Late 2016 15" MacBook Pro - i7-6700HQ/Radeon Pro 450 dGPU/HD Graphics 530 iGPU/16GB RAM/256GB SSD
macOS 10.13.2 - It was plug and play so nothing to report here.
Windows 10 1709 - I performed the same procedure as detailed in my Boot Camp setup guide. With the latest W10 Fall Creator Update, only PCI Express X16 Controller - 1901 to the dGPU needs to be disabled in order to resolve error 12. This is an improvement over older W10 versions which required both PCIe bridges to dGPU and left Thunderbolt 3 ports disabled for AMD eGPU to work.
I ran AIDA64 with nothing attached to the expansion ports to get a baseline of bandwidth (left). I ran ATTO benchmark on an USB SSD attached to the USB port of the Razer Core V2 then ran AIDA64 at the same time (right). You can see they are sharing bandwidth through a single Thunderbolt 3 cable.
Here are some benchmark numbers for this eGPU setup. They were run in internal display (left) and external monitor (right).
I use this eGPU with my 15" MBP in internal display mode only. It's been very reliable booting up with rEFInd. The dGPU remains disabled and iGPU attached to the internal display even when I boot it up without eGPU connected. If you experiencing flickering issues when there's no power source connected to the laptop, change the Intel Graphics setting to not use power saving.
Radeon drivers in Windows are getting better for external graphics use. With this pairing, I see very minimal loss using the internal display vs. external monitor. If you plan on using an AMD eGPU with an external monitor on a TB3 15" MBP, instead of disabling the PCIe 1901 - Radeon dGPU, try disabling PCIe Controller 1905 - left TB3 ports instead. This means you'll only able to use the right TB3 ports for eGPU and other peripherals. What you'll gain is ease of switching between macOS and Windows.
Unfortunately I didn't have time to run benchmarks in macOS prior to sending the media sample back to Razer.
I have a similar set up to you:
mid 2017 15" MacBook Pro - i7-7820HQ/Radeon Pro 560 dGPU/HD Graphics 630 iGPU/16GB RAM/512GB SSD
- that was previously hooked up to an Akitio Node Pro + nVidia 1080 GTX Ti egpu, which I managed to get working in Windows 10 (1709) via Bootcamp using your set up guide (thank you for such well detailed steps that even a total noob such as I could follow!)
However, since recently changing my egpu enclosure to Razer's Core V2 (circumstances forced my hand, otherwise if it ain't broke... right?), the set up no longer works 🙁
In windows, the eGPU fans are still humming along as if the card is being used; it is still detected, appearing in device manager, but now the applications (e.g. games) are using the internal display attached iGPU instead of switching to use the eGPU - evidenced by how hard the macbook's own fans start blowing upon loading up Crysis (not to mention its absence from the nVidia GPU Activity app in the system tray).
I haven't changed anything else besides the enclosures; I've looked for extra drivers or workarounds that might be needed to force usage of the eGPU to no avail...
@sooj When changing the enclosure, I'd recommend using DDU to completely uninstall the graphics drivers. Once your MacBook Pro can communicate with the new enclosure, manually installing the drivers will make it work.
Really nice build. I am thinking of getting the same setup. I use the late 2016 MBP 15" with i7-6820HQ (2.7GHz).
I am comparing benchmarking between different eGPUs at the moment and realized there is something odd about the numbers of your benchmark on Unigine. Im not a computer sage, but how did you get such good benchmark result on this version of Vega 56? It "almost outperform" (maybe not really) the GTX 1080Ti in the score with same laptop you use with most of your builds.
I am reffering to this build
This Vega 56 Mantiz benchmark doesn't score near as good as the Razer V2 build you did
Is it due encasing? Rendering it in Direct3D11 or perhaps difference between Win 8, 10 and Dawin?
@test1 I flashed RX Vega 64 vBIOS onto this RX Vega 56 so it's running a little better than a stock RX Vega 56. You were looking at RX Vega 56 Windows 10 performance in Unigine vs GTX 1080 Ti macOS performance. Compared both in Windows 10, the RX Vega 56 has no chance vs GTX 1080 Ti. Last but not least, the Mantiz Venus build was when the RX Vega 56 first came out. I ran it stock and Radeon drivers were not as optimized.
Oh okay, then I understand. It's only the samsung memory GPU that has the flashing capability, right? Do you think the Vega 64 is worth the money over a flashed 56?
EDIT: Oh and by the way, have you tried the RX 590?