Do these USB4 expansion cards support all motherboards?
IIRC, all the Thunderbolt expansion cards require extremely specific motherboards with Thunderbolt headers, which make them less expansion cards and more motherboard daughterboards.
All the TB4 cards (gigabyte, asus et al) have Thunderbolt 4 cards that support power delivery via psu power NOT just bus powered. Not sure why this card is so special because it does 100w PD.
Gigabyte GC-MAPLE RIDGE and ASUS ThunderboltEX 4 add-in cards have exactly the same Power Delivery specs. The Gigabyte card even has 2x 6-pin PCIe power connectors so you can fully power the ports without using any slot power.
Why on Earth are vendors designing extra long boards like these?
Plenty boards have an extra x4 slot, x8 is far harder to find or requires bifurcation.... So why?
It's similar with ACQ113 nice, which only require x1 on PCI 4, but nearly always come with a physical x2 connector, which is practically non-existent on Mainboards while underused x1 slots abound...
Seems like they really do not want to sell these or worse!
Power consumption is the likely reason why. 4x and higher PCIe slots are permitted to pull more amps off of the 12V power in the slot. Physical slots do have to adhere to the power requirements even if they don't provide the full bandwidth. For example a physical 16x slot but wires 1x electrically still has to provide up to 75W across the slot.
This card is electrically x4 and mechanically x8; which doesn't get it it anything power wise. x4 and x8 are both 25W from the slot.
Maybe the extra length is just for increased stability to reduce the risk of half-unseating the card while pulling a cable out of the card. A lot of x4 slots are open backed allowing longer cards to be inserted; but a PCB notch to allow it to install in close backed slots would have improved compatibility here.
Yeah, the x8 physical connector is a real puzzlement here. It gains you nothing in terms of power as x4 / x8 / x16 slots all have the same 25 W initial power up, 75 W maximum dissipation limits. It's even odder when you consider that Intel generally requires Thunderbolt add-in cards use PCIe lanes coming from the chipset and not PEG lanes.
For those of you who care about the math, per the PCIe spec, this card can pull up to 5.5 A from the +12V rail (which is really the only one we care about in this case) using the card electromechanical connector and 6.25 A from the 6-pin auxiliary power connector. That works out to 141 W, which is adequate to cover the full 127 W Power Delivery budget as well as conversion losses and other board requirements. The controllers and retimers probably have a combined TDP of around 5 W, so thermal dissipation isn't an issue here regardless of form factor.
This card, like all of the current Thunderbolt 4 / USB4 add-in cards is almost certainly using the Intel JHL8540 "Maple Ridge" Thunderbolt 4 controller, which is a dual-port PCIe Gen3 x4 device. The only thing that sets it apart is that MSI opted for USB4 certification instead of Thunderbolt 4 certification. If they're actually using a different controller, especially if it were one not sourced from Intel, this card would indeed be newsworthy.
If so, then there are so many questions like: what motherboards/chipsets are supported, will it have dma protection, what are the maximum transfer speeds, what benefits, if any, are there over maple ridge.
Oh, wow, fantastic. Finally a third-party discrete Thunderbolt host controller option! Somehow I missed the news about this a couple weeks ago.
DMA protection will, of course, be OS and UEFI dependent, but Windows 11 Kernel DMA Protection should work.
The ASMedia ASM4242 provides a PCIe Gen4 x4 connection to the host, whereas Maple Ridge is only Gen3 x4. And the ASM4242 also supports USB 3.2 Gen 2x2 (SuperSpeed USB 20Gbps) and Thunderbolt 3 Alt Mode for Thunderbolt interoperability. So this definitely looks to have an edge over Maple Ridge, offering capabilities more in line with recent integrated Thunderbolt 4 implementations.
The ~10W of 5V isn't useless. It's needed for back compatibility with low end USB devices. Not what you'd normally use a card like this for, but the support needs to be there. I think the initial negotiations prior to the switch into high power 12V mode probably needs to be done at 5V too (to avoid frying 5V devices with 12V).
There's no 5 V supply rail for PCIe cards though, just 3.3 V and 12 V. There is 3.0 A @ 3.3 V = 9.9 W available, and you're right in that it can be useful, but it isn't being used by the Type-C power path.
The Type-C ports absolutely have to support the standard 5 V Vbus at 500 mA when connected to USB 2.0 devices, 900 mA for USB 3.x single-lane devices, and 1.5 A for USB 3.2 dual-lane and USB4 devices. Additionally, the USB PD spec requires the 27 W rated port to supply 5 V at up to 3 A and 9 V at up to 3 A. The 100 W port must do likewise and additionally supply 15 V at up to 3 A and 20 V at up to 5 A. However, all of this is being sourced from the 12 V supply rail and converted to the appropriate levels by a buck-boost regulator on the card. The PD controller is likely capable of supplying a very broad range of outputs to support explicitly negotiated power contracts as well, e.g. 3-21 V in 20 mV steps and 0-5 A in 50 mA steps.
Aren't all the other options tied to motherboards with specific headers/support, which are generally only available for *Intel* platforms? This could be huge for unlocking Thunderbolt for AMD boards that don't support it. Does it support passthrough to external PCIe enclosures?
MSI's press release has a section titled "Intel Z790 MAX Series Motherboard, PROJECT ZERO Back-Connect Motherboard and Expansion Card" that says "MSI brings a unique new motherboard, starting with the popular back-connect motherboard - B650M PROJECT ZERO." talks about the motherboard for a bit, and then says "And also on display is the new USB 4 100W Expansion Card, which allows the motherboard to support dual USB 4.0 40Gb/s Type-C Ports with ultra-high transfer rate and up to 100W fast charging!"
So, that really seems to sound like this is just yet another expansion card tied to a specific motherboard, and not a general purpose Thunderbolt/USB4 add-in card. :-(
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
19 Comments
Back to Article
ikjadoon - Wednesday, May 31, 2023 - link
Do these USB4 expansion cards support all motherboards?IIRC, all the Thunderbolt expansion cards require extremely specific motherboards with Thunderbolt headers, which make them less expansion cards and more motherboard daughterboards.
Kevin G - Wednesday, May 31, 2023 - link
Looking at the picture in the article, the card has a TB header on it to connect to the motherboard.abufrejoval - Wednesday, May 31, 2023 - link
ACQ NICs not nice, need editCharonSin - Friday, July 14, 2023 - link
Excellent point.herozeros - Wednesday, May 31, 2023 - link
All the TB4 cards (gigabyte, asus et al) have Thunderbolt 4 cards that support power delivery via psu power NOT just bus powered. Not sure why this card is so special because it does 100w PD.repoman27 - Thursday, June 1, 2023 - link
^This.Gigabyte GC-MAPLE RIDGE and ASUS ThunderboltEX 4 add-in cards have exactly the same Power Delivery specs. The Gigabyte card even has 2x 6-pin PCIe power connectors so you can fully power the ports without using any slot power.
abufrejoval - Wednesday, May 31, 2023 - link
Why on Earth are vendors designing extra long boards like these?Plenty boards have an extra x4 slot, x8 is far harder to find or requires bifurcation.... So why?
It's similar with ACQ113 nice, which only require x1 on PCI 4, but nearly always come with a physical x2 connector, which is practically non-existent on Mainboards while underused x1 slots abound...
Seems like they really do not want to sell these or worse!
LiKenun - Wednesday, May 31, 2023 - link
Thunderbolt 4 requires 4 lanes of PCIe 3.0 bandwidth. Thus, 8 lanes are required to support 2 ports.The real question is why they aren’t designing them to use 2 PCIe 5.0 lanes?
Kevin G - Thursday, June 1, 2023 - link
Power consumption is the likely reason why. 4x and higher PCIe slots are permitted to pull more amps off of the 12V power in the slot. Physical slots do have to adhere to the power requirements even if they don't provide the full bandwidth. For example a physical 16x slot but wires 1x electrically still has to provide up to 75W across the slot.DanNeely - Thursday, June 1, 2023 - link
This card is electrically x4 and mechanically x8; which doesn't get it it anything power wise. x4 and x8 are both 25W from the slot.Maybe the extra length is just for increased stability to reduce the risk of half-unseating the card while pulling a cable out of the card. A lot of x4 slots are open backed allowing longer cards to be inserted; but a PCB notch to allow it to install in close backed slots would have improved compatibility here.
repoman27 - Thursday, June 1, 2023 - link
Yeah, the x8 physical connector is a real puzzlement here. It gains you nothing in terms of power as x4 / x8 / x16 slots all have the same 25 W initial power up, 75 W maximum dissipation limits. It's even odder when you consider that Intel generally requires Thunderbolt add-in cards use PCIe lanes coming from the chipset and not PEG lanes.For those of you who care about the math, per the PCIe spec, this card can pull up to 5.5 A from the +12V rail (which is really the only one we care about in this case) using the card electromechanical connector and 6.25 A from the 6-pin auxiliary power connector. That works out to 141 W, which is adequate to cover the full 127 W Power Delivery budget as well as conversion losses and other board requirements. The controllers and retimers probably have a combined TDP of around 5 W, so thermal dissipation isn't an issue here regardless of form factor.
This card, like all of the current Thunderbolt 4 / USB4 add-in cards is almost certainly using the Intel JHL8540 "Maple Ridge" Thunderbolt 4 controller, which is a dual-port PCIe Gen3 x4 device. The only thing that sets it apart is that MSI opted for USB4 certification instead of Thunderbolt 4 certification. If they're actually using a different controller, especially if it were one not sourced from Intel, this card would indeed be newsworthy.
Exotica - Thursday, June 1, 2023 - link
I believe this card is not maple ridge, but instead finally signals the arrival of a maple ridge competitor: the asmedia4242 usb4 chip.https://www.asmedia.com.tw/product/e20zx49yU0SZBUH...
If so, then there are so many questions like: what motherboards/chipsets are supported, will it have dma protection, what are the maximum transfer speeds, what benefits, if any, are there over maple ridge.
Exotica - Thursday, June 1, 2023 - link
https://www.youtube.com/watch?v=J08h3wz_ma8&t=...repoman27 - Thursday, June 1, 2023 - link
Oh, wow, fantastic. Finally a third-party discrete Thunderbolt host controller option! Somehow I missed the news about this a couple weeks ago.DMA protection will, of course, be OS and UEFI dependent, but Windows 11 Kernel DMA Protection should work.
The ASMedia ASM4242 provides a PCIe Gen4 x4 connection to the host, whereas Maple Ridge is only Gen3 x4. And the ASM4242 also supports USB 3.2 Gen 2x2 (SuperSpeed USB 20Gbps) and Thunderbolt 3 Alt Mode for Thunderbolt interoperability. So this definitely looks to have an edge over Maple Ridge, offering capabilities more in line with recent integrated Thunderbolt 4 implementations.
DanNeely - Friday, June 2, 2023 - link
The ~10W of 5V isn't useless. It's needed for back compatibility with low end USB devices. Not what you'd normally use a card like this for, but the support needs to be there. I think the initial negotiations prior to the switch into high power 12V mode probably needs to be done at 5V too (to avoid frying 5V devices with 12V).repoman27 - Friday, June 2, 2023 - link
There's no 5 V supply rail for PCIe cards though, just 3.3 V and 12 V. There is 3.0 A @ 3.3 V = 9.9 W available, and you're right in that it can be useful, but it isn't being used by the Type-C power path.The Type-C ports absolutely have to support the standard 5 V Vbus at 500 mA when connected to USB 2.0 devices, 900 mA for USB 3.x single-lane devices, and 1.5 A for USB 3.2 dual-lane and USB4 devices. Additionally, the USB PD spec requires the 27 W rated port to supply 5 V at up to 3 A and 9 V at up to 3 A. The 100 W port must do likewise and additionally supply 15 V at up to 3 A and 20 V at up to 5 A. However, all of this is being sourced from the 12 V supply rail and converted to the appropriate levels by a buck-boost regulator on the card. The PD controller is likely capable of supplying a very broad range of outputs to support explicitly negotiated power contracts as well, e.g. 3-21 V in 20 mV steps and 0-5 A in 50 mA steps.
richardnpaul - Monday, June 5, 2023 - link
"Meanwhile the card's second Type-C port can deliver up to 27 Watts, which is enough for smartphones and other mid-power periphreals."My 150W SuperVOOC charged phone would like a word.
hubick - Wednesday, June 7, 2023 - link
Aren't all the other options tied to motherboards with specific headers/support, which are generally only available for *Intel* platforms? This could be huge for unlocking Thunderbolt for AMD boards that don't support it. Does it support passthrough to external PCIe enclosures?hubick - Tuesday, June 13, 2023 - link
MSI's press release has a section titled "Intel Z790 MAX Series Motherboard, PROJECT ZERO Back-Connect Motherboard and Expansion Card" that says "MSI brings a unique new motherboard, starting with the popular back-connect motherboard - B650M PROJECT ZERO." talks about the motherboard for a bit, and then says "And also on display is the new USB 4 100W Expansion Card, which allows the motherboard to support dual USB 4.0 40Gb/s Type-C Ports with ultra-high transfer rate and up to 100W fast charging!"So, that really seems to sound like this is just yet another expansion card tied to a specific motherboard, and not a general purpose Thunderbolt/USB4 add-in card. :-(