It depends on both the dock model and the system.  The D6000 uses a DisplayLink chip, so it actually doesn't connect to any GPU directly.  Instead, it uses "indirect display" technology where a driver causes the CPU and GPU to render video data meant for the DisplayLink display, then compress it and send it as regular USB data traffic to the DisplayLink dock/adapter, which then decompresses it converts it to native video for output to the displays.  If you're thinking this isn't an ideal solution, especially for gaming, you're absolutely right.  Making matters even worse is that due to a Windows limitation, indirect displays can only be driven by the system's primary GPU, which will be the Intel GPU in a multi-GPU system, so the dGPU would not be able to accelerate any content on DisplayLink-attached displays.  DisplayLink is really intended for PCs that don't have any native docking or display outputs (or to add more displays to a system that's already using all available display outputs), and it usually works fine for typical business productivity use cases, but it basically falls apart for gaming and even full screen video applications.  The compression can create video artifacts, motion judder, and of course increased CPU/GPU utilization, and that gets worse whenever lots of the display area is changing at once and/or your USB bus is already saturated with other traffic, such as a large file transfer.

The WD15 and TB16 both actually connect to a GPU, but which GPU depends on how your system is wired, and I'm not sure about the AW17 R4.  In most multi-GPU systems on the market, the Intel GPU is the only one physically wired to the display outputs, and the dGPU when active works as a render-only device, passing completed video frames to the Intel GPU for output to the displays.  If you have this type of setup, then the dGPU will be able to accelerate content being output to the WD15 and TB16, with some limitations.  The limitations are that there are certain technologies that may be supported by your dGPU that only work if the dGPU is physically wired to the display outputs and therefore will NOT not work in the "Intel GPU passthrough" scenario.  These include G-Sync, 5K resolution, stereoscopic 3D, and VR.  There may be others.

The other possibility is that the dGPU is physically wired to the USB-C/TB3 port.  In that case, any displays attached to that port (or attached to a dock that itself taps into the native GPU connection on that port) will be driven directly by the dGPU, in which case you'd have no functional limitations at all.  The reason this design isn't used in most cases is because it means that the dGPU has to be running whenever a display is attached to any outputs that are physically wired to it, even if you're not running GPU-intensive tasks, and that takes a battery life penalty.  But this design IS sometimes found on gaming-oriented systems because the assumption is that gamers will care more about things like VR, G-Sync, etc. than battery life.  I just don't know if that's true of the AW17 R4.

Hopefully that helps.

In addition to all of the above, if you want to find out how the AW17 R4's ports are wired, you or another owner can run a pretty simple test.  Get a USB-C to DisplayPort or HDMI cable/adapter and connect it to an external display, then go to NVIDIA Control Panel.  Somewhere in there, it will show all of the displays on the system and indicate which GPU each of them is physically wired to.

The TB16 and WD15 will behave identically with respect to the GPU acceleration that's available.  The TB16 will support up to dual 4K 60 Hz displays and some triple display configurations, whereas the WD15 will only do dual displays up to 1080p or a single 2560x1600 display (or 4K but only at 30 Hz).

Sounds like I should stick with the TB3 interface, then?

Do you mean “the same” as in that the GT1060 won’t output over Thunderbolt either?

I’m wondering what the point of the high pricing of TB3 solutions boasting multi-display 4k outputs for creative arts and high end gaming markets are, if the solution itself is ineffectual in utilising high end graphics...

Unless I’m still missing something.

I understand TB3 eGfx solutions... but not 4k output only devices...

Who spends £300 to output their state of the art laptop GFX through discrete GPU’s to miltiple displays so they can view webpages, play with 1080p movies and spreadsheet simultaneously?

I also have a 17r4 with gtx1060.  If you want to run an external monitor I think you will get the best results if you connect it to your DisplayPort.  

You can only utilize your 1060 though either your hdmi or DisplayPort.  If you try any other ports like your tb3 it will only go through your intel 630 integrated graphics.

Your tb3 and usb-c ports will work fine for other things like powered usb hubs and external storage.  Just not anything video related.

Ok. So most convenient solution for me would be HDMI plus a USB-C dock for attaching multiple high speed devices and ethernet etc?

Yes, that should work fine.  You should always use either your hdmi or DisplayPort for external monitors so they will then be directly connected to your gtx1060.  Just note that if you use hdmi most monitors are then limited to 60Hz max re-fresh rates.  Most monitors today also have a DisplayPort and this is better than hdmi anyway.  However, for your 17r4 you will need a DisplayPort to mini DisplayPort cable and/or adapter because your 17r4 has a mini DP port.

---

@SampleX wrote:

Do you mean “the same” as in that the GT1060 won’t output over Thunderbolt either?

I’m wondering what the point of the high pricing of TB3 solutions boasting multi-display 4k outputs for creative arts and high end gaming markets are, if the solution itself is ineffectual in utilising high end graphics...

Unless I’m still missing something.

I understand TB3 eGfx solutions... but not 4k output only devices...

 

Who spends £300 to output their state of the art laptop GFX through discrete GPU’s to miltiple displays so they can view webpages, play with 1080p movies and spreadsheet simultaneously?

 

---

You CAN use the dGPU with Thunderbolt and USB-C solutions as long as they don’t use a DisplayLink chip. There might be certain features that are unavailable depending on your system’s internal wiring, but dGPU acceleration will be available. If you’re still confused, reread my first reply more carefully. I already spelled all of this out there.

(And on a somewhat related note, when you ask for help on a public forum and people volunteer their time to write replies, especially detailed replies, adding at least a simple “thank you” to your own follow-up posts wouldn’t be completely out of line, rather than just asking more questions and expecting more volunteered time to provide answers.)

Unfortunately I don't think that you can use the tb3 usb-c to bypass the iGPU with a 17r4.  That's just the way it's wired.  I have tested this before with a usb-c to hdmi adapter connected to my tb3 and checking the nvidia physix tab it only shows it connected to the iGUP (intel 630), not the dGPU (1060). 

So, the only connections you can use to utilise the 1060 with a 17r4 for external monitors is to plug them into either the hdmi or mini DisplayPort (or both).  the DisplayPort should support 4k 30hz, maybe 60hz depending on resolution.

The tb3 usb-c port on the 17r4 is still very useful for other applications like fast mass storage.  No sense using it for ethernet because there is already an ethernet port on the 17r4.

If you need more graphics processing power you may want to consider using an Alienware Graphics Amplifier with a GTX1080 founders edition GPU.