Hi, sorry for bit offtopic,

I'm seeing 1.6/2.1ms in Reaper 7.02 at 44.1k/24bit at buffer 32 samples set on RME HDSPe AIO in Windows 10 on 7 years old i7-5820K PC,
can't set buffer 16 for some reason though...


EDIT: I'll take a look at UFX III next week if you remind me

 

it's your cpu man, an i7 7700k doesnt have the power, thats why it's crackling. i'm on a ryzen 5700x, i can go to 192khz at 16 samples without crackling on standalone ampsims, with studio one i use 48khz at 16 samples, tons of plugins, zero cracks

 

I wrote something about this in another thread which may be interesting for you.

Latency Analyzing Application for Mac

I'm getting 4.6ms round trip latency on macOS with my RME Babyface Pro FS at 44.1kHz and 64 samples buffer size. In this post I described how I measure it myself and don't rely on any number reported by a DAW.

 

I usually use Ableton but that DAW has no buffer size 16. Usually I work in buffer size 64 when recording live stuff cause that makes a little room for plugins and stuff like that. Here is Cubase on buffer size 16 .....44.1KHz:


Of course no USB interface can get number like the Quantum.

 

The German c't tested 10 "budget" interfaces with USB-C earlier this year: https://www.heise.de/select/ct/2023/2/2232711203621768202
At 48 kHz and 32 samples buffer, the fastest was Steinberg's UR22C with 3.6 ms round-trip latency, followed by MOTU's M2 with 4.3 ms and the Scarlett 2i2 3rd gen with 5.5 ms. However, these numbers are so close together that you wouldn't notice any difference.

 

Last edited: Nov 3, 2023

That's great but not what he is trying to do. If you cannot get an i7 to record 44.1 16 bit, there is a problem. We could record multiple channels of 44.1 16 bit (and of course higher) without latency problems into Protools with nubus macs running about 250 megahertz processors pre-2000. The plugins and your DAW software are what are the CPU burden. Not simple recording of 16 bit 44.1 audio.

 

i looked at the steingberg's ur22c and it has a big driver problem with windows 10 im using windows 11 tho and i dont want to buy an interface i cant use it looks cool but if it has driver problems then it's a no go

 

Go look on reverb.com. Almost every UAD Solo brand new have new $200 price drops on them, as of now. And that is only going to continue.

You can find lots of complaints about them not having a ton of external DSP power, prices, and even about the company if you look around online. I have never seen a latency complaint that I have noticed, nor have I seen any RME users say they sound better. That takes more than just looking at some numbers. Multi-channel recording with slow as hell Macs was possible on Protools using Apogee AD8000 interfaces, 20 years + ago. You are getting 1 forum worth of advice from people who all have bought the same things. That same interface on a Mac will be no standout results. Look at above mac/rme user with 4 ms RTL. That is double my Motu and with a Apple Silicon CPU that is probably 3 times faster. Probably more than that.

 

Last edited: Nov 4, 2023

Unless each Yamaha Steinberg behaves differently despite using the same driver, I don't have any issues with my Yamaha mixer on Windows 10. Did you check the date of the post? Maybe it's an issue that is already fixed.

 

I made the work for you:

You can read more at this link: https://dawbench.com/images/DAWbench LLP Database-December-2022.pdf

RME Fireface UFX+ (other RME products that use the same driver should have more or less the same latency, you can check which products share the same driver here: https://www.rme-audio.de/downloads.html ) will give you 1.111 IN/2.063 OUT for a total measured (some interfaces don't report accurate numbers or don't take into account AD/DA conversion) RTL (round trip latency) of 3.187 ms at 32 samples.

Note that at this sample buffer you get 120 RXC* but 0 CV* and 0 NCV*(read below for explanation of this) so some work could result in dropouts, for example using kontakt libraries but could be good for tracking and using Vst or other Vst instruments other than Kontakt.

RME Babyface Pro (other RME products that use the same driver should have more or less the same latency, again you can check which products share the same driver here: https://www.rme-audio.de/downloads.html ) will give you 1.973IN / 2.358 OUT for a total measured RTL of 4.325 ms at 64 samples.
At this buffer all operations including using kontakt libraries should have no dropouts (within the limits of CPU intensity of plugins used of course)

RME Fireface UFX+ (and other RME products with the same driver) reports a slightly higher RTL of 1.837 IN/2.789 OUT
for a total of 4.646 ms at 64 samples but the added 0.321 ms will give you 141 RXC*/180 CV*/200 NCV* vs 122 RXC*/120 CV*/160 NCV* instances before dropouts, so you'd gain more CPU efficiency and you could do more work for an imperceptible amount of added latency but at a substancial higher price range.

*RXC= Reaper multiband compressor
*CV= Kontakt instance with convolution verb active
*NCV= Kontakt instance without convolution verb active

*Look here for a more detailed explanation of how the test (DAWBench DSP and DAWBench VI) to calculate CPU efficiency are conducted:
https://original.dawbench.com/benchmarks.htm

In conclusion RME Fireface UFX+ (or other products that share the same driver) has the lowest possible latency ( not considering possible dropouts when using Kontakt for example) at 44100 SR at 32 samples and a slightly higher (0.321 ms) than a RME Babyface Pro (or other products that share the same driver) at 64 samples.

Case closed.

EDIT:
what you heard about Apollo Twin (if MK2) is incorrect. *** read last line in the post

That interface has:

90 RXC/0 CV/0 NCV and a 2.653 IN/1.293OUT for a total RTL of 3.946 ms at 32 samples with Input delay compensation OFF.

125 RXC/180 CV/200 NCV and a 3.379IN/2.018OUT for a total RTL of 5.397 ms at 64 samples with Input delay compensation OFF.

*** I now saw that the Apollo Twin Mk2 is a thunderbolt 2 interface so you were talking about the Apollo twin USB 3 interface but unfortunately I don't know what latency this interface could give you.

 

Last edited: Nov 4, 2023

thank you for the work i will really consider saving more money and buying a babyface pro we are talking about latency at 44100 hz right? becouse if it's higher samplerate i cant use it with my current pc

 

Yes, as I wrote we're talking 44100 hz sample rate.

All considered, if you don't need more than 12 in/12 outs (using ADAT I/O) the Babyface PRO FS (FS is the current model that shares the same driver as the Babyface PRO) is the best USB interface you can buy not only for lowest latency but overall quality and CPU efficiency.

The only other interfaces I would consider are UAD Apollos if you're interested in their ecosystem and unison technology but I don't know if their USB only line have latencies that could satisfy you.

From what I read in this thread https://uadforum.com/community/index.php?threads/apollo-solo-usb-latency.51614/ they're not the best for low latency work inside a DAW.

 

Last edited: Nov 4, 2023

CPU is a 10400F. No issues here but to achieve it you might need to know how to config windows (this one is 11) and BIOS.

Kinda weird that you're getting 4.1ms on a gen3 Solo. Oblique measures 3.96ms.

 

Interesting difference between Buffer (16) and Round-Trip Latency (176)...

 

That is probably just the internal buffer, something like 0.4ms.

 

MOTU 828ES 16 samples at 44.1 khz in Cubase

 

This is what i wrote in another thread, it explains how buffer sizes, AD, DA and driver buffers sum up:


So far I know RME is the only manufacturer giving you FULL specs of their devices and drivers. I tried it with my Babyface Pro FS and my measurements accurately match these specs.

For example: using macOS, Core Audio is using a small buffer for recording and playback of 16 samples each. The Interface uses a small safety buffer of 32 samples for playback to stabilize performance. The BF Pro FS has very fast AD and DA converters which only need 5 samples AD and 7 samples DA at 44.1kHz.

Using 44.1 kHz sample rate and a buffer size of 64 this results in 5 + 16 + 64 + 64 + 16 + 32 + 7 = 204 samples or 4.6ms delay. And this is exactly what I measured myself.

Using Windows ASIO drivers, it is even faster because the 16 + 16 samples delay of core audio are absent. This results in 172 samples or 3.9ms delay @ 44.1kHz.

 

That shows how "slow" the interfaces are compared to RME, buffer size is not everything. The RME BF Pro FS in my above example reaches 172 samples or 3.9ms RTL in Windows with 64 samples buffer size because it has super fast AD and DA converters.

Also remember that all plugins that are running "live" need to run at the actual buffer size, and everything less than 64 samples taxes the CPU pretty hard, and some plugins may not run at all with 16 samples buffer size. Technically, the process loop of a plugin has to run four times as often at 16 samples than it has to at 64 samples.

 

If you really care about latency you should work with an hardware setup.

When I work with my HP Z600 I use a buffer of 256 samples, it gives me an 5.8 in latency and an 7.0 out latency with my RME USBface.
The lowest buffer of this sound card is 32 samples.
You can easily make music with 64, 128 and 256 buffers your brain will always compensate timing. (Your body always follow brain orders)

I get rid of noticeable latency with my Atari 1040STE and hardware samplers, instruments and effetcs, and recording with direct input monitoring, with RME TotalMix FX.

I'm of topic for not answering to your question but if with what you have now you are focused on latency instead of making tracks man, you will end up in a dead end, and you will be never satisfied of what you got. Just my thoughts bro.

PS: Don't forget you can get rid of latency, it's physics even in an 100% hardware setup, the speed of light have also latency, nothing in the universe is 100% instant.

 

Really?