Jump to content

Why I get higher roundtrip latency than others for a given buffer size? (and why it actually doesn't matter?)


Recommended Posts

Hey Guys,

(Sorry for the longer context but just wanted to be sure all are on the same page :) )

I've recently moved to a MacBook Pro M1 (13, 2020) and bought Logic as I want to dive more into home-recording etc. My audio interface is Focusrite Scarlett 2i2 (plugged via USB-C), as of now I'm mainly playing guitar through it, using Neural DSP plugins, but soon want to get more into programming drums, recording vocals etc.

The whole set works great as a gear for jamming to some music with Spotify as background, no latency whatsoever, sounds great (in the Neural DSP plugin I set the buffer size to 96 samples, the plugin in standalone mode shows 1.7 ms latency).

Yesterday I've installed Logic, wanted to give it a try, checked in the audio settings if it sees Focusrite as the main audio input/output, and yes indeed that was the case. I've lowered down the I/O Buffer Size to 32 samples and Logic showed 3.8ms output latency, 7.6ms roundtrip.

Ok, I thought that's pretty high, but then I've drag and drop some drum beat, insert 2 audio tracks, panned left and right, in both I loaded the Neural DSP VST. I've recorded two tracks, separately, with monitoring via Logic (so I can hear the sound of the guitar with distortion from Neural etc.) and actually... did not have any problems with timing, no delay of the sound going through my headphones, the recorded tracks were actually quite tight and on time.

And here are my questions:

- On different tutorial videos etc. people get much lower latency (or at least much lower values showed by Logic) when turning the buffer size to 32 samples (on average I would say 1 to 4ms roundtrip latency). Why is that the case, that latency showed by Logic on my MBP on 32 samples decreases only to 8ms? The M1 is a beast, Focusrite 2i2 is every's bedroom player audio interface so wanted to know if it's normal, if it should be a concern etc.?

- Is there a way to push it any lower? I do not experience any pops, cracks at 32 buffer size. Does the roundtrip latency depend more on the audio interface I'm using or the CPU itself?

- Is there any sense to push it lower? As I've said, I get no inconveniences from these settings as of now, but because of many tutorials I've seen on the web where the roundtrip latency Logic numbers being much lower on 32 buffer size (and in general), I've started to think if I just happened to not be that affected by 8ms latency when playing or the issue lies somewhere else?

Would be extremely grateful for any advices, comments, guidance etc.!

Link to comment
Share on other sites

15 minutes ago, kewson said:

Yesterday I've installed Logic, wanted to give it a try, checked in the audio settings if it sees Focusrite as the main audio input/output, and yes indeed that was the case. I've lowered down the I/O Buffer Size to 32 samples and Logic showed 3.8ms output latency, 7.6ms roundtrip.

Yes, that's what I get on my Focusrite.

Latencies are generally driver and hardware dependent, so different interfaces will report different latencies to the system. Also, higher sample rates will generally halve the latency.

A 1ms roundtrip latency seems... optimistic for native DAWS... where did you see this?

Edited by des99
Link to comment
Share on other sites

Hey Des, 

First of all thanks for the reply!

Really good to know that you get the same results on your Focusrite - so in terms of hardware apparently it is how it is and there's no additional latency coming from some of it being faulty. Are you normally recording on these settings and everything seems fine or you do some additional tweaking? I've heard the Logic Low latency mode and the fact that you can cap the latency to some value does miracles, but haven't tried that yet (+ as I've mention, I don't seem to be affected by the current latency, at least that much)

In terms of sample rates, changing it from 44.1 to 48 does a minor improvement (0.2ms), but I wouldn't like to go higher than 48 to be honest. 

In terms of roundtrip latencies from tutorials, yea I might presented it to optimistically 😅. From what I can see now it rather revolves around 3-4ms (which is still quite better than my results), selected examples:

https://www.youtube.com/watch?v=r3R_5rjPw3Y&ab_channel=HeinrichHouse - 3.4ms roundtrip, 13:54 in the video

https://www.youtube.com/watch?v=t0S1o19dMaE&ab_channel=OlaEnglund - 2.9ms roundtrip, 14:08

https://www.youtube.com/watch?v=MmQLmHTDZvY&ab_channel=MusicTechHelpGuy - 6.9ms roundtrip on 64 samples, so I guess should be closer to 3-4ms on 32 samples, 0:38

Looking forward to hearing from you!




Link to comment
Share on other sites

Generally speaking, in most cases, 4ms of latency is not really noticeable (there are exceptions). Humans find it hard to reliably distinguish anything under about 10ms, but again, depending on circumstances, you can sometimes *feel* the difference.

Honestly, I have no problems tracking MIDI parts at 128 samples. Playing a guitar through an amp sim, 64 feels "alright" to me, but 32 feels "better", so I'd go 32 in those cases. For regular tracking, vocals etc, latency is less important for me as I would generally tend to (out of habit) do monitoring and comfort reverb via an analog mixer, but really, whatever your setup, you have to try various things out with different performance and applications and figure out the approach that suits your needs best.

I have zero problems with a 32 buffer at those settings. Going from 44 to 48 won't affect latency much, but when you go from 44/48 to 96, that's when your latency is halved. As you say, most people don't need this...

Also, don't get confused between recording/audio driver latency, and latency induced by plugins, which is a whole different thing. It's still latency (in that it delays the audio signal) but they are caused by different things, and at least the plugin situation is more controllable.

Link to comment
Share on other sites

Anything under 10ms RTL is fine.  Its actually very very difficult to get less than 5ms of RTL in actual use.  For sure you would need to be running with a 32 sample buffer, which many computers can simply not handle without dropout city.  But further to that, many sound card devices and their drivers are simply not efficient enough to reliably provide audio with only 32 sample buffer.  Many will actually add additional internal buffering in the lower buffer settings to enable it to work, and thus can't get below the 5ms barrier in actual practice.

You can find a database of actual test results of many devices here:


You will notice that the devices getting the lowest latency are PCI based solutions.  If you're using USB it becomes very difficult to get actual measured latency below 5ms.  The driver either won't provide it or lowering your buffer to 32 samples will peg your CPU.  M1 Macs can possibly handle the 32 sample buffer, but still you have drivers that are doing whatever they are doing.

In order to know what RTL you are getting you need to actually measure it, not rely on what LogicPro reports, LogicPro only reports whatever number the sound card driver is reporting.  But the sound card isn't measuring anything its just hard coded to know how it processes the buffer and supplies that number to the DAW.  In order to know your latency, measure it with a loop back test.

In my opinion, anything under 10ms RTL is quite good and fine enough.  But really even at 10ms if you are using software monitoring while singing with headphones, for example, you're going to hear some artifacts in your inner ear.  Its better to use direct monitoring for that sort of thing, and forget about chasing some 3ms latency number that you heard was possible.  Even with your new M1 Mac, there are various moving parts involved with getting lower latency...and the M1 is not a magic cure all to get near zero latency...



Link to comment
Share on other sites

Des, Dewdman, 

Thanks so much for the detailed answers! I was mainly concerned and irritated by the fact that potentially something was not ok with the new gear I bought, unnecessarily comparing it with videos etc. so your replies that everything seems to be fine calmed me down a lot and I can focus on writing and recording! It seemed weird from the beginning since in any other aspect the M1 and Focusrite were killing it, so wanted to double-check. 

For singing, Focusrite has the built in direct monitoring functionality, so that shouldn't be a problem. Unfortunately, it doesn't work with guitar if you want to hear distorted sound and not a DI, but well 

As I said before, the Mac seems to work well with the 32 buffer size setting, no pops or cracks, so fingers crossed it will stay that way :)

Thanks again guys!

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...