really liked your ' Sum of Parts' blog post, particularly the 'soothing and calming work environment' analogy.

 

Thanks for reading my blog!

 

Can you point out exactly why you think I take DawBench as the "absolute truth"? Of course, the whole thing is one guy's results on one system. I'd appreciate it if you could point out some specific unreasonable conclusion I've jumped to from the DawBench numbers.

 

You did not, guess I was the one jumping to conclusions. My bad.

 

Can you elaborate on how exactly you measured the difference between OSX and Windows? 'Nearly as well' can mean anything; I've posted the specific numbers for my Dell builtin Vs. TC Konnekt. For example, how different were your measurements from the numbers I've posted? Of course, there's also the 'load processing power/buffer' part which I haven't posted.

 

I was surprised to learn that my old iMac's internal soundcard peformed nearly as well as the external Saffire LE I used to have.

I can't say I find it too surprising that any builtin card has a latency performance nearly as good as that of a bottom-quartile external interface.

 

I don't have that setup ready anymore, so I can't provide numbers at this time. From memory, compared to the Saffire LE my '07 iMac's internal soundcard at 256 samples added between 2 and 3 ms on both input and output, so probably double that for total RTL. As for VST load, most projects started with the Saffire LE could be run with the internal sound card.

 

I had WinXP bootcamped on the same machine, and while the Saffire LE had only slightly higher RTL's, the internal sound card could not handle anything near what it could on OSX. My guess is that's where having a system-wide pro-grade protocol like CoreAudio really pays off.

 

My point is the Mac's internal sound may well suffice for some people, especially if they use Mac-specific software. And I would recommend a low-budget interface primarily for the connectivity, not so much for LLP.

 

 

[Ashville.Guru]

All good, Zeph!

 

I had WinXP bootcamped on the same machine, and while the Saffire LE had only slightly higher RTL's, the internal sound card could not handle anything near what it could on OSX.

This is new to me. Would be nice if someone with a similar setup could provide actual numbers. Not doubting your experience, just trying to fit it into the larger scheme of things. Personally, I'd need to see the numbers to be convinced. Numbers aren't everything, for sure, but they are the first step towards a systematic validation of the truth.

 

My point is the Mac's internal sound may well suffice for some people, especially if they use Mac-specific software.

Fixed! Agreed, it's pretty well-established at this point that internal sound cards may well suffice for some people - Mac/PC, shouldn't matter either way (no judgement here whatsoever - you make good music with your setup, then more power to you).

 

But again, given that the OP already knows that, the only sensible reason for this thread to exist is to ask what's beyond the "may well suffice" point!

 

For those of us in the 'may well suffice' group, there's no point of further discussion. For those who'd like to look beyond, threads like this are invaluable and necessary - so long as there are more real-world measurements and less empty discussion!

 

- Guru

[zephonic]

All good, Zeph!

 

I had WinXP bootcamped on the same machine, and while the Saffire LE had only slightly higher RTL's, the internal sound card could not handle anything near what it could on OSX.

This is new to me. Would be nice if someone with a similar setup could provide actual numbers. Not doubting your experience, just trying to fit it into the larger scheme of things. Personally, I'd need to see the numbers to be convinced. Numbers aren't everything, for sure, but they are the first step towards a systematic validation of the truth.

 

Truth? :idk

When I gave the iMac to my son, I did a full system restore, upgraded to Lion and did no bother to reinstall WinXP. Now it turns out I can't bootcamp anymore, because Lion won't support XP and my iMac won't run Win7. I do have Win7 on my MacPro and could conceivably compare that, except that I'd have to install Cubase + all VI's on it, and get MacDrive to access all my internal HFS disks. I think I will one day, but not soon.

 

 

 

My point is the Mac's internal sound may well suffice for some people, especially if they use Mac-specific software.

Fixed! Agreed, it's pretty well-established at this point that internal sound cards may well suffice for some people - Mac/PC, shouldn't matter either way (no judgement here whatsoever - you make good music with your setup, then more power to you).

 

My point is that CoreAudio makes the internal sound device in a Mac different from your average Windows machine. As you can see in post #2 by BurningBusch above, his Mac's internal audio is faster than all but the ProTools PCI card. My experience with the Saffire LE is similar. You'd have to spend significant amounts of cash to do better in terms of latency on a Mac. I do get much better LLP with the Mbox3Pro, but that is a >$600 device.

 

In my (admittedly limited) experience, any entry-level interface will perform better than the stock sound chip in a Windows PC.

 

But again, given that the OP already knows that, the only sensible reason for this thread to exist is to ask what's beyond the "may well suffice" point!

 

For those of us in the 'may well suffice' group, there's no point of further discussion. For those who'd like to look beyond, threads like this are invaluable and necessary - so long as there are more real-world measurements and less empty discussion!

 

I am sorry to only contribute "empty discussion", but nonetheless my answer to the OP remains: not unless you want to spend a lot more money. With Anotherscott's description of his setup and workflow, I don't think anything under $500 is worth it in terms of just LLP. Like I said, I'd still recommend a dedicated interface for the connectivity and sound quality, but Macs built-in audio performs surprisingly well. You came on pretty strong saying:

 

And most of these will certainly perform better than the consumer-grade built-in interface in your Mac. Ergo, a good interface will certainly decrease latency.

 

and I am not convinced that is always the case. I have family staying over for the holidays now, but come Monday things will be back to normal, and I'd be happy to report Cubase's RTL's for my MacPro, SaffireLE and Mbox3Pro then.

 

[zephonic]

Afterthought, I get the impression that you do not use a Mac yourself?

 

You have made some pretty broad-sweeping statements here, most of them presumably based on the DAWbench results, but have you used a Mac with its internal audio?

 

For example, you say:

 

My humble TC Desktop Konnekt 6 has about half the latency of my laptop built-in soundcard, at 64 buffers. And I can use a much larger load at 64 with the TC than with the builtin soundcard.

 

Hence my confident assertion that a good external interface (at least as good as my TC Konnekt) will improve latency performance.

 

I do not doubt this for a second, as my experience with WinPC's are the same. But CoreAudio really makes that cheapo internal soundchip work. As Busch (whose opinion you seem to value a lot) stated:

 

With Presonus Studio One you can see input and output latency. Here are what the following devices registered with a setting of DEVICE BLOCK SIZE (buffer) of 128 samples (64-bit):

[font:Courier New]

Digidesign TDM (PCI): In 5.8 ms Out 2.9 ms

Built-in Audio: In 5.8 ms Out 4.4 ms

Konnekt 24D (FW): In 8.8ms Out 5.9 ms

Venom Synth (USB): In 10.93 ms Out 8.03 ms

Kronos Synth (USB): In 7.68 ms Out 4.68 ms

Edirol M16DX Mixer (USB): In 14.90 ms Out 6.98 ms[/font]

 

The Digidesign TDM card is the only PCI card in the mix which is undoubtedly why it does so well. I suspect the main reason the Edirol does so poorly is that it provides a lot more I/O (18 in/2 out) although the Digidesign interface provides 8 in/8 out and the TC Electronic Konnekt 24D also has numerous I/O. I believe that the main concern when running MIDI in and audio out is the output latency. FW doesn't do any better than USB. Given my interfaces, in a portable setting the Kronos would be my best bet.

 

Now this is at best a semi-educated guess on my part, but I bet that both your and his TC unit use essentially the same driver, with suitable modifications for differences in IO etc. (If not, I'd say his more expensive model should probably perform better than yours).

 

Yet on his Mac it does not achieve the same latencies as the internal sound, whereas you say yours has roughly half the latency of your built-in sound.

 

See what I mean here? Just because built-in audio on WinPC's are almost by definition inferior to dedicated units, does not mean that this necessarily the case on Macs as well, which -I think- is the point others here tried to make.

 

 

[Ashville.Guru]

Zeph - first off, tone and intent often translate poorly in text. If you and I were discussing this over a cup of excellent coffee, I could say the same words, and you'd read no judgement or dismissive intent! The 'empty discussion' part was not specifically aimed at you, but the general discussions we've had on interfaces and latency. Yes, I do have an agenda - which is simply to goad the community in the direction of more measurements, which is good for everyone in the long run.

 

You're right - I currently don't use a Mac (I have in the past), but that's only because I've never seen any convincing numbers (performance benchmarks) in favour so far. I'm quite open, and perhaps during the course of discussion you could change my mind!

 

On to the facts, then. I hadn't paid enough attention to BurningBusch's numbers until now, and yes, they do point towards what you're saying - but there's a caveat. I've read that numbers reported by the DAW are merely the same numbers reported to the DAW by interface drivers; and it's been established that drivers do a rather poor job of reporting this. Hence, tools like the CEntrance Latency Test Utility, which I used in my measurements. While I don't think reality is all too different from what BurningBusch reports, I'd still need to see a more reliable test before declaring any result.

 

So I really do look forward to any numbers you can provide. However, like I said, one shouldn't put stock in what the DAW - Cubase/Studio One reports. And the CEntrance tool doesn't work on Macs. But there's a better alternative - the simple 'hardware differential' test I described earlier in this thread. The playing field doesn't get more level than that. It measures only output latency, not RTL, and will work equally well on both Win and OSX. Just make sure the softsynth comes in both VST and AU format, and it's best not to use a standalone version rather than a DAW (eliminates one more variable). Pianoteq demo comes to mind as ideal for this purpose.

 

This test should be ideal on a dual-boot machine; but still latency measurement is only one part of the story. There's still the processor load to consider. This is where the DawBench VI Benchmark suite (download link) would be most useful. This would tell us how much polyphony can be handled without glitches, at each buffer size, on the two different OSes. Until this whole set of measurements is done, one can only say that the jury is still out. But I will say that you've made the strongest case I've seen for OSX so far...!

 

Mostly, though, you and I are in agreement. You'll see that I've acknowledged all of this as possible in my response to Reezekeys above (the 3 'nits to pick'). I've also specified a DawBench rating of 7 or above for beating builtins on OSX - which exactly corresponds to your >$600 threshold. But here's a request - let's not get into the $$ aspect, especially in a Mac/PC discussion!

 

And I do look forward to discussing this over a cup of excellent coffee. Which can only be filter coffee, grown in the hills of South India. Now that's not prejudice, it's fact, and you do not want to dispute it with me!

 

 

- Guru

 

P.S. It's not BurningBusch's opinions that I value, but his consistent propensity for action. Be it latency measurements, pushing the limits of what's capable with sample loading on the Kronos, or creating kickass EP samples for the community. And have you seen the pic of his erstwhile laptop rig? This man's posts have expanded my concepts of what is possible with technology, and I'm sure I'm not the only one. Come to think of it, I can't recall reading any post of his which qualifies as pure 'opinion'...! I'm happy and proud to give credit to anyone who can actually contribute to the community - the least I can do.

 

[Tusker]

Yes, I do have an agenda - which is simply to goad the community in the direction of more measurements, which is good for everyone in the long run.

 

Wouldn't it be nice if everything could be measured?

 

Unfortunately, there are measurements and there is the real world of complex electronics where one single factor doesn't explain performance. By the time all the yardsticks are properly defined and benchmarks are universalized, the bleeding edge of experience has moved on.

 

When we know something works we should share that with each other.

 

The macbook pros are simply "more than adequate" for decent sound quality and reasonable latency when using a modest number of plugins. There are probably some other (Dell XPS? Sony?) laptops that work off the shelf in the same trustworthy way. Let's not allow "waiting for measurements" to prevent this community from sharing what works, and what doesn't.

 

(OT - Currently attempting to convert an Ipad into a multi-app processing environment using audiobus ... )

[Dglavko]

The macbook pros are simply "more than adequate" for decent sound quality and reasonable latency when using a modest number of plugins. There are probably some other (Dell XPS? Sony?) laptops that work off the shelf in the same trustworthy way. Let's not allow "waiting for measurements" to prevent this community from sharing what works...

My two-month-old 11 inch entry-level MacBook air ($999 model) runs the new Pianoteq acoustic, wurli, rhodes and clavs, as well as broken wurli VST's flawlessly. I am using the headphone out jack and no audio interface (although I have one if needed) and can detect no sound quality deterioration or noticeable latency (not saying it is 0, just that I don't notice it). I am triggering it with a Crumar Mojo organd using a midi to USB cable only.