Hi there,
I am currently trying to optimize the latency of my overall setup a little bit. Already switched from a Focusrite to a MOTU audio interface, which already was quite some improvent (from 14ms down to 10.5ms). What are your measurements (using GigPerformer) for the roundtrip time and what are your exact settings? I am quite unsure whether my values are good or whether there is a “serious” problem with my setup.
For comparison my values:
Best regards
Wolfgang
Whether it is good or not, depends partly on yourself. I (guitar player), for myself, am not comfortable with a latency > 20 ms (I can hear the delay between my plectrum flapping and the signal via the in ear monitor). 15 ms I can definitely hear, but is doable. Less than 10 ms is fine.
Some vocalists don’t want to have any latency, because of the combing effects of his/her voice directly combined with that via the monitor. But keep in mind that a distance of wedge monitor on the floor to your ear (2 meters or so) already adds more than 6 ms latency.
TL:DR;
I would be happy with this latency.
I agree with Frank. I can also notice 15 ms.
~10 ms (or less) is OK.
Same here - 10ms is where I start getting jumpy.
Wouldn’t this also depend on the computer’s audio buffer settings for rate and number of samples? Did that change with the interface?
These settings are properties of the interface and it’s driver. Not really of the computer. Of course the os and the used computer interface (usb, firewire, pci) also might add latency and the speed of the cpu will also make a (little) difference, but the quality of the driver makes a lot of difference.
For instance, I’m using a focusrite with 64 samples @ 48khz. Theoretically that’s a round trip of 2,33 ms. In practice however, I end up with 7 ms due to the added latency of the usb interface, dpc handling, etc. But I also use a Behringer umc1820. Same settings. The roundtrip of this one is closer to 6 ms. Same computer. Same usb interface…
Supposedly RME has even lower latencies, but I’m not having the opportunity to put that to the test.
So in my opinion: on the same computer the audio interface and its drivers are an important factor, if not the most important, when it comes to latency.
RME Babyface Pro.
Measured round trip latency:
5.5 ms for 96 samples @ 44.1 kHz.
That is what I’m using for production.
… is quite acceptable, but there are interfaces with lower latencies at 128 samples.
You have to ckeck if it is good for you. If yes, then don’t care about this number.
Are you using ASIO drivers?
Yes, I am using the ASIO drivers. The RME Babyface has lower latency, I know. But for my Focusrite it was even 14ms at 128 samples. Wonder which other interfaces have lower latency.
I use an RME, but as far as I know MOTU M2/M4 have a quite low latency.
That’s why I switched to the MOTU 
I’m curious if anyone has fully implemented the GP PC optimization guide, and if so, how low were they able to get their latency.
(https://gigperformer.com/docs/ultimate-guide-to-optimize-windows-for-stage/The%20Ultimate%20Guide%20to%20Optimize%20your%20Windows%20PC%20for%20the%20Stage.pdf)
I’m planning to use GP with a Ui24R to allow use of VST’s in my live mix.
If I only use GP for send/return effects (Reverbs, delays) then the latency is just pre-delay and not supper critical.
But if I start using GP for inserts on channels (or on the main LR bus) then latency starts to become a problem. I would imagine anything over 10ms will start becoming apparent. There is also the problem that some signals could have multiple passes through GP - say a channel with an insert, and then an insert on the LR bus.
So how low can you go?
Alot depends on whether you are playing VSTis vs monitoring audio inputs.
You can easily get aways with those settings as a keys player.
Monitoring audio is where things get tricky.
Those settings are fairly common although getting under 10ms would be ideal.
That said most vocalists i record dont seem to notice the latency.
Out of 10s of people i recorded only one Rapper complained about latency.
I think many people actually can manage up to 20ms. As every once in awhile i will add latency inducing plugins and i start noticing it before the performer does.
i forgot to mention that going to 64 samples is the logical next step.
That should get you under 10ms and im fairly certain hardly anyone will notice.
You can go even with 16 samples, if your audio interface supports it, but - why bother? 
People won’t figure out. Here’s a comparison:
so now I have to ask, how does one launch this latency monitor in GP5?
It’s a good thing I can still hear because I obviously can’t see. How many times have I looked directly at that and didn’t see it?
It claims my Behringer UMC404HD has 25.3 ms. I have it set at 96k with a 1024 buffer and GP says 1024 samples results in 10.7 ms. If I lower my buffer size to 512, I risk occasional cut outs. No amount of cut outs are acceptable to me. If I raise it, the latency is obvious to me and affects my ability to play. I find these settings to be a reasonable balance.
That’s seems like an extremely high latency value. If you look at the examples on the user guide, you will see that at 128 buffer size at 48k the round trip latency is less than 6ms
I don’t know anything about that audio interface but you should be able to get way less than 10ms round trip with a buffer size of 128. Are you sure you’re using the correct driver?
Also, in my humble opinion, 96k is overkill for live performance. If you can’t run your system with 128 buffer size at 48k or 44.1k (the latter is what I use myself), then there’s something else wrong somewhere, either a driver issue or a computer that’s not fast enough.
@dhj
My computer:
It took some doing to get rid of all of the “nanny” services provided by Microsoft, Dell, Alien and the bios speed stepper. I think I still have one gremlin running every 15 minutes or so. But it doesn’t interrupt real time audio with the settings I have been using.
Some test results for you:
I prefer running 96k for a couple of songs I play where the keyboard stands alone attempting to sound like a small orchestra. For the rest of the songs, 48k is probably ok, especially since the band’s mixer handles sound at 48k internally. I run at 1024 samples because of that occasional glitch I was getting last time I tried running at 512. I do not find the latency of my preferred settings to be noticeable. I press a key, it makes a sound. 2048 @ 96k on the other hand is unbearable.
I have a break over the next couple of weeks. I will try running 48k. I will likely only drop samples to 512 (half the sample rate, half the buffer size). Latency isn’t noticeable and I see no need to disturb the gremlin. We will see if anyone other than me notices a difference in those few songs where I have felt I need to push for the highest possible quality.
