ASUS Reveals its USB4 Add-in Card with 60 Watt USB Power Delivery Support

[TheLostSwede]

Back at Computex last year we got a look at the first USB4 add-in card, which was from MSI. Some six months later, ASUS has finally revealed its USB4 add-in card that simply goes by the name of USB4 PCIe Gen4 Card. The general design of the two cards appear to be more or less identical, at least in terms of outputs and inputs. As such, both cards feature two USB4 outputs and two DP 1.4 inputs for those that want to use the USB4 ports to connect to a display. Other inputs include a USB 2.0 header, a custom USB4 header for communication with the motherboard—similar to Thunderbolt add-in cards—and a 6-pin graphics card type power input. The card supports up to two monitors and three devices or one monitor and four devices when daisy chained.

ASUS has gone for a shroud on its card, so it's impossible to make out any real details, but based on the manual, ASUS has installed a heatsink on the ASM4242 USB4 host controller from ASMedia, albeit a smaller one than MSI. However, ASUS has only gone for 60 W USB Power Delivery compared to 100 W for the MSI card. If this makes any useful difference in a desktop PC is up for discussion though and would depend on specific use cases. On the other hand, ASUS allows for 60 W through both ports, whereas MSI only delivers 27 W through its secondary port. The card should work with any ASUS motherboard that has a TB/USB4 header. As ASUS has only just put up the product page, there's no word on pricing.



View at TechPowerUp Main Site | Source

 

[Assimilator]

Looks basically identical to the Thunderbolt 4 cards previously released by various manufacturers. Don't understand why they don't use an 8-pin PEG power connector, which would give 225W total board power and thus allow for dual 100W PD.

 

[HairyLobsters]

Why PCIe Gen4 x4 and not gen3 x8?

 

[evernessince]

Why PCIe Gen4 x4 and not gen3 x8?

Making it x8 would essentially force most people to use up a full x16 slot and not everyone has an extra x16 slot to spare. There are not a lot of motherboards with an extra slot wired for just x8. In addition, secondary x16 slots often split or share lanes with the GPU / M.2 drives.

 

[Assimilator]

Why PCIe Gen4 x4 and not gen3 x8?

Why use more lanes when you could use less?

 

[thomasjpr]

It still just seems wild to me how many modern, current-gen PC desktop motherboards lack more than a single USB-C 3.2 port, much less something that is USB4 or Thunderbolt compliant. I'm at a place with my devices now that I usually only need a single USB-A for something (where I still never manage to nail the correct orientation on the first try), but everything else is USB-C.

PC laptops and Apple devices fully embraced USB-C long ago -- why are desktops so behind that these after-market solutions are the only way to get more than 1 or 2 ports? Yeah, the needed lanes are at a premium with so many other demands (especially for full-spec USB4 and Thunderbolt) but let's figure this out.

 

[TheLostSwede]

Why PCIe Gen4 x4 and not gen3 x8?

Why would a newly developed IC use an older generation of PCIe? Also, the chip was partially designed to interface directly with AMD's Ryzen 7000-series CPUs for best performance.

 

[Ownedtbh]

Can someone explain me, whats the purpose of the two DP's ?

Any usecase scenario would be helpful

 

[TheLostSwede]

Can someone explain me, whats the purpose of the two DP's ?

Any usecase scenario would be helpful

Inputs from your graphics output, in case you want to use the USB4 Type-C ports for connecting to monitors, as I wrote in the news post.

 

[trsttte]

Do these boards also require a special motherboard header like with thunderbolt cards?

Don't understand why they don't use an 8-pin PEG power connector, which would give 225W total board power and thus allow for dual 100W PD.

The problem is not how to get more power to the card, is not wanting to buy and include more expensive PD controllers

I'm all for using power delivery to it's full capabilities but all in all 60W is not that bad, there's not many devices you'd plug to your computer that would need that much power or, if they do, that support being powered that way.

 

[billEST]

Looks basically identical to the Thunderbolt 4 cards previously released by various manufacturers. Don't understand why they don't use an 8-pin PEG power connector, which would give 225W total board power and thus allow for dual 100W PD.

power supply : 250 w more from psu like a second graphic card , and 1 or 2 connector free

thunderbolt card use power supply connector

 

[LabRat 891]

The card should work with any ASUS motherboard that has a TB/USB4 header.

This angers and upsets me.
I thought that req. was specific to Intel Thunderbolt, not USB

Why has every TB3, TB4, and USB4 card required a stupid proprietary header? Since when did we need mobo-level support for a friggin USB card?
Also, a lot of X570 boards lack the header, but not B550 -what the hell?!

Edit:
Can the header req. be worked around? What signals are possibly needed above what a PCIe slot offers? (PCIe already has SMbus/JTAG pins)

edit2:
Looks like it

Thunderbolt 3 with threadripper on a Asrock x399 fatality mo...

I have been doing research and decided to buy the titan ridge thunderbolt card. I installed it in my slot 4 as instructed and connected the pci connec...

egpu.io

 

[Chaitanya]

Is it like those thunderbolt cards that are locked in with every vendor? Also are even going to available in retail? Over half of accessories these motherboard makers make isnt available for purchase anywhere.

 

[TheLostSwede]

Do these boards also require a special motherboard header like with thunderbolt cards?

Maybe try reading the news post before commenting? It's all in there.

This angers and upsets me.
I thought that req. was specific to Intel Thunderbolt, not USB

Why has every TB3, TB4, and USB4 card required a stupid proprietary header? Since when did we need mobo-level support for a friggin USB card?
Also, a lot of X570 boards lack the header, but not B550 -what the hell?!

Edit:
Can the header req. be worked around? What signals are possibly needed above what a PCIe slot offers? (PCIe already has SMbus/JTAG pins)

edit2:
Looks like it
View attachment 330527

Thunderbolt 3 with threadripper on a Asrock x399 fatality mo...

I have been doing research and decided to buy the titan ridge thunderbolt card. I installed it in my slot 4 as instructed and connected the pci connec...

egpu.io

It's possible to use the cards without it, but let's just say it's not recommended.
For starters, you can't hot-swap devices without that extra cable.
It's using various "slow speed" interface that are required to communicate information that isn't sent over the PCIe bus, although I don't know the specific ones, but it's things like GPIO, LPC, SMBus, I2C etc.
So if the card can't communicate over those interfaces as needed, your system might crash when you unplug a drive.

Hacking Thunderbolt Part 2: Retrofitting PCI Hot-Plug to the UEFI – Matt's Tech Pages

Part 1 here https://xpander.mattmillman.com/files/linked/hacking_thunderbolt_part1.pdf
Yes, this is old, but still applies.

There as an old HP card that simply listed the header as TBT GPIO, so it does at the very least rely on that.

Is it like those thunderbolt cards that are locked in with every vendor? Also are even going to available in retail? Over half of accessories these motherboard makers make isnt available for purchase anywhere.

The cards aren't vendor locked in as such, there just isn't a standard pin-out for the USB4/TB4 header, which is stupid.

Pinged someone I known at ASMedia and got a quick reply.
So the extra pin-header is not just about hot swapping, but it's also related to suspend and sleep, which won't work correctly without the extra connector.
This could result in the system crashing when woken up. Apparently there's a certain amount of UEFI integration here as well, so it's not working in the same way as previous USB generations.
This is related to how Intel implemented things with Thunderbolt and as USB4 builds on Thunderbolt 3...

 

[bonehead123]

AsAss may have revealed the card(s), but what they (once again) failed to reveal is how many arms, legs, left testicles, and 1st borne children they are gonna cost us, hehehe

 

[LabRat 891]

It's possible to use the cards without it, but let's just say it's not recommended.
For starters, you can't hot-swap devices without that extra cable.
It's using various "slow speed" interface that are required to communicate information that isn't sent over the PCIe bus, although I don't know the specific ones, but it's things like GPIO, LPC, SMBus, I2C etc.
So if the card can't communicate over those interfaces as needed, your system might crash when you unplug a drive.

Hacking Thunderbolt Part 2: Retrofitting PCI Hot-Plug to the UEFI – Matt's Tech Pages

Part 1 here https://xpander.mattmillman.com/files/linked/hacking_thunderbolt_part1.pdf
Yes, this is old, but still applies.

There as an old HP card that simply listed the header as TBT GPIO, so it does at the very least rely on that.

View attachment 330554


The cards aren't vendor locked in as such, there just isn't a standard pin-out for the USB4/TB4 header, which is stupid.

Pinged someone I known at ASMedia and got a quick reply.
So the extra pin-header is not just about hot swapping, but it's also related to suspend and sleep, which won't work correctly without the extra connector.
This could result in the system crashing when woken up. Apparently there's a certain amount of UEFI integration here as well, so it's not working in the same way as previous USB generations.
This is related to how Intel implemented things with Thunderbolt and as USB4 builds on Thunderbolt 3...

Thanks for the detail(s). Also, AGREED.

This makes me wonder if somewhere back in history, the SMbus on PCIe became 'depreciated in use'?
As far as I knew, PCIe slots were supposed to be able to communicate with the motherboard's Super I/O device. Which, handles a lot/all of those mentioned signals and tasks.
So, I'm still at a loss as to what exactly the header is for; why can't the USB4/TB host communicate over PCIe's SMbus for all those functions?

From what I've seen 'in the industry', I cannot help but feel like this is purposeful for some reason, beyond "technical limitation". (especially, when we look-back @ how Intel/Apple 'handled' ThunderBolt*.)

*To be clear, I'm not insinuating foul play on Asus', ASmedia's, etc.
I'm implying that there are 'left over'/'holdover' proprietary design elements within the TB3-derived USB4 standard, that complicate compatibility/integration.

 

[TheLostSwede]

Thanks for the detail(s). Also, AGREED.

This makes me wonder if somewhere back in history, the SMbus on PCIe became 'depreciated in use'?
As far as I knew, PCIe slots were supposed to be able to communicate with the motherboard's Super I/O device. Which, handles a lot/all of those mentioned signals and tasks.
So, I'm still at a loss as to what exactly the header is for; why can't the USB4/TB host communicate over PCIe's SMbus for all those functions?

From what I've seen 'in the industry', I cannot help but feel like this is purposeful for some reason, beyond "technical limitation". (especially, when we look-back @ how Intel/Apple 'handled' ThunderBolt*.)

*To be clear, I'm not insinuating foul play on Asus', ASmedia's, etc.
I'm implying that there are 'left over'/'holdover' proprietary design elements within the TB3-derived USB4 standard, that complicate compatibility/integration.

From what I was told, it's all going back to how Intel implemented Thunderbolt back in the day and they chose to do it this way for som weird reason. The Thunderbolt cards and now USB4 cards aren't simply plug and play like your normal PCIe card and require a fair amount of UEFI work, which obviously all the motherboard makers have figured out for quite some time by now, but that requires that extra "GPIO" interface or whatever the connector is, which is a bus that isn't possible to do over PCIe. Pin 5-9 is still SMBus/JTAG on PCIe, so that shouldn't be part of the problem.

My own guess here is that Intel didn't intend for add-in cards like what we're seeing when they developed Thunderbolt and that's why we have a janky implementation now.