Lowest Latency USB interface?

[Jim Roseberry]

1 minute ago, Gswitz said:

For all us regular humans, USB 2 carries enough data to support all our tracks.

 

So is usb c faster? What is the benefit? Just having it work on a USB c port isn't worth it to me.

FWIW, To my knowledge... all the new audio interfaces that have USB-C ports are actually USB-2 (not USB-3.1).

There's zero performance advantage.  Some of these new audio interfaces come with a USB-C to USB-2 adapter cable (to connect to standard USB-2 ports).

[Bill Ruys]

On 2/5/2019 at 6:42 PM, Jim Roseberry said:

The 896mk3 Hybrid interface (and all the Hybrid series) was (round-trip latency wise) a step backward from the original 896HD.

The original 896HD yielded 5ms round-trip latency at a 64-sample ASIO buffer size 44.1k

The Hybrid series added onboard DSP processing/mixing... and that increased round-trip latency to ~6.5ms at those same settings.

 

To get sub 4ms round-trip latency from MOTU USB, you have to be running one of the newer AVB models (or spin-offs).

The newer MOTU USB drivers allow you to tweak the safety-buffer size.

 

FWIW, Thunderbolt under Windows is not a crap-shoot. 

You just have to make sure you've covered all the details.

MOTU was one of the  first companies to have (release version - not beta) Thunderbolt drivers for Windows that support "PCIe via Thunderbolt" (allowing PCIe level performance).

Thanks for all the info, Jim.  I have also read that latency on the first generation Ryzen processors is a lot worse than Intel CPUs.  Apparently the 2nd gen (2000 series) chips are significantly better.  No news yet on the upcoming 3000 (Ryzen 2) chips.

Given that Thunderbolt is an exclusively Intel technology, I think my DAW experiment with AMD is over and I'll be building an Intel-based DAW next time with good Thunderbolt support and look at one of the newer MOTU AVB interfaces.

[Toddskins]

10 hours ago, Jim Roseberry said:

Sample-rate refers to the number of samples per second

44.1k = 44,100 samples per second

48k = 48,000 samples per second

96k = 96,000 samples per second

 

The higher the sample-rate, the quicker a given buffer size.

64-sample ASIO buffer size at 44.1k = 1.5ms

64-sample ASIO buffer size at 48k = 1.3ms

64-sample ASIO buffer size at 96k = 0.67ms

 

Thus at a given buffer size (64-samples in this example), as sample-rate gets higher... latency gets lower.

ie: Doubling the sample-rate (twice as many samples per minute) cuts the latency of a buffer in half.

One might then expect that doubling the sample-rate would cut your audio interface's round-trip latency in half.

The X-Factor is the driver's safety-buffer.

If the driver's safety-buffer (often hidden) is consistent size across all sample-rates, doubling the sample-rate will cut round-trip latency in half.

If the driver uses a larger safety-buffer for higher sample-rates, round-trip latency will be slightly lower at higher sample-rates.

 

Many audio interfaces don't allow using ASIO buffer sizes smaller than 64-samples when working at sample-rates above 48k.

 

So Jim, the latency that people experience is reduced by using a higher sample rate - because more music samples of sound are being streamed at any given time?  Is that the upshot of it all?

Again, so if we want to experience as little delay as possible in our monitors, then use as high as possible  sample rate (cpu permitting) that your device can go?

[Jim Roseberry]

On 2/8/2019 at 7:13 PM, Toddskins said:

So Jim, the latency that people experience is reduced by using a higher sample rate - because more music samples of sound are being streamed at any given time?  Is that the upshot of it all?

Again, so if we want to experience as little delay as possible in our monitors, then use as high as possible  sample rate (cpu permitting) that your device can go?

The higher the sample-rate, the smaller amount of time for a given ASIO buffer size (which reduces latency).

Using high sample-rates is one way to potentially mitigate higher round-trip latency.

The downside; CPU use will be considerably higher

Note that the audio interface's safety-buffer is the major X-Factor when it comes to round-trip latency.

[Rico Belled]

My uninvited 2 cents:

Most of my Mac buddies with overpriced Pro-Tools setups on their overpriced MacBook Pros never understood why I kept saying that once you go full input-monitoring you can never go back... until we did the sessions for my album XR7. I built a machine specifically for that record; had little money so it was a first gen i3 with 4GB of RAM, MOTU PCI-424 and a 2408 MK3, a 2408 MK2 and a 1224 that I got for less than the cost of a headphone amp. We recorded 4 or 5 guys at a time, with 4 independent headphone mixes, eqs/compressors/verb at 44.1K 24 Bit at 64 sample buffer size for days, not a single glitch ever.  I'm talking 8 or 9 mics on the drums, 2 or 3 on percussion, bass, guitar and keys in stereo. This on a machine that even at that time was low end! Sonar was flawless and these dudes could finally see the power of it.

Listen, if you're an amateur or even a pro with let's say more 'relaxed' sense of time latency is not as big a deal. If you have a real master drummer, anything over 5ms is just not gonna work; a lot of singers are the same. When Jim says 1ms my heart jumps! I don't wanna drop that kinda cash, because of you know, priorities, but I very much look forward to an even lower latency future.

As I write this, I'm thinking of the effects of a latency that's too high. What happens is that, just like on a big stage, a great musician will automatically start playing a little early, just so his part sits right in the mix he's hearing. When you play it back it's actually rushed!

Finally, the bandwidth increase in USB3 compared to USB2 has absolutely nothing to do with latency. The latter has enough to do hundreds of tracks. Overall system latency is a very complex interplay of factors, how hardware behaves, quality of drivers etc. Case in point: as long as you set sample rate to 48K, my Surface Pro 3 is amazingly stable at 64 sample buffer size with ASIO4ALL. I tried for a while at 44K and it just crackled but I guess that's because the built-in audio device always runs at 48 and the resampling the driver does causes problems.

 

R

[Larry Jones]

13 minutes ago, Rico Belled said:

My uninvited 2 cents:

Most of my Mac buddies with overpriced Pro-Tools setups on their overpriced MacBook Pros never understood why I kept saying that once you go full input-monitoring you can never go back... until we did the sessions for my album XR7. I built a machine specifically for that record; had little money so it was a first gen i3 with 4GB of RAM, MOTU PCI-424 and a 2408 MK3, a 2408 MK2 and a 1224 that I got for less than the cost of a headphone amp...

...I don't wanna drop that kinda cash, because of you know, priorities, but I very much look forward to an even lower latency future.

Sensible position. I also recorded a full rock band (one time only!) with an i3 laptop, a bunch of cobbled together outboard gear and SONAR. In my case it was a live recording and no headphone mixes, so latency didn't come up, but your point is well taken: We dream of better performance, but we can do good work on an extremely tight budget when we have to.

PS: I'm a PC guy, but I won't chime in with you on dissing Apple products. Those guys are relentless in promoting their little 5% market share.?

[Rico Belled]

31 minutes ago, Larry Jones said:

PS: I'm a PC guy, but I won't chime in with you on dissing Apple products. Those guys are relentless in promoting their little 5% market share.?

For me it's two things: first off the arrogance of the typical Mac user here is quite unbelievable! "I could NEVER use Windows, it's just AWFUL!" The smugness and cockiness just begs for a dose of truth and reality.

 

More importantly though, around here they are really starting to at least inconvenience themselves, but mostly really hurt themselves and not just financially.  When you combine the insane amounts they lay down for gimped products like the iPad Pro, fancy laptops etc with the fact that almost all musicians are complaining about not having enough money it just doesn't make sense anymore. They bring all their dough to the Apple store and Guitar Center for overpriced preamps and stuff, then say they can't pay a guy to play on their records or demos! Again, if we were all flush with money, go ahead and buy that $3000 laptop, $1000 iPad. Or you can get a used Surface Pro 3 (bought an i5 with keyboard and pen for $240 shipped!) and run Ableton Live on the same tablet you read your charts from, catch my drift?

An even better example of how the lovers of the fruit are hurting themselves is the XPS 18, an 18.5 inch mega-tablet that I got a few years ago. Not only is it a great machine with space for 2 internal drives, if you read music on a gig, there's simply nothing Apple makes that's even close. It's literally twice as big as the big iPad pro, which means 2 pages at the same time and anyone who reads knows the value of that. If I had a dollar for every time another musician that saw it for the first time went "Wow! Which iPad is that?", followed by me saying it's a Dell, followed by them immediately checking out and losing interest because they could never be seen with a Dell.

Then end result is: having to bring a laptop to run Ableton (which needs a special stand or something) AND an iPad for on the music stand.  Touch screen on you computer? Oh no, thanks to saint Steve they are forced to arbitrarily see the pad as different from a computer, while Microsoft has truly made Windows into a touch friendly OS. Amazingly, even applications that were made a long time ago still work with Touch because MS thinks way ahead. On MacOS it seems every update breaks compatibilty with half the software out there!

I have a good buddy who's pretty much married to Apple, trying to replicate what I do live with my XPS 18 by running a USB touch monitor into his MacPro (paid $3000 for a 5 year old machine with graphics cards he will NEVER put to use) and it even just doesn't work very well, because the OS is not touch aware and software compatibility is a nightmare. He just can't stand it, but won't see the light. I bought another XPS 18 with the Pentium chip used for $180, put an SSD in it, 8GB of RAM, and IT runs Ableton great too, WHILE reading charts.

The worst part is, and I say this because a lot of these guys are my friends and I want to help them move into the future, it's only gonna get worse. Again, it was Steve Jobs' doing to separate the pad from the computer, Cook is just running with it. My buddy was just happily telling me how they've now made a way for apps on the iPad to "talk to each other"! It's hilarious, having to learn ALL NEW SOFTWARE, for instance for writing charts, while I just put Sibelius on my Surface! You see the trend? And then they just keep saying to each other and themselves how it's easier, although when you ask how in the world it's easier to learn a program than just use the one you're intimately familiar with they have no answer. Propaganda is a dangerous thing.......

Have you seen the survey on Protools users? What if Apple decides to switch its Macs to ARM chips? Is Avid gonna make ProTools for ARM? How about all you favorite plugins? I still run the Native Instruments B4, EVP73, some of the best instruments ever made in my opinion. Installed Windows 10 on a computer from 2005 for my buddy, runs better than 7 and Sonar 5 which he was using works great! To me, THAT is simple.

Sorry for the rant guys, but we need to fight the propaganda if we care about each other......

R

[Gswitz]

4 hours ago, Rico Belled said:

 I kept saying that once you go full input-monitoring you can never go back...  We recorded 4 or 5 guys at a time, with 4 independent headphone mixes, eqs/compressors/verb at 44.1K 24 Bit at 64 sample buffer size for days, not a single glitch ever. 

When i record a band with input monitoring (meaning inputs into the rme are sent with virtually 0 latency to headphones or monitors) and no pc based sounds, I usually raise the sample buffer as high as it will go because the buffer is only a risk, not a benefit, given that it isn't impacting the musicians.

From your description it sounds like the musicians were listen to tracks in the computer as well. Still, I'm not sure your buffer needed to be low. 

It's a decent question to ask yourself when recording... What is the safest buffer i can have without trade-offs?

Edited February 11, 2019 by Gswitz

[Rico Belled]

43 minutes ago, Gswitz said:

When i record a band with input monitoring (meaning inputs into the rme are sent with virtually 0 latency to headphones or monitors) and no pc based sounds, I usually raise the sample buffer as high as it will go because the buffer is only a risk, not a benefit, given that it isn't impacting the musicians.

From your description it sounds like the musicians were listen to tracks in the computer as well. Still, I'm not sure your buffer needed to be low. 

It's a decent question to ask yourself when recording... What is the safest buffer i can have without trade-offs?

As I was saying, it was completely safe, that was the point of my whole post! I WANT to use Sonar to control the mixes for everyone, with the plugins I know and love, with the routing that's head and shoulders above any mixer that comes with your audio interface. Nobody was listening to tracks from the computer.

It means when we listen back to the take, it's exactly the same as what we heard when we were recording. Did did hours and hours of recording at 64 samples at it NEVER ever even popped once. Once you get used to that it's nearly impossible to go back to having to deal with yet another separate mixer application.

R

[Gswitz]

I think i didn't understand what you were doing. Glad it worked well for you.