CalDigit Launches Element 5 Hub With Thunderbolt 5

[theluggage]

One thing of note that is never made clear is that all the ports on this hub share the 80 Gb/s interface to the host so you can not have all the ports on the hub using their maximum supported bandwidth.

You'd think that would be obvious, but maybe it is worth stating! People keep asking for 3 displays, 10G ethernet and NVMe as well as loadsa downstream TB and USB bandwidth... all off a single host port.

I think that, less obvious (and much harder to clarify), is that most hubs, however many ports they have, will only offer you 1 or 2 streams of USB 3.1/3.2 at 5/10Gbps - distributed via internal USB 3 hubs (and, AFAIK, a 10Gbps USB hub doesn't mean it can run two 5Gbps devices at full speed simultaneously). You won't get 80Gbps of total USB 3 bandwidth out of these hubs. If you just want to connect a bunch of USB 3 devices (which includes many non-Thunderbolt USB-C peripherals) then an old school USB 3 hub may be just as good for less money. If you want to mix USB 3 devices with displays or "true" TB devices - which get a separate share of the total bandwidth - then a hub starts to make sense as a way of maximising the use of your TB bandwidth.

Given that all Macs except the Air and entry-level iMac now have at least 3 TB4 ports, my approach would be to use a TB4 hub to "concentrate" all the non-bandwidth/latency-critical devices on the hub, maybe alongside one display, to free up host ports for the more critical peripherals (whether it's a top end NVMe drive, ultra-high-bandwidth display or even a USB 2 device - like an audio interface - that preferred not to be on a hub).

 

[PaulD-UK]

@whitby "...all the ports on this hub share the 80 Gb/s interface to the host ..."

The PCIe bandwidth (using TB5 protocol) is actually only 64Gb/s, because that's the PCIe 4x4 limit.

@theluggage "Why would a hub that still supports TB3 bother to use USB4 USB tunnelling for its internal ports?"

Your linked data-path diagram of the Caldigit TB4 Hub implies that tunnelled USB 3.* data, output through the three downstream TB4 ports, might be in addition to the single 10Gbps data stream that feeds the two hubs?

I suspect Caldigit's diagram is oversimplified, and all the possible USB 3.* streams come, via switches, from the tunnelled USB 3.* 10Gbps data steam.
Unless there is no 'tunnelled' USB 3.* stream - when the hub is connected to a computer's non-TB USB 3.* Host port.

Also, is the USB 3.* data in addition to the PCIe 4x4 data stream, or tunnelled within it?

Maybe @joevt can clarify this? 😀

 

[unQuestionable]

I LOVE the T4 Hub they’ve released for a robust livestreaming setup via my M1 Max MacBook Pro, so once I upgrade to an M6 Max MacBook Pro, I’ll grab this, too!

 

[canonical]

It doesn't even have ETHERNET. What kind of HUB, especially for Mac notebooks, does not even offer ANY ethernet? Hopeless and useless.

 

[Ifti]

That OWC dock will be a Don't Buy.

Why?

 

[Ifti]

It doesn't even have ETHERNET. What kind of HUB, especially for Mac notebooks, does not even offer ANY ethernet? Hopeless and useless.

It's a Thunderbolt hub, not a dock. Their Dock will have more ports, including Ethernet.

 

[phrehdd]

And why USB-A ports when all the shiny things now are made with USB-C.

Some people have older devices, and what is the big deal? There are plenty of decent USB C to A and A to C adapters out there if you need one. I have the Element 4, and it has continued to work quite well for me. When the Studio Max M4 comes out, I see this particular new item on my short list.

 

[theluggage]

What kind of HUB, especially for Mac notebooks, does not even offer ANY ethernet?

Well, this is a direct successor to the TB4 Element hub, and a lot of people (including me) bought that so obviously there is a point. Especially for people with desktop Macs that already have ethernet and HDMI.

Hubs like this let you connect the maximum number of TB and USB devices to a single port. You can add Ethernet via a TB or USB adapter. Other "docks" try ro give you all the interfaces you'll ever need (HDMI/DP/Audio/Ethernet) in one box. Neither is the right or wrong way to do it - but the "kitchen sink" ones tend to be a bit more expensive (for the same brand and TB version).

Caldigit offered the TS4 dock alongside the TB4 Element Hub. Maybe they'll produce a TS5 - if not, Sonnet have announced a "kitchen sink" dock with TB5.

Take your pick.

 

[drrich2]

...not disagreeing but just to nitpick, "USB A" peripherals would describe peripherals with captive USB-A cables or "dongles" built into USB-A plugs (not that there aren't plenty of those in circulation). Peripherals with detachable cables should officially have type B or C sockets (I am aware of a few naughty ones with USB-A sockets but AFAIk they break the standard).

Good point. In a practical sense, most peripherals I've seen come with an included USB cable, and the cable end for the computer on many older ones had a USB-A connector. In theory, one could buy a USB-C to USB-A cable and use it.

But that means buying a new cable. A lot of people who've just taken their new gadget (e.g.: dock, hub, whatever) out of the box want to plug in what they've already got.

The cable thing sounds minor, and can be. Yesterday I did a little digging at Monoprice to see what some USB-C cables would cost, to have some spares on hand. To make an informed decision, I looked at the various versions of USB; this page at Kingston Technology offered a table that was helpful, albeit the topic remains complicated. Not only have there been multiple iterations of USB, but some of them have been renamed, and the naming gets confusing (e.g.: USB 3.2 Gen. 1x2 vs USB 3.2 Gen. 2x1, and USB 3.2 1x1 was previously known as USB 3.1 Gen 1 and USB 3.0, and USB 3.2 Gen. 2x1 was previously known as USB 3.1 Gen 2). I am not gonna remember all that.

I can see where a number of people would really like it if there was simply a matching USB-A port on their shiny new hub, and they plugged their printer's old cable into it, and it all just worked.

 

[EugW]

Some people have older devices, and what is the big deal? There are plenty of decent USB C to A and A to C adapters out there if you need one. I have the Element 4, and it has continued to work quite well for me. When the Studio Max M4 comes out, I see this particular new item on my short list.

Funny you should mention that. I just sold my M1 Mac mini, and besides the fact it's a used machine so it's cheaper, the buyer said one thing he valued on that machine is its USB-A ports, since the various equipment he'd use with it is USB-A.

The cable thing sounds minor, and can be. Yesterday I did a little digging at Monoprice to see what some USB-C cables would cost, to have some spares on hand. To make an informed decision, I looked at the various versions of USB; this page at Kingston Technology offered a table that was helpful, albeit the topic remains complicated. Not only have there been multiple iterations of USB, but some of them have been renamed, and the naming gets confusing (e.g.: USB 3.2 Gen. 1x2 vs USB 3.2 Gen. 2x1, and USB 3.2 1x1 was previously known as USB 3.1 Gen 1 and USB 3.0, and USB 3.2 Gen. 2x1 was previously known as USB 3.1 Gen 2). I am not gonna remember all that.

The main annoyance for me is knowing the speeds of each cable. Some of my USB-C cables support up to 10 Gbps. Some of them support only 5 Gbps, and the some support up to 40 Gbps. I've taken to labeling my existing cables, but for new cables I've only been buying certified cables, either certified Thunderbolt 4 cables (especially for Thunderbolt 4 hubs), or else certified USB 4 cables.

BTW, it's also a bit of a minefield with those adapters. A lot of them out there are junk, so you need to tread lightly when buying.

 

[fredn]

I don't know what's up with this, but I'm going to raise it internally. We use Shopify for our web shop, so it could be some weirdness with the platform. Either way, I'll see what's up with this and figure out why it wasn't properly reporting the currency it's showing you.

Hi Dalton, thank you for that!! It was truly a weird and unexpected thing.

 

[fredn]

The terms overlap somewhat but hubs generally function more like port replicators, with or without power, whereas docks are typically powered and have more added features like Ethernet or HDMI or SD card readers or whatever.

...

Thank you for this.

My takeaway is -- just buy the item you need with the ports you need at the price point you are OK with, taking into account any duplicate or unnecessary ports, and disregard any terms like "hub" or "dock".

 

[drrich2]

The main annoyance for me is knowing the speeds of each cable.

I've taken to labeling my existing cables, but for new cables I've only been buying certified cables, either certified Thunderbolt 4 cables (especially for Thunderbolt 4 hubs), or else certified USB 4 cables.

A label maker sounds like one good way to do it.

I'm torn between ordering higher end (e.g.: in terms of data transmission and power delivery capability) cables that cost more money (but might be used for higher data transmission tasks or charging high demand items), and cheaper cables that should serve fine for charging lower demand devices, all the while knowing I'm not likely to remember the particulars well.

Another option, if I don't stick to buying whatever is on sale, is to pick at least one category from a line that offers color choices. So, if all the 10 Gbps capable USB-C cables are orange and all the 480 Mbps USB-C cables are black, that could work.

 

[Jonas Danielsson]

Please make a device with 10GbE network and 5-6 usb c ports some of them could be thunderbolt. That would make i perfekt dock for some that has mutiolendevice connected.

10GbE for home network with nas
USB c for Wacom, keyboard, control panel , streamdeck, memory card reader , firmware update etc

 

[phrehdd]

A label maker sounds like one good way to do it.

I'm torn between ordering higher end (e.g.: in terms of data transmission and power delivery capability) cables that cost more money (but might be used for higher data transmission tasks or charging high demand items), and cheaper cables that should serve fine for charging lower demand devices, all the while knowing I'm not likely to remember the particulars well.

Another option, if I don't stick to buying whatever is on sale, is to pick at least one category from a line that offers color choices. So, if all the 10 Gbps capable USB-C cables are orange and all the 480 Mbps USB-C cables are black, that could work.

I have gotten to the point that I have nothing slower than cat 6/6a cables for Ethernet. My cat 5 cables all have new homes. USB A is now for items like printers that are often A to B. Nearly all of my cables are USB C and middle to high quality. I do have Tbolt rated cables but not many.

 

[joevt]

I didn't think that Mac supported the USB 3.2 x2 modes at all (and they don't seem to be widely supported, nor can I see them expanding now there is USB4).

macOS supports USB 3.2 gen 2x2 20 Gbps since Sonoma.
https://forums.macrumors.com/thread...-at-full-speed-on-a-mac.2443940/post-33591251
https://eclecticlight.co/2024/12/23/will-that-hub-or-dock-slow-your-ssds-or-even-make-them-faster/
https://eclecticlight.co/2024/12/25/how-your-intel-mac-can-use-usb-3-2-gen-2x2-drives-at-full-speed/

To get USB 3.2 20 Gbps from Apple Silicon, you have to disable USB tunnelling so that the USB controller of the Thunderbolt 5 hub/dock is used instead of the USB controller of the Mac. You can do this by placing a Thunderbolt 3 device between the Mac and the Thunderbolt 5 hub/dock.

USB 3.2 gen 2x2 is an inexpensive way to get more than 1100 MB/s.

Caldigit published a block diagram of the TB4 Element hub: https://www.caldigit.com/element-hubs-controllers-and-data-paths/ - but it doesn't distinguish between the TB4 (USB tunnelling) and TB3 case.

As far as I can tell, the USB3.1 bus on my Element 4 hub, connected to a TB4 Mac Studio, reports as "Intel corp" - and the front USB-A ports hang off a further USB 3 hub ("Caldigit") connected to that.

In that diagram, the USB Hub #1 is part of the Thunderbolt controller which is why it reports Intel as the vendor. Unlike other Thunderbolt 4 hubs, CalDigit connects a separate USB hub (#2) to the exposed USB only port of hub #1 because USB hubs are cheap and adding the ports doesn't make the Thunderbolt 4 Hub much larger.

The USB-A port of the OWC Thunderbolt 4 Hub is the exposed USB-only port of hub #1. Since there's no hub#2, it might be slightly faster than the USB-A ports of the CalDigit Thunderbolt hubs.

Why would a hub that still supports TB3 bother to use USB4 USB tunnelling for its internal ports?

I don't know what benefit USB tunnelling has over using the USB XHCI controller of the Thunderbolt controller of the Thunderbolt hub/dock except that the USB controller of the host is used which could be more compatible in some situations - such as if other USB controllers don't have the same support as the built-in USB controller during boot (which I don't think is true in the case of Macs and most PCs).

USB-IF probably wanted to add USB tunnelling to Thunderbolt to make USB4 so that USB4 has something to do with USB besides the USB-C connector!

I know in many cases the USB controller of a Thunderbolt dock performs better than the USB controller of an Apple Silicon Mac. Maybe later Apple Silicon Macs have improved USB performance?
https://forums.macrumors.com/thread...ally-10gb-s-also-definitely-not-usb4.2269777/

The USB controller of a Thunderbolt 5 dock/hub is definitely superior to the USB controller of the Mac - if you have a USB 3.2 gen 2x2 20 Gbps device.

Or, having a mix of USB-A and USB-C sockets leaves you perfectly free to use an A-to-C adapter to get extra USB-C sockets when you need them.

USB-A to USB-C adapters are against the USB spec for reasons - even though they do work. One reason it might be against spec is because if you use a USB-A to USB-C adapter at either end of a USB-C cable then it would let you connect two USB-A ports together which could cause damage.

The "extra" USB sockets - regardless of type - on this hub seem to be limited to 3.1g2 and 7.5W by the internals of the hub, so the type C sockets can't do aything that the type A sockets can't do.

Anyhow, the real point is that USB A/B- or even most "USB C" peripherals - only use USB 2, 3.0 or 3.1 protocols up to 10Gbps & use less than 7.5W - and gain absolutely no advantage from being connected to a USB C port (The Type A ports on this hub support up to 3.1g2 10Gbps & Apple doesn't support USB 3.2x2, which is the only flavour of USB 3.x that actually needs a type C connector to work). The current hub chipsets only support 3 downstream TB/USB4 ports so all the "extra" ports are limited to 10Gbps USB whether they're type A or B, and supplying 15W power to every port would need a bigger, more expensive power supply & more heat.

USB 3.2 gen 2x2 is supported since Sonoma.

The USB hub that is built into the Thunderbolt 5 controller of a Thunderbolt 5 hub or dock supports USB gen 3.2 gen 2x2 20 Gbps. CalDigit could have chosen a USB gen 3.2 gen 2x2 hub for the second hub so that the USB-C ports could support gen 2x2 speed while the USB-A ports are limited to gen 2x1 speed.

I think that, less obvious (and much harder to clarify), is that most hubs, however many ports they have, will only offer you 1 or 2 streams of USB 3.1/3.2 at 5/10Gbps - distributed via internal USB 3 hubs.

Thunderbolt 3 docks using Alpine Ridge usually had multiple PCIe USB XHCI controllers (though most of them were limited to USB 5 Gbps).
Thunderbolt docks since Thunderbolt 3 (Titan Ridge) have one USB controller and use a USB hub for all USB devices of the Thunderbolt dock (including additional USB hubs).

(and, AFAIK, a 10Gbps USB hub doesn't mean it can run two 5Gbps devices at full speed simultaneously)

That's a good question. Can two 5 Gbps devices connected to a 10 Gbps USB hub achieve greater than 5 Gbps performance using ATTO Disk Benchmark.app's multiple disk benchmark?

USB 10 Gbps is 9.7 Gbps of data.
USB 5 Gbps is 4 Gbps of data.
So full performance from two 5 Gbps devices might be possible.
Multiple read requests can be sent to multiple devices at once before the first read request completes.
Therefore, the two USB 5 Gbps devices can be transmitting data at the same time.
However, the USB 10 Gbps hub needs to be able to receive data from both devices at the same time in order to get beyond 5 Gbps.
In order to do that, the USB hub has buffers. I don't know if the buffers are sufficient for this situation.

@whitby "...all the ports on this hub share the 80 Gb/s interface to the host ..."

The PCIe bandwidth (using TB5 protocol) is actually only 64Gb/s, because that's the PCIe 4x4 limit.

Maybe that's an arbitrary limit, since the Thunderbolt controllers are not transmitting PCIe between each other. The PCIe data is tunnelled over Thunderbolt. The only thing that is definitely PCIe 4x4 limited is a device connected to real PCIe lines of a Thunderbolt controller in an external enclosure. On the host side, the upstream of an integrated Thunderbolt controller in Apple Silicon is not real PCIe.

PCs with PCIe gen 4x4 discrete Thunderbolt host controllers are of course limited by the PCIe gen 4x4 upstream of the Thunderbolt host controller but couldn't someone just make a PCIe Thunderbolt host controller with a faster upstream?

@theluggage "Why would a hub that still supports TB3 bother to use USB4 USB tunnelling for its internal ports?"

Your linked data-path diagram of the Caldigit TB4 Hub implies that tunnelled USB 3.* data, output through the three downstream TB4 ports, might be in addition to the single 10Gbps data stream that feeds the two hubs?

The single tunnelled USB data stream feeds the hub #1 which is inside the Thunderbolt controller. That hub has USB ports for the three downstream Thunderbolt ports and a hub #2 that is external to the Thunderbolt controller.

If USB is not being tunnelled then the XHCI controller of the Thunderbolt controller is used. I think it provides a single USB port for the internal hub. So basically it's not much better than the tunnelled USB situation.

I suspect Caldigit's diagram is oversimplified, and all the possible USB 3.* streams come, via switches, from the tunnelled USB 3.* 10Gbps data steam.
Unless there is no 'tunnelled' USB 3.* stream - when the hub is connected to a computer's non-TB USB 3.* Host port.

Three connection types:
1) USB Host: No Thunderbolt is involved. The upstream is USB from the host's USB XHCI controller.
2) Tunnelled USB: The upstream is Thunderbolt. USB from the host's USB XHCI controller is tunnelled by the host's Thunderbolt controller. The dock's Thunderbolt controller converts the tunnelled USB back to normal USB.
3) No Tunnelled USB: The upstream is Thunderbolt. The host sends commands via tunnelled PCIe to the XHCI controller of the dock which has a USB port.

In all 3 cases, there is one USB stream to the USB hub of the Thunderbolt controller of the Thunderbolt dock.

Also, is the USB 3.* data in addition to the PCIe 4x4 data stream, or tunnelled within it?

Tunnelled USB is separate from tunnelled PCIe is separate from tunnelled DisplayPort is separate from tunnelled other stuff (Thunderbolt networking or Thunderbolt target disk mode).
If USB is not tunnelled (such as the case when a Thunderbolt 4/5 device is connected upstream to a Thunderbolt 3 device or host), then the PCIe USB XHCI controller of the Thunderbolt 4/5 device is used and the host communicates with that using PCIe tunnelled over Thunderbolt.

 

[drrich2]

Per an Apple Insider article -

CalDigit's new Element 5 hub brings more Thunderbolt 5 ports to the Mac ​

This interesting tidbit - "So far, there haven't been very many Thunderbolt 5 hubs — yet. We suspect that most vendors are awaiting a recently announced Intel controller chip to take maximum advantage of the new technology."

Apologies if that's already been discussed, but what do people anticipate that new Intel controller chip is going to add?

 

[whitby]

Apple Thunderbolt ports can have up to two DisplayPort connections over Thunderbolt. An LG UltraFine 5K display uses two DisplayPort connections from Thunderbolt to achieve 5K60. This connection type is called dual tile SST. It requires two DisplayPort connections with HBR2 x4 link rate.

The Apple Studio Display supports that mode for old GPUs that don't support HBR3 or DSC. For new GPUs, the Apple Studio Display can use a single HBR2 x4 link with DSC to achieve 5K60. I don't think the Apple Studio display has HBR3 support which could allow 5K60 without DSC (but only up to 8bpc).

An Intel PC with Thunderbolt 5 could support an Apple Studio Display and an LG UltraFine 5K display from a single Thunderbolt 5 port using a Thunderbolt 5 hub (3 DisplayPort connections - 28.1508 Gbps for the LG and 11.232 Gbps for the Apple). Actually, that adds up to less than 40 Gbps, so perhaps even a Thunderbolt 4 hub could do it - since each display is a Thunderbolt display with their own DisplayPort outputs (meaning none of the DisplayPort outputs of the Thunderbolt hub are used). Well, it's very near 40 Gbps (closer to 40 Gbps than dual tile SST Apple XDR display) so it might not work. 28 Gbps is for the older LG UltraFine 5K display (27MD5KL). The newer LG UltraFine 5K display (27MD5KL) is at least 29 Gbps so that's a no go there.


If the displays use DSC@12bpp then 5K60 only requires 11.232 Gbps.
If a display doesn't support DSC, it may be possible to use an MST hub that supports DSC decompression to reduce the bandwidth used by the display. However, all the DisplayPort 1.4 MST hubs I have examined do not support 10bpc decompression (such as the CalDigit SOHO). I haven't looked at the new DisplayPort 2.1 MST hubs. Reducing bandwidth to the MST hub may require disabling some DisplayPort pins. One way to do that is to connect the MST hub to a USB-C hub that supports USB 3.x (such as the CalDigit SOHO).


I also don't know when 120 Gbps gets used.

This is very helpful. I had not remembered that the 5K displays use 2 Display Port streams and TB 5 only supports 3. Hence the inability to get 2 x 5K displays to work on a single port. The Apple display supports DSC and the LG does not support DSC which is moot since we have the available data rates without compression but not enough Display Port streams. What is interesting that you can connect 2 x 5K displays to one TB 5 on the Mini and they work but blank occasionally. I am not sure why they work at all.

I moved one display to a port on the Mini and left the other on the Hub along with a relatively low bandwidth TB 2 external RAID array. This seems to work. My low data rate USB devices use the down stream USB C connections on the Apple Studio.

Your technical knowledge in this and other ports has been extremely useful and I thank you.

 

[Allen_Wentz]

And why USB-A ports when all the shiny things now are made with USB-C.

Perhaps for convenience because it is a dock and there are a zillion legacy old things with USB-A ports. Vendors just try to guesstimate what buyers want.

 

[xpaulso]

5K (LG UltraFine 5K or Dell UP2715K) and 8K (Dell UP3218K) use dual tile SST to achieve 60Hz. Then can do single tile SST at least 30Hz.
macOS did support MST for dual tile 4K60 MST displays which existed before single tile 4K60 SST displays.

Thanks, I wasn't aware it used "Dual Tile" SST. Are they essentially alternating every other frame of (2) 30hz mirrored displays?

 

[xpaulso]

This is very helpful. I had not remembered that the 5K displays use 2 Display Port streams and TB 5 only supports 3. Hence the inability to get 2 x 5K displays to work on a single port. The Apple display supports DSC and the LG does not support DSC which is moot since we have the available data rates without compression but not enough Display Port streams. What is interesting that you can connect 2 x 5K displays to one TB 5 on the Mini and they work but blank occasionally. I am not sure why they work at all.

This is interesting, what cables were you using to daisy chain the (2) Studio Displays? Also how long are the cables? This feels like a low signal issue.

 

[xpaulso]

You'd think that would be obvious, but maybe it is worth stating! People keep asking for 3 displays, 10G ethernet and NVMe as well as loadsa downstream TB and USB bandwidth... all off a single host port.

I think that, less obvious (and much harder to clarify), is that most hubs, however many ports they have, will only offer you 1 or 2 streams of USB 3.1/3.2 at 5/10Gbps - distributed via internal USB 3 hubs (and, AFAIK, a 10Gbps USB hub doesn't mean it can run two 5Gbps devices at full speed simultaneously). You won't get 80Gbps of total USB 3 bandwidth out of these hubs. If you just want to connect a bunch of USB 3 devices (which includes many non-Thunderbolt USB-C peripherals) then an old school USB 3 hub may be just as good for less money. If you want to mix USB 3 devices with displays or "true" TB devices - which get a separate share of the total bandwidth - then a hub starts to make sense as a way of maximising the use of your TB bandwidth.

Given that all Macs except the Air and entry-level iMac now have at least 3 TB4 ports, my approach would be to use a TB4 hub to "concentrate" all the non-bandwidth/latency-critical devices on the hub, maybe alongside one display, to free up host ports for the more critical peripherals (whether it's a top end NVMe drive, ultra-high-bandwidth display or even a USB 2 device - like an audio interface - that preferred not to be on a hub).

I'd love to see some of these docks use a dual TB cable, with one going to the TB switch and one that is essentially a TB passthrough.

The iVANKY FusionDock Max, actually uses both TB4 ports and has (2) TB4 Controllers, allowing massive bandwidth and pretty much any monitor configuration you can think of.

 

[whitby]

This is interesting, what cables were you using to daisy chain the (2) Studio Displays? Also how long are the cables? This feels like a low signal issue.

Sorry I was not clear. I am not daisy chaining the displays, I was using a hub (an OWC Thunderbolt 5) to connect two 5K displays to a single Mini TB 5 port. The cables are identical in length. However, to reiterate, based on the fact that they are not 4 display port streams on a TB 5 port, two 5K displays should not work at all.

 

[joevt]

What is interesting that you can connect 2 x 5K displays to one TB 5 on the Mini and they work but blank occasionally. I am not sure why they work at all.

Apple Silicon Mac mini with two Apple Studio Displays? They should not blank occasionally. That just means poor connection quality or noise.
They work because they each use a single DisplayPort HBR2 x4 connection with DSC.

You can force an LG UltraFine 5K to use a single HBR2 x4 connection but then it would be limited to 5K39 (with a custom timing added to the system).

Thanks, I wasn't aware it used "Dual Tile" SST. Are they essentially alternating every other frame of (2) 30hz mirrored displays?

Minor correction: the displays are dual tile; the connection is dual link SST (not related to dual link DVI). The Apple support documents that mentioned dual link SST can be found on the wayback machine.

For 5K60 dual link SST, the two DisplayPort connections are each sending 2560x2880 60Hz - one for the left half and one for the right half of the display. I think the signals may be synced such that each line from each half are sent nearly at the same time so they can be drawn on the 5120x2880 screen at the same time. I'm not sure how synced they need to be if at all. I suppose one could try sending two unsynced 2560x2880 signals (if you can convince the system that the two connections are not part of a multi-tile display - which would involve modifying the EDIDs of each tile before the system sees them).

 

[theluggage]

USB-A to USB-C adapters are against the USB spec for reasons - even though they do work.

...more honoured in the breach than the observance, I suspect. Lots of USB-C peripherals come with them. The "correct" solution would be to use USB-A to C cables. Pity the USB specs didn't ban devices with captive cables (esp. since a USB-C extension cable wouldn't be 'legal' either...)

Still intrigued by the reason for USB tunnelling if all peripheral chips have USB controllers anyway (perhaps they want to phase these out of future chipsets?) - I'd prefer to use the USB controller in the hub (but it's not worth interposing a TB3 device for!).