Madiface USB plus ADI 9632 did this.

Unfortunately the ADI-6432 is legacy product.

Thanks for the info. Two Sonnet expansion chassis is the last resort option but I'm trying to avoid having a mess of extra stuff involved and potential bottlenecks or additional points of failure compared to the amazing Mac Pro M2 Ultra with all the PCIe slots built right in. Clean, fast, reliable, only power cable (for the computer only) vs. 3 or more power cables when you factor in the chassis and docks etc.

it's not about cost, it's about aesthetics, speed, reliability, mess, etc.

Here a proposal, check with a RME reseller/partner (e.g. Synthax) whether would be feasible
and whether the mentioned Appsys products would be a feasible solution just in case you do not get ADI-6432 on used market.

Thunderbolt based Sonnet expansion chassis with HDSPe MADI FX card.
This gives you three MADI buses (2x optical, 1x coaxial).

Option A) RME ADI-6432 from the used market

If it is ok for you to buy from the used market, I saw on eBay an ADI-6432 with a 3-year warranty.

Option B) Appsys FLX-AES3 (with AUX-MADI-SFP) based, check with RME or your reseller

The only alternative to the ADI-6432 that I found in the internet is from Appsys
    https://appsys.ch/products/flexiverters/flx-aes3
in combination with this MADI module
    https://appsys.ch/?option=com_hikashop& … y_pathway=

The FLX-AES3 delivers 16x16 channels of AES/EBU = 8x AES I/O like the HDSPe AES.

Ideal would be IMHO the following Setup

Mac---USB---optional: ARC USB
|
| Thunderbolt
|
Sonnet Chassis        +---AES---optional: ADI-2 Pro FS R BE or ADI-2/4 Pro SE
     +                      /
     + HDSPe MADI FX---MADI Bus 1----[OM3/4]----Appsys AUX-MADI-SFP Module---Appsys FLX-AES3 (8x AES I/O, 16ch)
     + HDSPe MADI FX---MADI Bus 2----[OM3/4]----Appsys AUX-MADI-SFP Module---Appsys FLX-AES3 (8x AES I/O, 16ch)
     + HDSPe MADI FX---MADI Bus 3----[OM3/4]----Appsys AUX-MADI-SFP Module---Appsys FLX-AES3 (8x AES I/O, 16ch)
                                               SC-Plug                LC-plug

You need to order the optional RME HDSPe OPTO-X module for the HDSPe MADI FX.
This replaces the daughter card to get a 3rd optical MADI port.

MADI gives you the advantage of having an extremely long multicore.
When using multimode fiber according to OM3 or OM4 standard, the cable length can be up to 2 km.
MADI gives you also the opportunity to serial chain devices on one MADI bus, then the cable length can still be up to 2 km between each of the devices in a chain.

Note: When multiple devices are connected serially, the MADI I/O of each OctaMic XTC causes a delay of 3 samples at single speed (6 samples at double speed). To avoid this very little delay I propose to connect one Appsys FLX-AES3 per MADI bus. You have three MADI buses anyway, so why not use them ...

This solution would have the advantage, that you have one MADI card with 3 MADI busses.
and you can route everything directly on that card with the lowest latency.
You former solution with the three HDSPe AES had the little drawback, that you had to route through the DAW with much higher latency and depending on the chosen buffersize, if you wanted to route an audio flow from an input of one card to the output of another card. With the HDSPe MADI FX you can route any input to any output as everything is on the one card.

Further advantages of the HDSPe MADI FX
- AES I/O e.g. for connecting one of RMEs reference converter for monitoring (ADI-2 Pro FS R BE or ADI-2/4 Pro SE)
- Full implementation of FX chip
- RoomEQ and Crossfeed
- Headphone output (1x analog I/O, unbalanced)

Also nice for cabling. In case you have multiple rooms or racks, you can now place the Appsys FLX-AES3 close to the location where the AES devices need to be connected. Then you have one long and "slim" optical multicore (OM4 multimode cable) and can keep the much stiffer and more expensive AES cables short.

The list of devices
- Sonnet Thunderbolt Chassis, best one Supporting Thunderbolt 5
- HDSPe MADI FX
- HDSPe OPTO-X
- 3x optical cable; OM4 Duplex patch cable, with SC to LC plug
- 3x Appsys FLX-AES3
- 3x Appsys AUX-MADI-SFP Module
- Note: its unclear whether power supplies are included with the Appsys FLX-AES3, check with your seller.

Optional:
- ARC USB connected to your Apple computer to remote control TM FX control room
- RME Reference Converter for monitoring through active monitors and headphones

Regarding the Sonnet Thunderbolt chassis: As the new Mac Studio supports Thunderbolt 5, I would take a look at the new Sonnet chassis, which also supports Thunderbolt 5. These two might fit:

ECHO-EX-SE1-T5 - Echo SE I T5 Thunderbolt 5 Expansion System
    - Two Thunderbolt ports
    -  One x8 mechanical (x4 electrical) PCIe 4.0 [Thunderbolt 5 spec]
    -  One half-length+ (maximum 7.75 inches long), single-width, full-height x8 PCIe card and one daughter card
    - 3.0 x 11.13 x 5.75 in. (8.5 x 28.3 x 14.6 cm)
    - 1.13 kg
    - 0.8- meter Thunderbolt 5 cable, ThunderLok-A connector lock
    - NOTE: Check with RME whether the length of the HDSPe MADI FX card fits into the case

ECHO-2DDV-T5 - Echo II DV T5 Desktop System
    - 2U Thunderbolt desktop enclosure
    - Four Thunderbolt ports
    - Supported cables: Thunderbolt 5 or Thunderbolt 4
    - Two x16 mechanical (x4 electrical) PCIe 4.0
    - Two full-length, full-height, single-width PCIe cards
    -  3.5 x 16.7 x 9 in. (8.9 x 42.4 x 22.9 cm)
    -  9.7 lbs (4.40 kg)
    -  Two 0.8-meter Thunderbolt 5 cables
    -  Three ThunderLok-A connector locks

I also see a rack-mount version, but this might be overkill.

Thanks for all the info, I'll have to digest it more later when I'm not in the middle of a large project with a tight deadline.

I should clarify though...

The 3 AES cards are all doing different jobs and not working together.

AES card #1 with the daughter card is used for my analog mastering processing chain, where I need a a few different DA and AD options, plus monitoring, metering, etc. and all my devices use AES in/out. It's clocked externally via my Crane Song HEDD or Lavry SAVITR.

AES card #2 is totally separate and I use it just for working in WaveLab "in the box". In theory, I just need two separate stereo outputs...one for main monitoring and one for reference monitoring. I don't need all the I/O that the card has but it works well and is in my Mac Pro for the two stereo AES outputs I need it for. I use the internal clock since it's basically just used for listening to the stereo output of WaveLab and the alt 3/4 output for reference.

AES card #3 is also separate and I only use it for 1 stereo INPUT for my metering software. It's an older AES card version I had laying around so while it's overkill in terms of I/O, it works amazingly well. The Digiface AES may be an option but I haven't gotten a clear answer from RME as to whether it can adapt to incoming sample rate changes over AES like my AES card(s) can...something that RME somewhat recently added at my request thanks to the great support from MC at RME. As I listen to different sample rates on my Grace m908, the digital meter output adjusts to match that sample rate and both the AES card and my metering software flawlessly adapt to these changes automatically. Again, I'm not sure if the Digiface AES can do this or if it's even the best option to simple get an AES signal into my metering software on my Mac.

My AIO Pro card is used exclusively for listening to Qobuz as my Grace m908 monitor controller has many AES inputs.

So really, what I need is a lot of AES I/O but not working as one big unit. This is why I have the 3 AES cards and an AIO Pro card in my Mac Pro and while some these things can maybe be converted into USB devices and others put in a Sonnet chassis, I'm trying to minimize external things and points of failure, power connections, etc.

I have no experience with MADI so I need to see what those offer over what I current use but right now I do not understand MADI.

None of my hardware devices use MADI so ultimately, I'm just looking for a lot of AES connections over XLR to get in and out of my existing equipment which is easily doable with the Mac Pro and my RME cards but moving forward, so changes are going to need to happen on my end.

It would be so great if RME made a dedicated box with many AES I/O with Thunderbolt 5 for super fast and reliable speed.

The Mac Pro will last for a long time technically speaking but the offline rendering speeds in iZotope RX and WaveLab are already becoming a slow down compared to my Mac Studio M4 Max and MacBook Air M5 so when the Mac Studio M5x comes out I plan to replace the Mac Pro with the Mac Studio M5 and I'm trying to make a plan for how to keep using all the stuff I'm using right now inside the Mac Pro with as little external boxes as possible.

I'm not trying to save money or be philosophical about it. I'm just trying to get the best speed and performance with as little clutter and messing around as the Mac Pro offers which is zero clutter, zero external boxes, and incredible bandwidth/speed/performance when it comes to my AES I/O needs.

Just to have this more clear,  you don't mean different sample rates at the same time ?

Page 24 in the user manual :

The Digiface AES utilizes a very user-friendly, intelligent clock control, called AutoSync. In Auto-
Sync mode, the system constantly scans the digital input for a valid signal. If any valid signal is
found, the Digiface switches from the internal quartz (Clock Source – Current Internal) to a clock
extracted from the input signal (Clock Source – Current AES, SPDIF, ADAT). The difference to a
usual slave mode is that whenever the clock reference fails, the system will automatically use its
internal clock and operate in clock mode Master.

That means YES it does adapt to outer sample rates.

Yes indeed, and also an output.

Don't you use Digicheck for metering ?

Digicheck NG is not available as plugin, it is standalone software with connection to RME audio channels.
you can select any input channel or software playback channel.
For example the output of WaveLab, very comfortable on an extra display between the loudspeakers.

Your demand is to have less clutter and less equipment.

your metering software uses an input signal, there is no need for a input device I guess.

Totalmix Loopback brings your monitoring signal to an input signal you can select for metering in MIRA or Digicheck.

Skip out one of the cards...

This section describes basic sync mechanisms, not how the card behaves when connected to macOS via USB driver. Or Windows.

But the answer here is indeed yes, the system sample rate (Core Audio) for the DF USB will change when the hardware input sample rate changes (Follow Clock feature).

To switch the complete OS has the most impact on a setup and would be a real pain. So I understand his demand not wanting having to switch from Apple to a Windows PC to get PCIe sockets.
Another issue is that there are only a few PC mainboards available with a good socket design.
The mainboards are usually filled with up to four M.2 sockets for storage which is a waste of PCIe lanes and sockets.
You would need special workstation or server mainboards with better socket design, but those are very special
and bring further problems as then you need to look at Threadripper or Intel Xeon based platforms.

It would take weeks and months before he would be familiar with the setup and operation of the computer.
Everything would have to be installed from scratch.
This will need serious time and effort over weeks and probably months and a parallel installation, and you would have to switch the AES-connected end devices back and forth between those two solutions.

But I see it the same as you; it is rather unlikely that RME will introduce such a Thunderbolt-based unit with 8-16x AES I/O.
At best, RME would reactivate the ADI-6432 or ADI-642, but which would again be ... MADI based.
So a Thunderbolt dock with PCIe sockets and a HDSPe MADI Fx would be the basement anyway.

I assumed that he might need a solution "just right now" or for the foreseeable future... Thats why I made the proposal to him.
And the best possibility is to use a proven MADI-based design without any additional complexity based on the proven HDSPe MADI FX card.

To answer his question, "What is MADI" .. To keep it simple, similar to ADAT/TOSLINK, only much much better.
1. Instead of 8/4/2 channels at single/double/quad speed you get 64/32/16 channels per MADI bus
2. It uses standard fiber cables from the network area. Usually multimode (OM3 or OM4).
3. Unlike ADAT you can even chain multiple MADI devices on a MADI bus, e.g. eight 8-ch preamps at single speed (8x8=64ch).
4. Unlike ADAT which has a max cable length of 10-15 m, the maximum cable length for MADI is 2 km.
    Even 10 km when using single mode (needs special purchase order for older devices without modular transceiver (SFP)).
5. Allows you to remote control preamps by "MIDI over MADI". Then RME uses unused bits in the MADI protocol header for
    the transmission of remote control commands to preamps which otherwise would require MIDI cabling.
    So audio and remote control over one standard fiber link.
6. For a local installation, you can use standard patch cables from the networking area, which are quite inexpensive and
    usually available in length of 0.5 - 50m. Some shops also offer custom lengths.

Two examples from my setup, how nicely the units can be connected through MADI cabling.
Here two MADI devices in series: RME UFX+ --- RME Octamic XTC --- RME 12Mic --- back to UFX+
https://www.tonstudio-forum.de/file-download/1834/

Here with the proposed HDSPe MADI FX card, which has three MADI buses, so every devices has its own MADI bus
- MADI Bus 1 (ch 1..64) for the M-1620 Pro
- MADI Bus 2 (ch 65..127) for the 12Mic
- MADI Bus 3 (ch 128-192) optional: for the UFX III as DURec recorder
https://www.tonstudio-forum.de/file-download/1986/

Why did I propose this solution to him? Because I do not know of any existing solution for him that looks as he would wish
and it is rather unlikely that RME will build such a TB based box with 4, 8 or 16x AES I/O.

My proposal comes closest to the PCIe based setup with three HDSP AES cards, that he has now.
Because Thunderbolt is "external PCIe".

With the Sonnet case he gets the PCIe sockets back, well connected through a Thunderbolt ("external PCIe") based connection.

From there, rock solid MADI standard components: HDSPe MADI FX, offering 3 MADI buses (3x64 channels).

Three MADI buses he will need when using higher sample rates, because also for MADI sample multiplexing takes place for double and quad speed. So for double speed he has only 3x32ch, and for quad speed only 3x16ch.

It boils down to the point: either to think about my proposal or to wait for a feature request that most likely won't come, which means no solution at all.

PC/Windows is not an option. I'll stop mastering and deliver packages for Amazon before I move to a PC. PC/Windows itself is much more of a headache than staying on Mac and having multiples external boxes and docks connected to a Mac Studio to get back what I have now with the Mac Pro.

I think I missed one detail from your 1st post. You said 3 HDSPe AES cards, but only the last used the daughter card.
Theoretically you would be fine with a Sonnet Echo Express SE IIIe, which supports 3 PCIe cards and one daughter card.
You only need to check with RME/Sonnet whether the chassis is long enough.
Then you have the same state as now, three independent cards, and your only investment is the Sonnet case.

But sorry, I thought that all three HDSPe AES cards were fully equipped, but this is not the case.

To MADI, this is a solid, professional solution. Google a bit about that.
It also has the benefit of being a solution dedicated to your recording equipment, without any additional requirements compared to AVB or Dante. And if you change sample rates often, then I would either stick to the three HDSPe AES cards or to a MADI solution. Simple and robust.

Besides the amount of AES I/O and a little scalability in mind, the other reason why I mentioned the MADI solution with the HDSPe MADI FX card and additional MADI AES equipment is because then you can directly route through all AES ports "any to any".

If you do not need this or other features like Room EQ, well, then stay with the old setup to have three independent working cards where you would have to route through the DAW in case you need to route from one to the other card.
It will definitively save you some money.

If you only need 4x 4 AES I/O then you could also use only two of the three HDSPe AES cards
and use both with their daughter cards, then you can keep one card as a spare, and you only need two PCIe sockets in the Thunderbolt case but in total 4 slots to house the PCIe cards with their daughter cards.

An HDSPe MADI FX-based MADI solution will definitively be more expensive due to the additional MADI equipment.

If you need 4x 4 AES I/O then this is in total 16 AES Ports with 32 channels.
If you need to work up to quad speed, this means you would need at least two MADI buses.

MADI Bus #1 @quad speed: ch 1..16
MADI Bus #2 @quad speed: ch 17..32

You could achieve this setup with

HDSPe MADI FX

   + MADI bus 1 (ch 1..16)   + ADI-642#1 Ch 1..8
                                          + ADI-642#2 Ch 9..16

   + MADI bus 2 (ch 17..32)  + ADI-642#3 Ch 17..24
                                          + ADI-642#4 Ch 25..32

For example four ADI-642 from used market. 2 serially chained at each of the two MADI buses.

Or you need to check with 3rd party components what combination is best to get 16x AES I/O (32ch) @quad speed.

The best thing is to check with Synthax or any other company of your choice that can consult you about what's possible and to stay within budget (which you didn't mention).