HP Thunderbolt Dock 280W G4 w/Combo cable review and teardown

[Comparison of 80 Thunderbolt 4 and USB4 docks]

---

This article, by Leaves, has been translated and re-posted with permission.

[Original article link]

[Buy the HP Thunderbolt 4 G4 Dock on Amazon] (affiliate link)

---

Introduction

It’s been a long time since my last teardown post, and this is the first time for me to disassemble a Thunderbolt 4 dock. For this post, I’ve used a miniature photo studio to improve the photo quality as well as picture content of my topology speculation.

[Teardown of HP’s last gen Thunderbolt G2 Dock]

As one of the three main PC manufacturers, HP’s ability to design related accessories is indeed first-class. There are already many HP docking station teardown posts in my column. From the actual internal structure or use experience, HP products do imply a certain level of reliability. However, HP also likes to stuff some sub-optimal features into current products for some known and unknown reasons:

1. The stupid naming pattern, this is not only a problem with the docking station, but other HP product lines, here I won’t go into details. [Editor’s note – G5, G4, G3 and G2 are confusing because each higher number is NOT a functional superset superset. And while this G4 dock is Thunderbolt 4, there is also a G4 USB-C only model that isn’t even USB4. The G2 dock can be either Thunderbolt 3, USB-C or USB A and the G3 and G5 are USB-C only.]
2. The HP 7.4mm DC power jack has the same specification as the Dell’s, but only HP’s own power supply is allowed to use with HP’s special recognition mechanism. Last time I tested the G2 Dock, this feature was very annoying that even a DC adjustable power supply can’t power the dock.
3. The new docking station has no audio module and one cannot be added by the user.
4. This time, the power brick is really bigger than the docking station itself.

HP’s official promotional video

HP’s spec sheet:

Compared to the previous generation G2 Dock, some features have been cut:

1. 3.5mm audio port
2. Top replaceable module for audio etc.

Changed/added some features:

1. The VGA port is changed to HDMI 2.0, which is good for the external monitor use.
2. The original Thunderbolt 3 port was upgraded to Thunderbolt 4.
3. The original Gigabit Ethernet port has been upgraded to the 2.5Gb/s version, and it has become a dual-NIC design with backwards compatibility for USB-C systems.

And… some weird parameters that didn’t change:

1. In theory, the single JHL8440 controller for Thunderbolt 4 devices can provide enough downstream TB4 ports to replace the original downstream DP Alt Mode USB-C port, but there is none included.
2. In the G2 Dock test, it was clarified that all USBs on the bottom board of the G2 are supported by VL820, all of which are 10Gbps, so the manual identification is wrong. It seems that the G4 manual still has this error.

The compatibility of the host port is still marked. It can be seen that, like the G2, it is compatible with the DP Alt Mode host port and the Thunderbolt port; however, a distinction is made between 2Lane Mode and 4Lane Mode. I don’t know if it means G4 could auto switch in two modes or it is just a detailed distinction based on the compatibility mode of the host port.

Predicted design

The basic structure of Thunderbolt naturally needs to be changed to suit the form of JHL8440, but there are only two Thunderbolt ports in the G4 Dock, and the PD controller shall also be reduced to one. The base structure should consist of CYPD5235, JHL8440 and FL5801, although the possibility of using a Texas Instruments PD controller exists.

The rest of the structure should follow the last gen product:

• There should be an MCU as the management device of the Dock
• At least two USB hubs, one 10Gbps and one 5Gbps
• A DP MST solution which provides at least two DP1.4 and one HDMI 2.0 ports downstream; still likely to be a Synaptics solution
• Dual network chips, one 1G and one 2.5; the 1G network card should be Realtek USB, and the 2.5G network card should be Intel PCIe. There will be corresponding multiplexing structure switching between them.

• DP Alt Mode USB-C port, since it is clearly marked that the function is mutually exclusive with the adjacent DP port, the possibility of TB4 port mislabeling is directly excluded. It should be implemented by an independent MUX and PD controller as in the G2 Dock.

• The rest of the ICs used for docking station management and detection are not predicted

So the corresponding specific IC predictions are:

• It is possible to use Cypress’s CY7C series as the docking station controller, or TI’s solution. Of course, it is not impossible to use an independent MCU
• G2 uses VL820 as 10Gbps USB hub and USB5807 as 5Gbps USB hub. Although there is a possibility to keep the design, it is uncertain.
• Should still be Synaptics VMM series, the part number is unknown
• Realtek’s USB network card should use RTL8153, Intel’s i225 is more likely to be used just like Caldigit TS4; the multiplexing structure is unknown
• The possibility of using G2’s PS8802+PS8468 and Cypress PD controller design is high

Unboxing

It is disappointing that the the audio module extension has been removed, HDMI is still at version 2.0, and there is still only one downstream TB4 port. Although I am quite interested in HP’s design and structure, I am not very happy with the total feature set. However, the HP G4 has at least some bright spots, and I have no desire to get an alternate.

It comes with the 230W combination cable and the massive power brick is also a high-end 280W version.

G2 vs G4.

G2 with conference audio module vs XBOX Elite 2 controller vs G4 for size comparison.

It can be seen that due to the cancellation of the module replacement structure, the air outlet of the G4 doesn’t have a screw in the middle.

Teardown

The G4 Dock seems to be basically following the structure of the G2. The foot pad bottom shell can be detached after removing the set screws.

However, the mechanism for fixing the wire tail has been changed compared to G2. In G4, it has become a mechanism that can be rotated by two resin hinges. In G2, it is a very direct tooth-like combination.

Remove the 4 Phillips bottom case screws and remove the bottom case. Except for some differences in the stickers, there is really no obvious difference with the G2.

When further dismantling, first remove the fixing screws on the lower part, take out the double PCB – metal heatsink as a whole, and then disassemble further, otherwise it will be very troublesome.

When it is separated, there is a 4-pin thin wire connected to the button module fixed on the shell. It has to be said that the connector is much simpler after the function is cut.

The overall close-up of the interior is still the same as the G2 PCB – metal heatsink, a sandwich structure. But what is interesting is that because the video output port adopts the interface package of DP stacked DP, the shape of the iron cooler has also been adjusted accordingly. The eDP cable and the columnar docking interface are still very familiar just like the G2, the blowing direction of this fan is confusing but is probably enough to get some airflow.

In the subsequent disassembly process, the steps taken for the G2 are repeated. The top board in the docking station is called the Top board, and the bottom board is called the Bottom Board. First separate the top fan and IO shield. As soon as the fan is removed, you can see a shield and MUX facing the network port.

Detach the Bottom Board. Compared with G2, it is a little simpler, and the three inter-board interfaces are separated. Since the eDP is fixed with tape, it is not separated here. The operation of adding a cooling pad to the sticker here is a bit confusing. The pink thermal pad that was torn and folded has dried. It appears it was like this when it left the factory, not due to disassembly.

Separate the top board, you can see that the DC port is still not soldered on the board, but fixed on the metal heatsink; the DC port extends two cables, and a group of black and red multi-wire interfaces are connected to the bottom board, which should be the main power supply function. Another thin cable connected to the top plate should be the identification mechanism responsible for the power supply; The Power cable is also divided into two connectors in G2, but both are connected to the Bottom Board.

Separate the metal heatsink.

Fake re-assembly after separation

Remove the shield from the top board.

PCB Close-up

Bottom Board

Top Board

IC Tracking

I will not pay attention to MOSFET, Flash, and some peripheral power devices, but will only examine the main functional ICs.

PCB Details

Bottom Board

In general, compared to the PCB of the G2 dock, the G4 has little difference in the layout of the main components, but the high-speed signaling is mostly moved to inner layers which makes tracking them difficult. You can clearly see the high-speed differential pair wiring of the JHL8440 to the downstream TB4 port and the associated coupling capacitors and diode protection; the slightly larger TVS above should be used for the USB2.0 interface protection. However, the wiring to the upstream TB4 interface on the other side of the board is basically untraceable.

The traces from JHL8440 to VMM5323 cannot be checked to the details on the surface, and only part of the downstream port wiring can be seen; it can be seen that one set of DP wiring is a straight-through dual DP female socket, and the other set (an extra pair of differential lines is uncertain – whether it is AUX or other functions) point to a MUX. It is determined by the operation of the MUX whether to connect the DP port or the Multi-function USB-C port.

The most prominent part of the Bottom Board should be this empty soldering position.

Obviously it is an audio design. On the G2, this part is the RealTek ALC4102. On the G4, the layout has been adjusted – beside the 3.5mm seat is no longer a crystal oscillator, but it seems to be an 8pin Flash package.

That is to say, the audio design of G4 should have a high degree of completion. I don’t know at which stage in design it was abandoned for what reason. It may be the cause of the interference as questioned in the G2 teardown, but who knows the specifics? [Editor’s note – perhaps it is a supply-chain issue?]

There are 4 built-in onboard interfaces in total, and the estimated functions are:

1. Connect the DC input interface to receive power
2. Connect the hard connector of the Top Board to transmit power to the Top Board
3. Connect the cable of the Top Board, presumably the downstream USB cable of the USB7206, to provide a USB interface to the application on the Top Board
4. A cable similar to eDP, connected to the Top Board, presumably including a PCIe transmission line, to provide an interface to PCIe-related applications on the Top Board

Top Board

There are 5 pairs of differential lines coming out of the eDP interface. The top two pairs are directly connected to the i225 network card. The pair on the left is equipped with a coupling capacitor, which seems to be the RX TX line of PCIe X1; the clock line of PCIe is not very clear.

The other three pairs of differential lines… are connected to the USB5744.

In terms of network card, before the LAN port MUX, it is clear that 4 pairs of differential lines are connected to i225, and the differential pair indicated by the yellow arrow is connected to RTL8153.

The other three differential lines connected to the RTL8153 can also be seen on the back. Downstream of the MUX, you can see the three pairs of differential lines connected to the network port protection IC, but the fourth pair is hard to find.

The USB5744 is connected to the RTL8153’s USB link, indicated by the yellow arrow.

In addition to several large connectors between the two boards, the Top Board also has some built-in onboard connectors:

1. Connect to the 4pin socket of the switch/illuminated indicator module on the top of the housing
2. The 5-pin interface, not used in G4, is similar to the position of the VGA output socket of the previous generation G2, but the number of VGA socket pins of G2 is much more, in short, the function is unknown
3. Connect to the HP identification communication part of the DC port

And there is an interface for the fan on the other side.

I can’t figure out the connection logic of CY7C65219, but theoretically it should be connected to the USB 2.0 link provided by the Bottom board.

Topology Speculation

In order to look neater, I have omitted the USB2.0 connection related to the FL5801. In addition, I have no evidence to confirm which device is connected upstream of the USB5744 on the Top Board. The connection logic of the CY7C65219 Dock management solution is also uncertain. It is more reliable to speculate that this IC should also control a considerable part of the Bottom Board’s functions through the adjacent board to board cable; the fan, the switch module on the top of the housing, and OnSemi buck-boost controller should also be under the management of this IC.

Functional Verification

A new machine is used in this test. The test is divided into two parts, namely the Thunderbolt operation mode and the DP Alt Mode operation.

PC specs:

• Mainboard: MSI X299 Creator
• Processor: i9 7980XE
• Graphics card: 2080ti
• Thunderbolt expansion card: GC-Titan Ridge

DP Alt Mode testing is powered by the native USB-C on the 2080ti. Although I don’t have a Thunderbolt 4 interface platform available, Titan Ridge is sufficient for testing as provided by the expansion card.

USB Performance Verification

Using a Western Digital P50, a portable SSD that supports USB3.2 Gen 2×2, to test… as long as the read and write test can exceed 5Gbps, it can be assumed to be a 10Gbps USB port, and if it cannot exceed 5Gbps, it is a USB3.2 Gen1 hub.

The ideal read and write speed close to 10Gbps can be obtained by testing on all Bottom Board USB ports.

And the ports on top board is close to 5Gbps

Obviously… G4 USB design structure is still inherited from the G2, but the labeling errors in the manual are also inherited.

Display Functional Verification

Thunderbolt 3 Host

I have limited materials on hand so unfortunately I will only do the simplest tests. The USB-C port of this portable screen only supports the 4Lane DP1.2 mode and the display supports up to 4K 60Hz. In theory, when the G4 is in Thunderbolt mode, the effect of connecting the screen to the multi-function C port or the downstream TB4 port is the same (both 4K@60Hz). However, in fact 4K 60Hz display can be achieved when connected to the downstream TB4, but only 4K 30Hz can be set under the multi-function USB-C port. Theoretically this problem is caused by VMM5323 compatibility.

Mobile Phone Host

Here is the test applied with Samsung S20. After linking the G4 to the phone, the P50 portable SSD and monitor are connected. The monitor can trigger Samsung DEX normally, and the S20 can also read the contents of the portable SSD, but the display is limited to 1080p 60Hz, which should be the limitation of the phone itself.

LAN Port Functional Verification in Dual Mode

In Thunderbolt mode

In HWinfo, you can easily observe the enumerated network card under the Thunderbolt Bus. Although HWinfo does not directly display the device name, it shows i225 in the device driver section.

In DP Alt Mode

G4 is no longer in the Thunderbolt state, and main devices are transferred to mount under the USB Root hub. In this state, only the Realtek network device is active, and it hangs under the USB5734, but the paradox is why the RTL8153 can’t even be seen in the thunderbolt mode? It seems that in addition to the LAN port MUX, HP should have other designs for dual-mode NIC operation

Power Supply to Host Verification

I don’t have a load device here for the time being, and I don’t have a notebook with enough power consumption for testing. In theory, it is still necessary to carry out PD messages detection and actual load testing in order to reach a more reliable conclusion. Here we can only catch PD broadcasts in a simple way.

Protocol sniffing on the host port by YZX Studio ZY1280… well as a notebook-only docking station product that only supports PD, there is nothing to blame, but only 75W is a little unexpected. If the ZY1280’s investigation is true, HP may still think that there is no need to give so much USB-C power if the DC plug is combined anyway? I don’t know if there is a difference in the PD configuration on the G4 120W Version. [Editors note: Hopefully the 120W version supplies at least 85W to the laptop]

Dock Topology Observation

In general, we use HWinfo to probe the hardware structure of G4, mainly in terms of the USB hubs and the basic design of Thunderbolt dock. HWinfo is still not very reliable for relatively complex structure exploration so a teardown is still needed.

DP Alt Mode

All major devices are mounted under the 2080ti USB controller, and HWinfo reports major devices named as follows:

For the device PID=800A at the beginning, HWinfo did not detect the specific device name. According to the tree structure parallel to FL5801, it can only be determined that the device is JHL8440. In USB mode, it only acts as SuperSpeed+ hub. Hub functions of USB2.0 is completely handed over to FL5801.

At this time, I connected the P50 portable SSD to the downstream TB4 port, and it will appear in the Port1 position under the device PID=800A. Therefore, it is reasonable to speculate that JHL8440 in DP Alt Mode has 4 subordinate interfaces; Port 1~3 are the original downstream TB4 ports, and Port4 is the native 10Gbps USB 3.2 interface (here is connected to USB7206).

It is also worth noticing that the USB hub mounted on USB7206 Port2 is recognized by HWinfo as USB5734, but in the actual teardown, the device is USB5744. According to the identification principle of USB devices, the device name shall be provided by the device during the enumeration process. This name problem should be caused by MicroChip, who might use the USB5734’s firmware content for the USB5744.

The USB4206 at the lower position, was not found during the disassembly process. In theory, this is the USB2.0 part of the USB7206 that is connected to the FL5801 and enumerated. In fact, MicroChip does not have a product named USB4206.

Conventional Super Speed USB hubs can actually be regarded as composed of SuperSpeed and High Speed USB hub combined. It is precisely because of the unique design of JHL8440 which separates SuperSpeed and High Speed USB, that the downstream USB hub is also separated into USB3. x and USB2.0 – this is rarely seen on conventional USB hub products.

Continuing on, there is also a USB2.0 hub with PID=FF01 without a device name under USB4206, which should be the USB2.0 part of USB5744.

Note that RTL8153 is mounted on Port 1 under USB5744.

The device that is mounted on USB4206 port3 and recognized as HP Thunderbolt G4 should theoretically be CY7C 65219.

Thunderbolt Mode

HWinfo captures some interesting changes in topology when the G4 is switched to Thunderbolt mode. Unfortunately, the JHL8440 is still not recognized by HWinfo. I don’t know much about the Thunderbolt bus topology, and there are still some problems with Titan Ridge’s Bios support settings, so I can only do some preliminary interpretations.

The USB tree structure is basically the same as the previous DP Alt Mode, but the RTL8153 originally under the USB5744 is gone; the P50 is still mounted on the downstream TB4 interface, and it is still located at the Port 1 position under PID=800A. Although PID=800A should belong to the 10Gbps USB hub mechanism of JHL8440, more mechanisms of JHL8440 emerged in HWinfo when in Thunderbolt Mode. It is reasonable to think that the hub mechanism for Thunderbolt/USB4 Bus and PCIe are separate from the USB hub mechanism we mainly observe, but the related investigation work will not be carried out here.

In addition, there is an unexpected feature. Whether in Thunderbolt Mode or DP Alt Mode, after connecting the P50 to the front USB-C port adjacent to the host cable (as shown in the figure below), the P50 is determined by HWinfo to be mounted on Port 3 under the device PID=800A.

When the P50 was previously mounted on the only downstream TB4 port of G4, HWinfo showed that its position was Port 1 under PID=800A, so in theory, Port 3 is a peer position, which is the downstream TB4 interface directly provided by JHL8440.

So is this actually another TB4 interface?

Sadly no. It is a pity that this USB-C port does not show the functional response that a TB port should have by mounting the monitor and other Thunderbolt devices. The USB-C port should be a downgraded USB port from the downstream TB4 port of JHL8440. Although it is equipped with a PD controller to support advanced power supply capabilities, it is very disappointing that there is no thunderbolt function. [Editor’s note: this could be an artificial firmware limitation to enable upselling a customer to a higher-capability dock or could be to workaround for signaling issues they had with the PCB]

The revised topology is as follows:

Conclusion

Some of the problems left over from the G2 remains in the G4. After the Thunderbolt controller was replaced with JHL8440, there are new problems in G4:

• The top board still has to use the 5gbps USB host, difficult to accept this.
• It is a pity regarding the downgraded TB4 port. The multi-function USB-C that is hung under the USB7206 and VMM5323 may not be as easy to use as the TB4 port.
• Although the use of the lackluster downstream PCIe x1 is worthy of recognition, but only one i225 2.5G network card is attached. Seriously, why not hang a 10GbE one, it’s better than the i225 even if true 10Gb/s speed can’t be fully reached.
• This network card switching is quite acrobatic, but the positioning of the G4 dock is still placed in the conference room, whoever does the presentation and who will pick it up; this kind of docking station that tends to be used for public purposes uses i225 that still requires additional drivers; someone who decide to use the i225 was probably kicked in the head by a donkey [editor’s note: hahahaa]
• HDMI 2.0 is provided again, but not HDMI 2.1

The basic Thunderbolt structure of G4 is similar to the Caldigit TS4, but Caldigit’s functionality is still somewhat insufficient; G4 is clearly the leader in terms of functionality. In contrast, the functions of other public TB4 docking stations is too simple; compared to the Lenovo TB4 docking station or Dell’s WD22TB4, the G4 has certain advantages.

It’s just that people like me don’t really need such a powerful dock; I still keep my original opinion. The regular Intel model TB4 docking station is not as affordable and easy to use as the Thunderbolt G2 docking station, let alone this G4. But for me, the elimination of of the extension of the top module means it is less interesting overall.

After understanding the internal structure, it is enough to end my interest in the G4.

---

[Buy the HP Thunderbolt 4 G4 Dock on Amazon] (affiliate link)

Or check out 80+ other Thunderbolt 4/USB4 docks