Topic: HDSP Aio Pro or Babyface Pro?

Frohes Neues an alle!

Es steht ein neues Audiointerface für meinen 2019 Apple MacPro an, und ich schwanke zwischen HDSP AIO Pro und Babyface Pro FS.

FRAGE:
Ist die Latenz der HDSP AIO wesentlich besser als die vom Babyface?
Ich spiele Gitarre darüber, also ist jede ms ein Gewinn.

Mit ist klar, dass ich für die HDSP noch einen preamp brauche, aber das macht nichts.

Meinungen?

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

Re: HDSP Aio Pro or Babyface Pro?

Under high cpu load a pie card could perform better as usb. Under low load you can probably use the same buffer size. So it depends if you are just playing guitar or playing guitar in a complex arrangement in a DAW, with high cpu loads.

Vincent, Amsterdam
https://soundcloud.com/thesecretworld
Babyface pro fs, HDSP9652+ADI-8AE, HDSP9632

Re: HDSP Aio Pro or Babyface Pro?

Thank you - I record guitars pretty early in the arrangement, so probably no high cpu loads. Thank you!

I just saw somewhere that you can set the AIO to 16 buffer size, which would be INSANE.

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

4 (edited by ramses 2021-01-09 13:06:26)

Re: HDSP Aio Pro or Babyface Pro?

Der Unterschied bei beiden Produkten ist nicht groß, ich würde mich da mehr auf die Features des Recording Interfaces konzentrieren, was Dir für Deine Anwendungen mehr zusagt.

Beide Produkte wurden kürzlich aktualisiert und haben aktuelle und sehr schnelle A/D und D/A Wandler verbaut, das kannst Du in den Handbüchern unter "Latenz und Monitoring" (Kapitel 27.2 bzw 28.2) nachlesen. 5/7 samples Latenz bei AD/DA und "single speed" ist schon echt klasse.

Einen größeren Einfluß auf die Latenz als Wandler Performance hat die Latenz, die sich durch den Transport von Audio über  Fw/USB/PCIe/TB und zurück ergibt, die sogenannte RTL (Round Trip Latency). Die RTL wird bei höheren Sample Rates geringfügig besser, aber bei Windows im Wesentlichen durch die von Dir gewählte Größe der ASIO Buffersize bestimmt.

Eins noch zur Sicherheit: die Angabe der ASIO Buffersize im weiteren Text bezieht sich auf single speed (44.1/48). Bei Double Speed (88.2/96 kHz) ist die ASIO Buffersize automatisch doppelt so groß, bei doppelter Datenmenge braucht man natürlich doppelt so große Buffer. Wobei die RTL (und auch die Wandlerlatenz bei Double Speed) bei höheren Samplerates insgesamt etwas geringer ausfällt.

Ich spiele auch Gitarre und habe das mal mit einem Gitarren VSTi von Kuasse ausprobiert.
Dabei konnte ich noch mit ASIO Buffersizes bis 256 (also RTL bis 13,2 ms) problemlos einspielen.
Allerdings ist das auch das Maximum, bei Verwendung einer ASIO Buffersize von 128 samples (RTL 7,4ms) wirkt sich das beim Einspielen doch etwas "tighter" an.

Daraus würde ich die Faustformel ableiten wollen, dass man beim Einspielen unter Verwendung Virtueller Instrumente darauf achten sollte, dass die RTL möglichst unter 10ms liegen sollte und rund 13,2ms nicht überschreiten sollte.
Was ich noch nicht untersucht habe: es könnte hier noch mit reinspielen, wie hoch die Latenz des VSTi selbst ist, denn manche VST/VSTi sind ja rechenintensiver, als andere, da könnte sich dann auch eine andere Latenz durch das VSTi selbst ergeben.
Bei einer ASIO Buffersize von 512 Samples (24,8ms, bei Single Speed) ist ein Einspielen mit Mithören was Du da spielst unmöglich, es kommt rapide zu einer Verzögerung Deines Tempos auf 0. Das Ohr will dann immer erst hören, was Du da gespielt hast und wenn das eine bestimmte Latenz überschreitet, funktioniert das ab einer gewissen Verzögerung einfach nicht mehr.

Was bedeutet das für die Produktauswahl ?
Nimm einfach das Interface, das Dir am besten gefällt.
Alle RME Recording Interfaces werden bei einer ASIO buffersize von 128 samples eine RTL von unter 10ms liefern.

Rein theoretisch sollte ein PCIe Treiber gegenüber einem USB Treiber effizienter arbeiten und unter Last vielleicht auch besser performen, was auch immer man darunter verstehen möchte. De facto habe ich beim Vergleich zwischen UFX+ (USB3) und RayDAT (PCIe) keine größeren Unterschiede feststellen können hinsichtlich "knackseranfälligkeit" oder höherer CPU Last. Kannst ja mal hier meinen Test mit einem "fingierten" großen Projekt mit 400 Spuren und 800+ VSTs sehen...
https://www.tonstudio-forum.de/blog/ind … cks-de-en/

Ich glaube, dass alle RME Treiber einfach sehr gut geschrieben sind, vergleichbar hervorragende Performance liefern und man sich wirklich mehr auf die Produktfeatures konzentrieren sollte.

Hier ein Vergleich der RTL zwischen verschiedenen RME Produkten die ich im Einsatz hatte. Dabei habe ich anhand der Angaben im RME Handbuch berechnet, wie es ausschaut, wenn die AD/DA Wandler hinter ADAT oder MADI stecken. Insgesamt ein sehr positives Bild für alle RME Produkte:
https://www.tonstudio-forum.de/index.php/Attachment/2343-UFX-UFX-RayDAT-Latencies-v2-jpg/

An der Stelle "Zehntel ms" zu zählen würde ich erst dann, wenn vielleicht auf Rechnerseite defizite bestehen durch:
a) VSTi hat selber eine sehr hohe Laten
b) Nicht ausreichende CPU Gesamt-Performance auf Rechnerseite, insbesondere Defizite bei der Single Thread Performance, weil der DAW Track mit dem VSTi und ggf noch weiteren Inserts durch einen CPU thread berechnet wird.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

Re: HDSP Aio Pro or Babyface Pro?

berndk wrote:

Thank you - I record guitars pretty early in the arrangement, so probably no high cpu loads. Thank you!
I just saw somewhere that you can set the AIO to 16 buffer size, which would be INSANE.

The HDSPe driver allows AFAIK down to 32 samples (at least what I remember from RayDAT which uses the same ASIO driver).
The former USB driver allows down to 48 samples.
The new MADIface USB driver also allows down to 32 samples.

16 samples ASIO buffer size sounds for me to be too risky.

It maximizes the stress for your computer to be able to process audio in time. There is not much time left for your PC to process audio in time, if the portions of audio (ASIO buffersize) in which the transport of audio samples to/from the digital interface happens, becomes smaller and smaller.

You can check this easily by taking a bigger DAW project.
Now play it back using different ASIO buffersizes.
The smaller the ASIO buffersize is, the more the stress increases for your system:
A) CPU load
B) e.g. ASIO load in Cubase

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

6 (edited by berndk 2021-01-10 09:45:52)

Re: HDSP Aio Pro or Babyface Pro?

Ramses, thank you so much for your detailed reply.

I saw this chart before and looked for it EVERYWHERE smile

I now decided for the Babyface Pro for the moment. Reason beeing that the Analog expansion Boards for the AIO are seemingly discontinued (read a post from MC about it).

As to latency: I mostly work with real tube amps and AxeFx3. Axe has a RTL of 2ms. I can clearly feel anything above 7ms, and with vstis having latency of their own, a RTL of 3-4ms would be perfect. But it seems at 64 buffer size, this should be real with the new babyface pro fs.

I'll let you know my findings when it arrives next week!

Thanks again very much.

EDIT: Ich merke gerade, hätte auch auf Deutsch schreiben können

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

7 (edited by ramses 2021-01-10 09:54:06)

Re: HDSP Aio Pro or Babyface Pro?

> Reason beeing that the Analog expansion Boards for the AIO are seemingly discontinued (read a post from MC about it).

URL bitte. Das kann ich mir eigentlich nicht vorstellen. Da ging es vermutlich um die alten 8-Kanal boards, die es schon länger nicht mehr zu kaufen gibt.

Von einer Abkündigung kann ich auf der RME Produktseite nichts erkennen: https://www.rme-audio.de/de_expansion-boards.html

Die Expansions-Board sind auch bei allen größeren Shops in Deutschland sofort lieferbar: Thomann, Musicstore, DA-X

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

8

Re: HDSP Aio Pro or Babyface Pro?

Noch. Da die neue AIO Pro die nicht mehr unterstützt werden die irgendwann ausgehen...

Regards
Matthias Carstens
RME

Re: HDSP Aio Pro or Babyface Pro?

Ok, dankeschön für die Klarstellung.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

Re: HDSP Aio Pro or Babyface Pro?

Also ich habe jetzt mal die Handbücher der AIO und des Babyface pro fs verglichen -
der Unterschied der Wandler ist wohl:
- AD Wanlder: Gleiche Buffersize vom 5
- DA Wanlder: AIO 5, Babyface 7

Das ergibt bei meiner genutzten Smaplerate von 48khz einen Unterschied von 0,02ms. Hmm... ich denke das ist echt völlig egal! smile

Danke für Eure posts,
es wird definitiv Babyface Pro FS.

Und wenn sie rauskommen sollte UCX mit thunderbolt. Um ein paar mehr ein und Ausgänge zu haben und damit ich für spdif keinen konverter mehr brauche.

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

Re: HDSP Aio Pro or Babyface Pro?

> UCX mit thunderbolt

Kann mir nicht vorstellen, dass es sowas geben wird. Design-/Produktphilosophie von RME ist, Preise nicht unnötig in die Höhe zu treiben. Für ein 30 Kanal Interface brauchst Du kein Thunderbolt. RME schafft über USB2 bis zu 68 I/O ports (68 IN und OUT), siehe beispielsweise MADIface Pro oder 64 I/O ports beim MADIface USB.

Thunderbolt ist immer noch nicht so verbreitet, die Kabel sind wesentlich teurer, die Stecker filigraner und die maxmale Kabellänge ist im Vergleich zu USB2 (5m) auf 2m begrenzt. Wohingegen USB2 an jedem PC zu finden ist und USB3 ports vollständig abwärtskompatibel sind.

Wenn Du oben nochmal in der Tabelle schaust, dann beträgt beispielsweise beim UFX+ der Unterschied zwischen USB3 und thunderbolt gerade mal 0,449ms. Der Unterschied ist so gering, dass die Nachteile nicht aufwiegt, mehr ausgeben zu müssen und dafür auch noch die Auswahl an Rechnerhardware einzuschränken.

Andere Hersteller mögen den Schritt zu Thunderbolt basierten Recording interfaces mit sehr wenigen Ports aus anderen Gründen gegangen sein. Möglicherweise, weil sie sonst nicht die geringen RTL oder auch Treiberstabilität hinbekämen.
Immerhin ist Thunderbolt externes PCIe und entbindet den Hersteller davon auf Rechner und PC Seite mit 3rd party Kommunikationsbausteinen zu tun zu haben, die hin und wieder auch mal ihre bugs haben und diese auch noch stabil und performant ans Performen zu bekommen.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

12 (edited by berndk 2021-01-11 11:23:45)

Re: HDSP Aio Pro or Babyface Pro?

Hallo Ramses,

ich halte es - leider - auch für sehr unwahrscheinlich, aber TB hat ja außer der 0,5 ms besseren Latenz noch den Vorteil, dass es evtl. das ganze UCX mit Buspower versorgen könnte - speziell für Laptops interessant - und dass man für Apple Laptops keinen Adapter mehr braucht. Das wäre schon fein.

Aber ich fürchte, Du hast recht, und RME wird da den "vernünftigen" Weg gehen.

Es wäre aber schon genug, wenn es das "UCX FS" geben würde, mit den neuen Wandlern. Im UCX sind ja noch die vom ersten UFX.
Laut Deiner Tabelle von UFX Usb der Unterschied zu UFX+ Thunderbolt sind bei 64 Samples schon Satte 2,5ms. Das finde ich extrem.
Vielleicht mit USB-C stecker? Haupsache der Coax SPDIF bleibt.

Würde es dann wohl sofort kaufen wink

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

13 (edited by ramses 2021-01-11 11:43:47)

Re: HDSP Aio Pro or Babyface Pro?

Mit diesem USB-C Stecker stehe ich persönlich auf Kriegsfuß. Kleinere Stecker halte ich für grundweg instabiler und weniger haltbar. Insofern hoffe ich auf USB Stecker, die auch bisher verbaut waren.

Von wegen Stromversorgung, meinst Du das reicht bei Thunderbolt ?

Ich habe zu TB 3+4  auf Wikipedia folgendes gefunden:
- bis zu einer Leistung von 100 Watt
- 15 Watt für vom Bus versorgte Geräte

UCX Manual: Strombedarf bei 12 Volt Betriebsspannung: 910 mA (11 Watt)

Bin jetzt aus der Zeit .. meine beim FF400 hieße es, dass eine Versorgung über TB nicht ginge.
Oder verbraucht das ältere FF400 mehr als das UCX ??

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

Re: HDSP Aio Pro or Babyface Pro?

Also das scheint mit 11 Watt echt etwas knapp zu sein.

Gerade nachgeschaut: Fireface 400 - Buspowered über FW. Die UCX leider nicht, auch nicht über FW sad

Trotzdem wäre es echt super - ob jetzt über TB oder USB3 / C ist eigentlich egal.

USB-C ist mit guten Kabeln eigentlich kein problem. Wobei die schon den Stecker hätten stabiler machen können. Oder vielmehr: Di Buchsen.

Lassen wir uns überraschen, ob es eine UCX+ oder UCX FS gibt. Ich fänd es super!

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

Re: HDSP Aio Pro or Babyface Pro?

I believe it is the TB to firewire adapter that doesn't transfer the full power., not that the TB power is not enough.

Vincent, Amsterdam
https://soundcloud.com/thesecretworld
Babyface pro fs, HDSP9652+ADI-8AE, HDSP9632

Re: HDSP Aio Pro or Babyface Pro?

vinark wrote:

I believe it is the TB to firewire adapter that doesn't transfer the full power., not that the TB power is not enough.

That is definately true, the adapter has no power, like the tb2 to tb3 adapter, but the RME Manual states that the UCX is not running with buspower.

I mean since firewire is performing worse than usb on latency, I would never use it.

With TB3 though, I think it would be possible to drive it with buspower. Which would be great, but I doubt they will do it.

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

17 (edited by ramses 2021-01-12 19:31:41)

Re: HDSP Aio Pro or Babyface Pro?

> I mean since firewire is performing worse than usb on latency, I would never use it.

Are you sure ? I see no big difference for e.g. an UFX .. around 0.5ms is a little nothing ...

https://www.tonstudio-forum.de/index.php/Attachment/2343-UFX-UFX-RayDAT-Latencies-v2-jpg/

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

18 (edited by berndk 2021-02-12 22:52:39)

Re: HDSP Aio Pro or Babyface Pro?

I just could not leave it alone and ordered a HDSP AIO Pro.

Guess what... Latency is WORSE (!!!) than with the Babyface Pro FS!
Not by much, but I was expecting better performance with a brandnew pcie card than with the babyface pro fs.

The Manual solves it: The Core Audio Buffer in the Babyface ist 16 Samples, in the AIO 24.

I could just not believe it!

Sending it back now. I am still in shock.

Here are the Numbers (I measured it with ableton live 10 on my macpro 2019)

At 48khz 32 Samples:
AIO Pro: 3,94 RT - Babyface: 3,58 RT

At 48khz 64 Samples:
AIO Pro: 4,60 RT - Babyface: 4,25 RT

...and it goes on like this. Both Devices work without any issues from 64 samples on, under 32 there is a little noise once in a while (1 short crackling).

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

19 (edited by ramses 2021-02-13 04:57:57)

Re: HDSP Aio Pro or Babyface Pro?

Sad that you didn't took values at 44.1 kHz, then you could even compare with my list.

And in terms of these differences, these are both excellent values and I do not know why you complain.
A delta of around 0,36ms is nothing which should worry you in any way.

I played myself through virtual amps (Kuassa) and could even play using an ASIO buffersize of 256 which is 13,22 RTL.
With 512 (24,8 ms RTL) you immediately slow down playing, clear.
But comforably I felt already with an ASIO buffersize of 128 ms = 7,4ms RTL.
I heard also from other sides, that a RTL under 10ms is being regarded as fine/sufficient.

You should see it from a different angle .. a low RTT is fine .. but .. if you loose audio this is definitively worse than debating about 0,3-0,4ms. And RME maybe had some reasons to increase it slightly.

And as soon as you have a few tracks together and use VST and VSTi, then a very low buffersize (for Windows this would be something around 32 or 64 buffers) would be anyway something that I wouldn't recommend.

The more tracks and VST/VSTi become added I would anyway recommend to choose higher buffersizes.
And the fine thing is that with RME you stay below 10ms RTL even with a ASIO buffersize of 128 which gives you already some "headroom" for your DAW project.

The other thing is .. with time critical stuff you can also start or freeze tracks so that lower settings can still be used a little longer in a project until it comes to the mixing and mastering phase.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

20 (edited by waedi 2021-02-13 07:36:11)

Re: HDSP Aio Pro or Babyface Pro?

I would go with the Babyface.
No latency with both system.
A latency can come from software when you use a soft amp and lots of plugins.
But the Mac and the RME interface wil not get to their limits.
The Babyface gives you much more freedom and flexibility
Recordings at Übungsraum, play live with soft amp.
You can record a new Riff and visit the bass player and he makes a bassline in his home while you empty his fridge.
The singer can scream his part directly into the Babyface where ever you want.
With the cheapest Macbook Air you have a music workstation in your hand.
You may also consider that software makers will no longer support pre M1 Macs sooner or later.

M1-Sonoma, Madiface Pro, Digiface USB, Babyface silver and blue

Re: HDSP Aio Pro or Babyface Pro?

ramses wrote:

Sad that you didn't took values at 44.1 kHz, then you could even compare with my list.

And in terms of these differences, these are both excellent values and I do not know why you complain.
A delta of around 0,36ms is nothing which should worry you in any way.

I played myself through virtual amps (Kuassa) and could even play using an ASIO buffersize of 256 which is 13,22 RTL.
With 512 (24,8 ms RTL) you immediately slow down playing, clear.
But comforably I felt already with an ASIO buffersize of 128 ms = 7,4ms RTL.
I heard also from other sides, that a RTL under 10ms is being regarded as fine/sufficient.

You should see it from a different angle .. a low RTT is fine .. but .. if you loose audio this is definitively worse than debating about 0,3-0,4ms. And RME maybe had some reasons to increase it slightly.

And as soon as you have a few tracks together and use VST and VSTi, then a very low buffersize (for Windows this would be something around 32 or 64 buffers) would be anyway something that I wouldn't recommend.

The more tracks and VST/VSTi become added I would anyway recommend to choose higher buffersizes.
And the fine thing is that with RME you stay below 10ms RTL even with a ASIO buffersize of 128 which gives you already some "headroom" for your DAW project.

The other thing is .. with time critical stuff you can also start or freeze tracks so that lower settings can still be used a little longer in a project until it comes to the mixing and mastering phase.

You cant compare the Ableton Values and I cant use your tool for measuring latency (PC only afaik), so you cant compare it that good anyway.

You misunderstand me - I am not complaining about RME for making the best interfaces in the world, I just don't get why a USB 2.0 Interface has better Latency than a brandnew PCIe card! I think that is insane. Not saying the AIO is bad in any way, but I just cant get my head around why it does not perform better than the babyface. Which of course also says what an amazing product the babyface is!

I am glad I tried it, though. So I know that PCIe ist dead - at least for me.

Waiting for a next generation UCX FS with the new Converters now so I can get rid of my old Fireface 400 which I use to convert Spdif to Optical for my AxeFx - and also for some additional input channels should I need them.

By the way the difference in sound when using the ff400 Instrument input compared to the Babyface is huge! RME has come a long way, and the ff400 was a very good product back in the day. The fact that i can still use it on my M1 MacBookAir - come on, what company does that?

RME Forever and always!

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

22 (edited by ramses 2021-02-13 14:59:16)

Re: HDSP Aio Pro or Babyface Pro?

I do not use a tool, I use the values that the RME ASIO driver reports to the DAW.
I think you also looked at the DAW to get "inbound + outbound latency = RTL".

And pls take whatever device you like most.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

23 (edited by berndk 2021-02-13 17:14:21)

Re: HDSP Aio Pro or Babyface Pro?

I can at least give you the values of the Babyface at 44,1 - really appreciate your comparison!
For the AIO pro you just have to add 8 samples with 1ns , because that is the only real difference between the 2.
At 44,1 this is about 0,2ms according to the technical Reference in the RME Manual. Althougt through my testing it is more like 0,35ms difference.

Here it is, Babyface Pro FS under MacOSx Catalina, Ableton Live.

32 smp: 3,90 ms
64 smp: 4,63 ms
128 smp: 7,53 ms
256 smp: 13,3 ms
512 smp: 24,9 ms
1024 smp: 48,2 ms
2048 smp: 94,6 ms

That is all ableton will give me.

For whatever reason, Logic audio is giving me better values for 32 samples - Who knows why! The other Values are the same. Strange.

32 smp: 3,2ms
64 smp: 4,6ms
128 smp: 7,5ms
256 smp: 13,3ms
512 smp: 24,9ms
1024 smp: 48,2ms

If you like, you are welcome to include that in your table!

And hey, I know this is nitpickin' at milliseconds. But I want to settle for the best and then stick with it smile

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

24 (edited by ramses 2021-02-13 19:46:16)

Re: HDSP Aio Pro or Babyface Pro?

I think this is one of the differnces between Windows with ASIO driver vs Mac OS Core Audio.

In Windows you have only the ASIO driver to access the HW.
In Core audio you have to go through Core Audio and if I remember right the application can also change some buffer settings.
In Windows all is being controlled by the ASIO driver and its ASIO buffer size settings.

I if I remember right then Windows ASIO RTL's are a little bit smaller compared to Core Audio. But I can't tell you anymore where I read it.

If you compare your number with e.g. the RayDAT then pls don't wonder, that RayDAT values are slightly better.
RayDAT is a full digital working card so the values for A/D and D/A conversion of any analog port is missing.
Therefore the other two colums with UFX connected as preamp or an XTC connected through MADI.
Then the values are more compareable.
Values for A/D and D/A you also get from RME manuals of converters and preamps.
The higher the sample rate the lower the additional latency for conversion.

BTW thanks for the offer, but I do not intend to add other interfaces to this list.
The intention of this list was to show RMEs driver quality (low latency) and the USB/FW/TB/PCIe based solutions are on par.
Not much of a difference. To demonstrate this, this list of interfaces that I owned is fully sufficient for my purpose.
And I want to limit this definitvely to Windows, as I have only Windows here and I am only using windows.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

Re: HDSP Aio Pro or Babyface Pro?

Hey! Just in case, I will post the values of the AIO Pro buffers measured by me on my PC. As you can see, on PC the latency of AIO Pro is less than on Mac.

32 smp: 2.517 ms
64 smp: 3.968 ms
128 smp: 6.871 ms
256 smp: 12.676 ms
512 smp: 24.286 ms
1024 smp: 47.506 ms
2048 smp: 93.946 ms

26 (edited by ramses 2021-02-13 19:45:29)

Re: HDSP Aio Pro or Babyface Pro?

Yeah this is want I meant, core audio has a little higher latency compared to ASIO.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13

27 (edited by berndk 2021-02-13 20:11:03)

Re: HDSP Aio Pro or Babyface Pro?

Damm it! That is quite a difference!
Good thing I spend 10 Grand on this machine... sigh...
That is almost 1,5ms better!!!!

@ramses, it would be really interesting so see the babyface pro fs and/or aio pro in your list. According to the Manuals, the AD/DA Conversion of the converters got much better.
From a FF400 to the Babyface, it is about 1,7ms better! And that is only the converters, not the safety buffer!

And also, I totally agree that under 7/8ms is feeling good with guitar, a little safety margin is very good when you want some EQ or compression plugin which gives you additional latency.

Damn it, I should have listened to my pal when he said "build a windows machine, f*** mac"...

Babyface Pro FS +FF 400 used as SPDIF Adat conv. and more inputs.
Waiting for the UCX FS smile

28 (edited by ramses 2021-02-13 20:46:28)

Re: HDSP Aio Pro or Babyface Pro?

> From a FF400 to the Babyface, it is about 1,7ms better! And that is only the converters, not the safety buffer!

From manual: FF400 (Firewire!):
AD 43,2 samples: 0.98ms  + DA 28,0 samples: 0,63ms = 1,61ms
Safety Buffer on playback side only: 64 samples (1,45ms)
Core Audios Safety offset: 64 samples (1,45 ms)

From manual: BBF Pro (USB2):
AD 12,6 samples: 0.28ms  + DA 7 samples: 0,16ms = 0,44ms
Safety Buffer on playback side only: 32 samples (0,73ms)
Core Audios Safety offset: 16 samples (0,36 ms)

From manual: BBF Pro FS (USB2 and new quick A/D converters)
AD 5 samples: 0.11ms  + DA 7 samples: 0,16ms = 0,27ms
Safety Buffer on playback side only: 32 samples (0,73ms)
Core Audios Safety offset: 16 samples (0,36 ms)

From manual: UFX+:
AD 12,6 samples: 0.28ms  + DA 7 samples: 0,16ms = 0,44ms
Safety Buffer on playback side only: 32 samples (0,73ms)
Core Audios Safety offset: 16 samples (0,36 ms)

If you do not route everything in the DAW, then the safety buffers do not matter much...

Example: You hear a backing track in your phones and are recording your vocals

You have a submix that audio from mic is directly routed to your phones output.
So the "near-zero" latency is only the sum of AD and DA conversion
and "a little nothing" for the routing in the routing matrix inside of the device.
And this is 1,51ms for old FF400, 0,44ms for BBF Pro (and UFX+), 0,27ms for BBF Pro FS.

For recording a band through Mics and analog ins the RTL is also not important.

RTL is only important if you play through a virtual instrument on DAW.
The time from e.g. playing the string of your guitar which is connected to the INSTR input of recording interface,
A/D conversion, transfer of audio through USB, processing in DAW/VSTi, way back through USB,
D/A conversion (until you hear sound from VSTi in phones).

And if this is below 12ms, this is not critical. Best under 10ms.
This means you can use nicely an ASIO buffer size of 128 (and even 256) on UFX.
Of course you can try whether 32 or 64 is also feasible.

Similar story on Mac, with buffers around 128 samples you seem also to stay below 10ms and this is fine.
And if you do not play through a VSTi, then the same is applicable as I told for windows, then you have more or less only the converter latency for AD and DA which  much smaller compared to any transfer through USB/Firewire/Thunderbolt/PCIe (back and forth). Especially with more modern converters with smaller conversion time.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub13