Hey everyone!

I'm in the process of setting myself up to mix in one location and record in an overdub studio in another location.

I currently have a 12 x 16 Digilink Burl Mothership. After much deliberation and looking at every product I could find that would allow me to record in the overdub room with Pro Tools vanilla and other DAWS, I believe I've come to the conclusion that a Burl B16 with either their USB or Dante cards would be the way to go. Since I can bring my Ship to other studios for high track count recording, I thought it would be pretty logical to remain with the Burl sound at the overdub room and just slide some of my cards into a B16. Albeit only 4 inputs (can't afford the new 8 input card and don't like that it's a different sound from the 4 input card).

My question to the community is, what motherboard for the B16? USB or Dante? I need to work with Pro Tools vanilla and Studio One. A marked advantage of the Dante is I can take that card from the B16 and slide it into the B80 to eventually move away from Pro Tools HD if I want too (it's starting to feel like a jail cell..).

Thoughts, ideas, things I've not thought about? How would you implement Dante if you were me? Or just keep it simple with USB?

Thanks !

I guess what I'm asking is can Pro Tools HD Native software work with a Dante converter without being connected to any PTHD Native thunderbolt box? And can any other DAW do the same? Or does one NEED to have a Rednet type box that goes from Dante to Digilink?

I'm doing a lot of research but can't seem to find a definitive answer. Lots of conflicting information out there.

is your pro tools software asio compliant? (as in can you take an asio interface and use it with your software?)

The only real reason to use digilink now these days if it you wanted to substitute the box. On older versions, the logic behind it was the digital mixer and tdm plugs used the dsp hardware on the cards. Now these days its the same as everyone else (running inside the computer). I would use the dante because its a expandable archetecture that is not track or hardware limited like pro tools and Usb has issues with some computers, as far as noise floor performance.

a lot of people are waiting for dante b80's. They should have made more dante and dump the digilink since protools is slowly on the decline (thank god. Finally Karma is catching up to them as thousands of unhappy PT users that waste money only for them to chase the hype and upgrade game).

Thanks for the reply!

Dante Burl B80s already do exist!

Well I can open PTHD and use it on it's own using the Mac as a sound card. I also once was able to work it with a UAD Apollo for a few minutes before it began crashing constantly.

Yep. I agree with what you say, I believe Dante might be the future of Audio (Burl has Dante on their Bombers now) and want to get away from PTHD gradually. Not to mention that Burl USB is limited to 16 tracks. If I got the Dante card I think in the future it could go into a B80.

But I'm having a hard time wrapping my mind around the relationship of Dante to Pro Tools HD and Pro Tools vanilla, as well as some basic concerns like is dante monitoring latency acceptable with the virtual sound card or should one invest in a Rednet PCIe card and thunderbolt chassis. So many variables!

You can plug any Dante device directly into any gig network port on your computer. That plus Dante Virtual Soundcard is all you need for up to 64 channels in and out. The limitation is the driver that Dante Virtual Soundcard simulates - if it's a windows machine it sees Virtual Soundcard as an ASIO or WDM driver, and with a mac it's a Core Audio driver. The latency of Dante hardware devices talking directly to each other is the insanely fast times you see in advertisements and specs; the latency you get with Dante Virtual Soundcard is the same latency limitations you'd see with that type of driver used with any soundcard. Yamaha has a PCIe card that theoretically can get your latency down to about 0.5ms - I don't know if it's only for Yamaha hardware or not, but I can't imagine that it would be; you can connect any number of different brands of equipment to the same Dante network, and if the card is working on that same level I don't see how it could differentiate between Yamaha devices and other devices, as they're all speaking the same language. But I do not know.

For what it's worth: I just opened Dante Virtual Soundcard on my un-special 2012 macbook pro and I can get 4ms of latency. That number is completely hardware-agnostic - as long as the device(s) are Dante devices, that number should stay the same.

I'm pretty sure the Focusrite card only works with Focusrite Rednet devices. Though I'm not 100% sure on that either.

And re: future expandability and so forth: that's a big consideration in my book - Dante is pretty much futureproof at least to the next 5-10 years, by my estimation. And re: monitoring: the latency of Dante device into and back out of a computer isn't the best, but you can send the signal from one Dante device to another and get the Dante device-to-device latency for monitoring. You'd need an additional device that has the routing capability to work with a headphone amp and so forth, or you could get something like an Aviom system that's a monitoring system based around Dante. But taking your computer out of the equation removes the whole monitoring thing from your list of potential issues.

Thanks for the super informative posts, much appreciated ! Helps to understand Truly feels like the world is wide open when it comes to Dante!

It seems I'll be getting a B16 with Dante then. I'll test it out with the DVS and see how it feels. You point out that it's no worse than usual USB or Firewire based systems. I don't remember how those systems feel honestly, having been on Pro Tools HD for 3-4 years hehe. If the latency is not acceptable I'll look into PCIe card options! It seems that they can fit with a chassis to use with a laptop via Thunderbolt. This all feels like an excellent long term solution. And I'll still have my HD Native system with Avid Box in the toolshed.

what I like to use for dante monitoring is that little focusrite AM2

From the two mentioned options, I'd say Dante, but neither one is super appealing to me.
I've already expressed my concerns at its NAMM announce thread.. They have something like five motherboards with all kind of advanced or proprietary connectivity, but missing elementary AES or ADAT motherboards.. both has still biggest market share and works pretty well for expanding of existing systems or in a combination with vendors, who did audio interfaces with good real-world performance (RME, MOTU).

I'll try to break motherboard choices for B16 from my point of view.

- Dante
Established proprietary networking audio standard, multi-vendor devices are interoperable without issues. This applies also to RedNet, which is based on Dante.
If some vendor adds another features (like network control of preamp gains), it is done separately (eg. working as another service at the IP network) from common audio streaming, routing and device discovery, which is standardized by Dante and interoperable with other devices.

Disadvantage is, that DVS (eg. straight connection of computer NIC to Dante network with virtual soundcard) is really designed and usable just for tracking with external monitoring, live show recordings or generally some tasks, where latency isn't important. 4ms, which was reported by Wrenc, is not complete real-world latency of that converter (regardless what's being showed in DAW or some control panel)..
There can be lot of confusion around that. 4ms is default delay for Dante network stream, when one communicating endpoint is DVS.. However you have to add also Core Audio or ASIO buffer size and latency for possible routing and DSP in hardware and actual conversion to the equation.
Dante with most of high performance hardware endpoints (like PCIe cards) allow you to cut down that transport latency down to 150us. That's because time critical part of the network protocol is offloaded to hardware with tight deterministic latency compared to software, where performance of computer and operating system scheduling needs longer buffers.
Other related thing is, how such system will perform with loaded DAW project (typically overdubs), dedicated PCIe (or TB) card would have certainly some edge with short ASIO or Core Audio buffers over completely software solution.

So to sum it up, for tracking, performance and monitoring through Dante connected B16, I can really imagine going through DVS (it wasn't designed for that), but for example RedNet card in TB chassis or even better their Red4Pre makes that maybe interesting choice for you. Former Red4Pre will give you another advantage of it's local I/Os for another inputs and monitoring, plus its Digi Link connectors will allow you to connect to some HD/HDX system.

- USB
It can be good, when done well (RME), but its biggest disadvantage is lack of expansion. To use two or more different devices with DAW, you have to use OS X based software audio device aggregation, which is PITA and bring several performance and reliability challenges. Also in real-world, only handful of vendors (typically only ones, who do their own USB audio implementation at FPGA or their dedicated chips) offers reliability and satisfactory low latency performance.

I'd think about that only in case, when B16 will be fitted also with D/A card (for monitoring without software aggregation) and when no further I/O channel expansion will be considered. Also I don't really believe (of course I can be wrong), Burl USB interface will be somewhat great latency performer.. Typically those converter companies picks some 3rd party USB class compliant solution (like XMOS) and add it to their hardware.. which might be O.K., but performance wise, it's far from RME.

- MADI
Well oldest from the bunch, but still very good multi-channel option for some setups, it's synchronous interface, simple, reliable with negligible latency (few samples). You can still get lot of very good peripherals and interfaces via MADI.
If you for example combine B16 with some USB or TB MADI card (like RME MADIFace, MADIFace Pro with local analog I/Os for monitoring or new UFX+), then you will gain very good latency performance and also great RME features like TotalMix for cue mixes etc.
AFAIK, only disadvantage there is, Burl MADI motherboard doesn't really counts with daisy chained setups with other MADI hardware.. and doesn't support arbitrary channel mapping to MADI stream. Otherwise I can imagine for example another combinations with other gear.. like first 16 channels from B16, another 16 channels from some Lynx or Antelope converter etc.

- Digi Link
This might be still good option even for native DAWs, because if you get HD Native TB interface, you'll get very good latency performance.
Disadvantage over previously mentioned RME or Red4Pre might be in lack of low latency monitoring DSP mixer (only one stereo AUX in PT) and arguably also 3rd party software support from AVID (their Core Audio driver for HD Native interfaces was broken for several years, AFAIK they finally fixed it recently along with 12.7 release).
Second disadvantage might be necessity of another Digilink capable converter for D/A monitoring.

- Waves Sound Grid
This was just announced IIRC.. I haven't really used it, but it's very similar principle as Dante DVS with virtual Core Audio driver, although they achieve better latency with this purely software solution according to all reports, which I've read.
Personally, I won't go there, because I'm not really interested about their effect "servers" and I don't see especial appeal in another proprietary networking protocol.. Dante is also proprietary, but at least widely adopted (SG is being used by Waves and Digico.. otherwise I haven't really seen anything else). Other thing is, if you find latency isn't really fitting for you or DAW performance craps out, when doing demanding stuff, then you don't have the possibility to get some hardware card or an interface.
If you're interested, maybe search here about Sound Grid, IIRC user with nick ProPower had written several posts about its usage.

So I'd probably go either with Dante and Red4Pre.. or with MADI and say RME MADI Face Pro or UFX+. Those looks like reasonable interfaces to match such expensive converters and retain good performance and reliability.

Michal

3.5ms is the average I get on it. the Virtual sound card is different from the asio hardware you're used to. Its host independent. each dante interface has a small embedded computer that you send control messages that the interface sends the audio data via the ip protocol (this is why its a ~ $450 up charge because you pay for an embedded computer card). So you can have several computers recording (even those joco black boxes) plus others controlling and routing(even wireless). The only DAW function you are not allowed is the loop-back, or i/o daw inserts on the same interface. Mainly because dante is unicast on its inputs and ip assigned on its outs. So separate i/o is required for that (one i/o box for out one i/o box for in)

I forgot to add, I use the dante system to assign my monitoring mixes, the daw just records and its monitoring functions are disabled.
.

Hi audiospacific,
yeah.. that usage with direct routing of input channels to monitoring device certainly works well.. I've discussed different situation, when someone assume DVS as "normal" high-performance audio interface just connected via Cat 5 patchcord to the converter, where one can monitor through DAW, play virt. instruments etc. That's what I've meant by "it wasn't designed for that". Othewise, if you essentially mult input signals to more destinations and one of them is DVS, which is used in tape-like fashion, then of course, it's great and one of main points of whole networked audio.
However even with that kind of tracking situation, one should imagine all implications of that for his particular workflow, so for example if your artist is receiving "direct" feed from input converter to his Dante monitor, it's all fine, but as soon as he would like to hear what's recorded at his phones (basically switching CUE feed from pre DAW to post DAW), it would require rerouting different source to "his" destination.. which might be for example that 2-ch Focusrite Dante D/A box. Of course there are ways around that like with external mixer and not everyone works that way. But to me, there are quite natural advantages of having some low cue latency mixer available at interface.. that's why I've mentioned for example TotalMix for MADI or Focusrite Control, where you can handle those situations relatively easily.
But again, I don't really know the context of OPs tracking, maybe he's always tracking at commercial studios with external CUE monitoring systems.. so this won't apply.

Michal

Thanks msmucr for such a thorough answer!

I just found out today from the horses mouth that Burl Dante is not actually shipping. So I can't actually buy it any time soon. So strange, it's been on their website for like 2 years or something, why even put it there..

Interesting to see that you would match the Dante with the Red 4Pre as a first Dante choice. I guess that does make much sense, and especially so for me since I have a Burl Mothership with Digilink card. The red 4Pre would allow me to interface with the Mothership and I suppose if I got a B16 with Dante in the future that as well. Although if I get the Focusrite it is meant to play the secondary interface role at an overdub studio that I had envisioned for the B16 with Dante- which I now find out is not actually available. Bummer!

Quite the conundrum. I wonder if the MADI Burl card is also not actually available like the Dante.

Essentially, the situation is that I have a mix room where i can't record. In this mix room I have a 12 x 16 Digilink Burl mothership with PTHD and a console for mixing.

I recently got the opportunity to set up a recording rig in another room. This room is not quite pro but will still do nicely. It's the kind of place where I can do overdubs of vocals and acoustic instruments, electric guitars etc.. but probably not drums and full live bands. Since then I've been tearing out my hair trying to find the right secondary recording interface for me.

It's a bit limiting in the sense that I would like to get something that could be folded into my current home system if I left that room. So ideally PTHD friendly. Or if not I can acquire PT Vanilla. The guy I share the space with works with Studio One and can't afford PTHD. I certainly don't want a SECOND PTHD license. Been wondering if I even need PTHD really lately. I initially thought of the Red 4Pre since it provided a REALLY nice basis for an overdub room at a reasonable price (incidentally I'm short on preamps too) The Digilink connections made it friendly to integrate into my Mothership system should I centralize, even providing some extra I/O that the Mothership doesn't have like SPDIF.

Then I read about how the Red 4Pre is not sample accurate with PT HD Native which turned me off some, but then neither is the Mothership. And secondly I love tracking with Burl conversion! I think it would make me sad to use Focusrite converters with 12 channels of Burl inputs sitting at home. The Burl magic is largely at the input stage! I'm not keen on shuttling the Burl mothership back and forth obviously. I wouldn't leaving the Burl at the secondary space but then would need a second Non PTHD interface for mixing. The red4Pre only has 8 outs for that.

Enter the B16 haha. I figured get that, pop some cards I already have into it (though it would reduce my mix rig greatly) and record with Burl on site. I stay in the Burl family, can centralize elegantly in the future, and open the door to Dante. But it turns out Dante is not available yet for the B16, and even if it was I would need a Focusrite RedNet card or the Red 4Pre anyway to make Dante truly useable for overdubbing regardless, making the Dante proposition fairly expensive at that point.

Sorry for the explanation haha but context is always good indeed!

It is, no worries on that score. My friend has one, works great.

Really? They told me no it wasn`t shipping yet.. :(

MADI? that's connected to my friends Mothership unless I'm sorely mistaken. DANTE and USB are missing in action, tho.

Ah ok that makes sense !

One thing I`m not clear on regarding the Dante option though, does the Red 4Pre provide the same service as the RedNet PCIe card- accelarating the lantency of Dante? I`m Under the impression it doesn`t and that it doesn`t help Dante anymore than DVS. In other words, only the RedNet card speeds up Dante.

Using a Burl Mothership Madi with RME Madi FX here. Running great! Probably one of the best converters (Burl) with the best latency and drivers (RME) available.

Definitely.. thanks for that.

You've also touched possible usage of Red4Pre with Mothership via Digi Link, just for clarification, that's possible only via additional ports at your HDX or HD native card - eg. if you have some free Digi Link ports there, you can connect it to DAW. It's not possible to hook it to Mothership somehow.

HD sample accuracy with 3rd party converter is pity, but ball is at AVID side I believe. They somewhat tolerate(don't support) those 3rd party vendors. So only choice is to design their gear to behave exactly as AVID I/Os. When they use converter chips with higher latency (for whatever reason like sonics), then it's not technically possible. If they use converters with lower latency than emulated AVID I/O, then they have to employ additional internal delay at hardware for that alignment (essentially degrading performance for someone, who wants shortest roundtrips).
Hopefully, AVID will figure out, it's beneficial for their whole ecosystem to add simple input/output delay offsets for 3rd parties.. because users will use it regardless of that and this is just pure annoyance, when analog inserts are used.

Yes it does. AFAIK there are actually two options for hw accelerated Dante interfaces.. First is that PCIe interface by Focusrite or Yamaha (it's the same as it's likely made by Audinate) and the second choice is Red4Pre (8Pre) for TB.

However Red4Pre has an advantage of its own low latency DSP mixer.. Dante I/Os appears there just as another channels.. that's what I believe is useful for that cue mixes.
PCIe card is just I/O interface - if you want some mixing (like input and backing track), it's necessary to do it at DAW (buffer size involved) or externally using some mixer.

With regards to options and availability, I'd probably call directly to Burl.
They should give you accurate info regarding product launches and walk you through all the options.

Otherwise, as I've mentioned, I have no doubts MADI board+RME combination will be also very good choice (for example in the case, when Dante board won't be launched in some foreseeable future).

All the best,

Michal

Dante Virtual Soundcard will always be the weakest link. For higher track counts and lower latency based around a DAW, the Dante Accelerator Cards should be considered vital to improving performance.

By comparison, the Merging Ravenna CoreAudio driver allows 2.1ms latency with 128 tracks at 24/44.1k and 2.6ms at 24/48k also with 128 tracks, which is simply not possible with Dante Virtual Soundcard.

Sans DAW and Dante Virtual Soundcard, Dante is fantastic.

Basically as reynaud and others have said, using a computer's NIC will not get you usable low latency for real studio work, you need their propriatary Dante PCIe card. On the other hand USB will likely be twice or more the latency of the PCIe card and carry less load. So if serious for your Burl options between USB and Dante, go Dante with their PCIe card (just look at the Antelope Orion threads with their choice between third party Madi interface and USB). Remember there are other interfaces and converters out there too so if you are not locked into that Burl, options open up.

While I generally agree with you Bassmankr.. I believe it's bit confusing for OP to say the only one option is PCIe card.. Red4Pre is another choice, which provides Dante hardware endpoint for the computer with low latency and less buffering for audio streaming than DVS.. I don't know, which exact chipset is being used by Focusrite there, but I suppose, it can be for example Brooklyn II. 32 I/O channels are supported there with low latency mixing, so unless this count will be limiting (I don't think with B16), he don't necessarily need a PCIe card.

Michal

Focurite makes no mention of using the Red4Pre without DVS in any of their documentation, nor in any of the online product information. The Red4Pre offers Thunderbolt connections, and the device latency only matches the Merging Ravenna CoreAudio driver (almost exactly), not improving on it. My testing with the Ravenna driver matches Focusrites configuration - Mac Pro, OS X.11, Reaper, except I used the built in NIC.

Notice, Focusrite doesn't say which interface is used to achieve the lower latency, but I suspect it's via the Thunderbolt port. i.e. when using Dante, and low latencies are required, the Dante Accelarator card is still required.

And, the Dante ports on the Red4Pre are both primary ports. So, no redundant networking here.

Hi,

I'm not sure, what you can achieve with Merging driver at OS X, but it's bit irrelevant, unless Burl would have Ravenna motherboard.
As previously mentioned, DVS itself reach pretty much subpar RTLs, because of increased buffer lengths for network streaming compared to HW Dante endpoints.
However Red4Pre isn't really meant to be primarily used with DVS (although it's of course possible for instance when another computer would access it simultaneously along the primary DAW and both will be mixed to common analog outputs for example), because there is TB or Digi Link mode.
So in case of normal native DAW configuration, you'll use TB mode and in case of Pro Tools HD configuration, you'll connect it to HDX or HD native card.

If you'd like to use it with some other Dante hardware, then it will appear just as another HW endpoint at network and you can use Dante Controller software to establish required routing and clocking setup. Then channels routed to Red4Pre will be accessible at DAW either via its Core Audio driver in TB mode or via that Digi Link, so DVS isn't used at all and no actual network streaming is going through DAW computer.

https://support.focusrite.com/hc/en-...Pre-with-Dante

I don't think so. That PCIe card is required only when you need 128ch. I haven't used Red4Pre in any project, but as it's hardware endpoint based on Audinate chip, it won't add any higher network transport latency than other similar hardware "bridging" units such as RedNet 5 or HD32R, which are reliably running at shortest network buffers (~ 150us) with simple network topologies.
Of course the actual performance with DAW will be also affected by Red4Pre native driver, but this doesn't have anything with network streaming latency, which is just added to that figure (two times for RTL).

But so far, it looks like pretty good performer (just looking at sheer figures), which is comparable to other TB interfaces at OS X.
https://us.focusrite.com/red-range-i...#latency-table
Avid Pro Audio Community - View Single Post - Latency of 3rd party interfaces compared to HD Native
* of course, there are used built-in converters at Red4Pre.. Burl I/O cards might have different chips, so it's necessary to extrapolate a bit together with that mentioned Dante transport.

It's rather convenience feature, which saves you Ethernet switch in simple setups. For example, you can use the first port for the connection to Dante converter and the second for your computer (for Dante remote control and routing). I don't really think, network redundancy is such important at OP's setup or small fixed installations at similar project studios.

Michal

My point was that Dante Virtual Soundcard will always be the weakest link.

The equivalent software, Ravenna CoreAudio driver, as I've pointed out previously, offers much lower latency and higher track counts than DVS using the built in NIC (even offering DSD).

I am running 128 channels over Ravenna using the built in NIC and a software driver, that previously required a Dante Accelerator card to match the versatility or get close to the performance offered by the Ravenna CoreAudio driver. I am now scaling down the Dante portion of my systems and scaling up the Ravenna portions.

My suggestion is to not use DVS for higher track counts or when low latencies are required. It more a convenience than anything. Rather, consider the Accelerator card if using Dante and require a high performance system.

Having not tested the Red4Pre, I will reserve judgement on suggesting it as an option. Suggesting it also doesn't solve the original poster's issue as it's another hardware purchase, which doesn't expand his Burl Motheship, and which then opens up discussions about other perhaps more appropriate options.


Thank you for the link, that article explains things clearly. So they suggest using Thunderbolt to the computer and then routing through the Red4Pre to Dante.

So yes, DVS is thankfully not required and the low latencies are, as I expected, via Thunderbolt.


It is also required when track counts via Dante above 32 channels (beyond what the Red4Pre provides) or 64 channels (beyond what DVS offers) are needed, to achieve lower latencies and to ensure improved network system performance.


That's the table I was quoting indirectly when comparing my Ravenna CoreAudio system latency results to the Red4Pre latency figures.

The Ravenna CoreAudio driver using the built in NIC achieves lower latencies than the figures in the link you provided and are almost matched by the Red4Pre via Thunderbolt:
Avid Pro Audio Community - View Single Post - Latency of 3rd party interfaces compared to HD Native

It has to be asked why Audinate isn't able to offer the same performance as the free Ravenna CoreAudio driver in the paid DVS software. Why is the hardware Dante Accelerator card required to match a purely software solution?


Perhaps not today, but it may be critical if the system is ever expanded or as requirements change. I prefer buying once, knowing that in future expansion is possible, rather than buying many times as I expand.

Since you are discussing Ravenna, AES67, please note there are some performances differences with standardized and premium versions [that are included with Merging Hapi and Horus interfaces] as well as Mac/PC capabilities.

msmucr, the Thunderbolt interfaces although basically being an exterior PCIe slot are not currently behaving that way when tested likely due to their drivers. While they should be just slightly above PCI / PCIe slot numbers for latency / load, they are coming in much higher and more along the line of in between PCIe and USB numbers. Details are in the the interface database thread in this section. What's actually usable for an operation depends on their specific needs but for low latency at load and higher channel count Dante in the studio you are looking at using their specific PCIe card. In general for other options you should be looking at PCI/ PCIe slot interfaces. Likely this is due to the use of FPGA chips on the interfaces for some offloading and very good drivers.

You're probably basing that at one mention at Vin's thread, where he suggested that kind of finding about UFX+ at his system, however it would be incorrect to use it as a general statement about TB or PCIe interfaces.

First of all, besides different hardware, most of newly released external TB interfaces has different drivers than previous internal PCIe interfaces.. so it's not really apples to apples comparison. This applies also to RME, where all previous HDSP(e) interfaces shares the same unified driver, which has been finely tuned for many years, while UFX+ driver is completely different package.

Other thing is, situation might be very different at OS X platform (I assumed OP uses Mac), where TB interfaces are established for about 6y now and currently only vendor supported computer bus for high perf. peripherals. compared to PC/Windows platform, where it's still bit of mess with varying support of different TB versions by motherboard and audio interfaces vendors and Microsoft (up to W10 TH2 it was purely 3rd party vendors business).

Finally, I've made bunch of TB tests, years ago, when first Macs without internal PCIe slots were introduced and also when first Supermicro PC boards with TB socket appeared on market.. External chassis, audio and video interfaces, 10Gb NICs and Fibre channel adapters, storage adapters etc.
Well if there were some problems, then it was mainly affecting long term reliability, disconnections.. etc. this is another story (which was later sorted out mainly by driver updates), but basically there weren't any issues with different performance with chassis compared to internally mounted cards.. even at much more demanding applications than audio streaming, we're talking about.
So for example, when you mount PCIe audio or DSP card into external TB chassis, then it works very well and TB protocol overhead won't affect its performance.. Lot of people working at such setup every day (eg. AVID HDX, RME, UAD PCIe cards are most common) and there isn't any issue with that.

There isn't any inherent problem with TB performance for this application. If someone achieves different performance with internal PCIe interface and other native external TB interface, then it's not because of TB, but likely due to different drivers and/or design of those units.

Michal