I still would wish for something like an update to the current UFX III with two MADI buses instead of one. Maybe per SFP to get the space on the back.

Because only then you have the possibility to
- record in double speed and
- to expand analog and mic ports at the same time

For example:
- M32-Pro AD and DA (bus1) and
- two 12Mic and one XTC on (bus2)

Or two split devices to the two different buses if you want to split devices according to the remote management tool
Auxdevice and MIDI Remote.

Or you need to change the configuration / setup to use the one or other devices for higher sample rates.

I've had really good experiences with USB over the years, so I disagree with your idea that USB3 alone would not be well enough for an interface like the UFX III.

Back in 2014, I started using a UFX with USB2 and FW400, but USB ended up working better than Firewire.

Then, in 2015, I got a RayDAT and tested whether a PCIe card would give me any advantages in terms of CPU load or latency. I used the UFX in standalone mode, connected through ADAT, and found that it gave me no visible benefit over USB2 in terms of CPU/system load or stability. So, in my experience, USB2 worked just as well as a PCIe-based RayDAT.

I upgraded to the UFX+ in 2016 and found that using it via USB3 worked great, even with its high channel count of 188. And for even more channels, the MADIface XTC has 394 channels and is also well-supported by USB3.

In 2017, I wrote a blog article comparing UFX+ via USB3 and RayDAT via PCIe under load. I used a big Cubase project with 400 audio tracks, 2 VSTs per track, and 3 in the sum, so 803 VSTs in total. Even at the lowest ASIO buffersize of 32 samples at 44.1 kHz (or 64 samples at 96 kHz), there were no audio drops during playback, and the CPU and ASIO load was handled very well by my system (meanwhile a 7-year-old CPU). The conclusion was that there was no visible difference between using UFX+ USB3 vs. RayDAT PCIe based. Thunderbolt is "external PCIe" so it is compareable.

BTW, I just uploaded a new screenshot where you can see this test project with 400 tracks and 803 VST running on my current system with last recent Cubase 12.0.60 Pro:

Latency-wise, the difference between USB3 and Thunderbolt is tiny. I have a comparison chart of different solutions that I own(ed) that you might find helpful, see here: https://www.tonstudio-forum.de/attachme … es-v2-jpg/

If you're interested in getting some more information, like e.g. differences between UFX+ and UFX III and some recommendations, then please have a look at these two threads:
https://forum.rme-audio.de/viewtopic.ph … 62#p200962
https://forum.rme-audio.de/viewtopic.ph … 02#p217002

I would encourage you to try USB2/3 with the UFX series for yourself instead of relying on hearsay or unproven expectations. I also recommend getting a USB3 PCIe card with a supported USB3 chip, like the Sonnet card with FL1100 USB3 controller, to isolate your recording interface from the rest of your USB infrastructure. This can proactively prevent the addition of other USB-based devices from negatively impacting your connected recording interface.

One proactive tip, like I recommended in the above-mentioned thread (again the URL: https://forum.rme-audio.de/viewtopic.ph … 62#p200962). I would directly get a USB3 PCIe card with supported USB3 chip like the Sonnet card with FL1100 USB3 controller to fully isolate the recording interface from the rest of your USB infrastructure. It can proactively help, that adding more and more USB based devices (headsets, Bluetooth adapter, etc.) have no bad impact on the connected recording interface. For example, this product: https://www.sonnettech.com/product/alle … 4port.html

BTW, Win10 already includes the drivers for the Sonnet card with FL1100 USB3 chipset.
Also nice: the drivers use the more efficient MSIs (message signalled interrupts), see
https://forum.rme-audio.de/viewtopic.ph … 26#p145426

Maybe the problems that you have were based on "bad luck", incompatible HW with 3rd party USB3 chips.

Such incompatibilities in HW can sadly always occur and, if I remember right, even with thunderbolt there could be issues.

The best is to ask people upfront or to try something out or you get a turnkey system where all such tests already have been made.

Sorry, my bad. The full channel count does already work via USB3. Is this working also like a dream on AMD CPU's? Ryzen 5000 to 7000 series?

What a beautiful topic!

I'm in a same boat with rfpm and some other guys.

I have RME Fireface UFX+. I use it with Mac and absolutely love it BECAUSE of Thunderbolt connectivity.

Whoever the speaking, Thunderbolt is faster and speaking directly to CPU, and I also can't imagine any reason to use USB over Thunderbolt.

Thus, when Fireface UFX III had been released, I was upset.

I don't know what is going on on Windows-department, but we, Mac-users always prefer Thunderbolt. It's a gold standard for high demanding devices.

I was planing to update my Fireface UFX+ to Fireface UFX III since I hear that UFX III use AKM chips and logics as in my also beloved ADI-2 DAC FS.

I also hear some advantages in Fireface UFX III over Fireface UFX+:

• Better SNRs

• Better THDs

• SteadyClock FS

But my question is: are these advantages worth it? Will I receive any real better quality of sound? I'm a purist guy.

Is the loss of the Thunderbolt port worth the increase in sound quality? Is it, the sound quality, much higher? What do you think, guys?

Also I have questions to dear MC:

• Are we going to see any Thunderbolt Fireface UFX in future? Or this is a final decision and I should try (buy) Fireface UFX III with its USB?

• Which AKM chips uses UFX+ and which UFX III?

Thank you!

Hello dear ramses!

You write in every single topic on this forum ! Are you from Marketing Department? It's cool that you are so involved, but your responses are sometimes a bit off-topic/off-axis, for my taste.

Yep, we all have seen this latencies comparison and Thuderbolt is faster. Not so much, but it is. Thank you for your detailed measurements.

There is no more discussion about what is "worse" or "cooler", imho. Everyone can make a decision by yourself. Some people like USB, some people don't. We don't have any fresh RME device with Thunderbolt anymore. And this is a fact we have to live with.

Anyway, I would like to get answers to the specified questions above, and not to others. Especially from respected MC, if it's possible.

Thank you!

Dear ramses, please don't take offense, I in no way meant to offend you!

On the contrary, it's cool that you help people, chew up some things.

I just wanted to say that sometimes you want to come to the answer itself faster than to everything around it.

Peace!

Nice! These are the results on Win10, lowest ASIO buffersize (64 samples) @96 kHz.

https://oblique-audio.com/rtl-utility.php

Hi all!

Been lurking on this thread for a few months whilst researching new interfaces, and whilst setting this one up.
It's been INVALUABLE so MANY THANKS  to all!
Particular thanks to @ramses as your posts have saved me A LOT of time and effort.

LOVING the FF UFX III, it's a WONDERFUL, highly capable device, which i'm having some serious fun with!

Well, I've made some measurements with RTL Utility, recommended by Matthias, and the Fireface UFX III is goddamn faster!

The results:

Some additional screenshots:

You write 32 samples @96 kHz, where did you configure 32 samples?

At least for Windows ASIO drivers
32 samples are the lowest ASIO buffer size for single speed (44.1/48 kHz sample rate)
64 samples are the lowest ASIO buffer size for double speed (88.2/96 kHz sample rate)
128 samples are the lowest ASIO buffer size for quad speed (176.4/192 kHz sample rate)

The reason is, the higher the sample rate, the higher the amount of data that is being transferred,
so with double the amount of data at double speed, you also need the double buffer size.

Even if Apple does not use ASIO drivers, the underlying transfer mechanisms are actually the same, and at double speed I would have expected Apple to use a higher transfer buffer, 64 instead of 32 samples. Is it perhaps a typing error, or where do the 32 samples @96 kHz come from, where are they set?

Thanks for clarification.

Ok, many thanks for delivering this information, this is quite interesting.

Well, I am wondering, that ASIO has to use higher buffer sizes compared to Apple, although ASIO provides direct access to the recording hardware compared to Apple where audio has to go through Apple sound system additionally.

I would have expected that Windows and Apple share the same basic mechanisms for the selection of buffer sizes.
Therefore, I wonder, what the ASIO buffer size at double speed has to be twice as large compared to Apple
and wanted your confirmation of numbers, thanks for providing this.

It would be interesting if RME could explain the technical background.

That’s interesting. Didn’t know about that. Why then the numbers between real RTL measurements and Logic’s measurements are the same? Very good predication from RME’s side? :-)

You already answered your question by yourself here...

...by comparing these with your results from your RTL measurements.

Some geeks are discussing, and comparing interfaces here:
https://gearspace.com/board/music-compu … -base.html

Their current List: https://dawbench.com/images/DAWbench%20 … e-2024.pdf
Note they publish the results tested on different OS - so please consider it more like an orientation

You surely also have noticed, that RTL generates a real "Ping" Impulse. Logic, just reads the values, reported back - thats the Difference.

Made a Test (just) for you, to explain what i meant in my last post:

Left side: reported values by Logic for IN/OUT of my FF802... Right Side:  I just switched the Input to Mic of a "dumb" Plantronics USB Headset.
So according to Murphys Law:
1,4ms - 1,1ms = 0,3ms Latency for this Mic??? - Never Ever!

If there are still people believing DAW calculations are accurate:
Buy cheap Headsets instead of expensive Units!
But before throwing your Units to Trash, contact me. I will recycle them for you ^^

Anyhow, if i may add my personal opinion:
I do not understand the hunting for nanoseconds in general here, if you make Music.
Why? Because this is not a "Real-Life"Usability Test. Its more like the gap between the consumption values of fuel by car manufacturers (which may be correct) and the real life on street.
   
Longer time ago, i made 2 scenarios, just for myself to see if the physics, once i learned are really true, lol...

1) Looping Back via a physical Cable, gives a quite more realistic Test. As it involves the DA/AD processing by the Unit (described in the RTL Manual here: https://oblique-audio.com/downloads/RtlUserGuide.pdf) - You will see, there are a few more milli/nanoseconds than via digital loopback. But i understood the sense of the loopback Button in TM. Anyhow - But still not a real life Example...

2) So I also made the RTL Measurement with a professional Kondensator Microphone, connected to IN of my FF, and the Monitor Speakers to OUT of FF. The distance between Mic and Monitors is about 4 meters in my case. As we all know: Speed of Sound (c) -> c = 343 meters/s - lets say ~ 3 meters/millisecond...
RTLs Results confirmed this absolutely. I had more than 20ms (if i remember right)...

-> So: If you just play a Synth, and we ignore all the equipment related Latencies, but your Speakers are 4 meters away from your Ears - calculate it yourself.

A other quite good explanation can be found here: https://www.logicprohelp.com/forums/top … ent-906870

Hope, it helps a little.
regards, meg33