2013 27" iMac (GTX780M) [4th, 4C,H] + RX 580 @ 10Gbps-TB1>TB3 (ASUS XG Station Pro) + macOS 10.15.3 [truedichotomy]  

This is my first eGPU setup and post here!!!  Thank you all for the awesome resource, especially goalque!

System specs

• iMac 27-Inch "Core i7" 3.5 (Late 2013) (Nvidia GTX780M + Intel Iris Pro 1536 MB)
• OWC Helios TB3 with NVMe PCIe card with Samsung SSD PRO 512GB connected via TB1 (internal Fusion Drive has failed)
• macOS Catalina 10.15.3

eGPU hardware

• ASUS XG Station Pro
• XFX Radeon RX 580 (8GB)
• Daisy chained to OWC Helios TB3
• Apple TB3 to TB2 adapter
• Apple 1m TB2 cable
• TB3 cable 1.5m
• USB drive for automate-eGPU EFI

Installation steps (what did you do to get it all going?)

• Installed the XFX card into the XG Station Pro eGPU enclosure
• Formated an USB flash drive via command line (i.e. "diskutil eraseDisk FAT32 EFI MBR /dev/disk4"). Named it anything but EFI!
• Downloaded automate-eGPU EFI from https://egpu.io/wp-content/uploads/2020/02/EFI.zip
• copied unzipped EFI folder to USB flash drive according to https://egpu.io/forums/mac-setup/automate-egpu-efi-egpu-boot-manager-for-macos-and-windows/
• Updated config.plist and automate-eGPU.efi files in the EFI folder according to https://egpu.io/forums/mac-setup/automate-egpu-efi-egpu-boot-manager-for-macos-and-windows/paged/24/#post-67493 for Catalina.
• Connected all the hardware together, powered up all TB devices, and powered on iMac while holding 'option' key.
• Select EFI_Boot, Hit 'q' in EFI menu and select Boot Mac OS.

Benchmarks

Geekbench 5 Compute Benchmark for eGPU:

• OpenCL: 36270
• Metal: 43110
•  

Comments

I just got the setup working this afternoon. I also tested the same eGPU setup connected to my 2018 MBP 15'' (6C, i7, Radeon Pro 560X). Had to disable secure boot options to make that work, but otherwise worked with no apparently problems. the eGPU is roughly 7x faster than the iMac's NVIDIA GTX780M and 3x faster than the MBP's dGPU. Plan to try an Ubuntu Linux boot on the iMac using another USB drive with the eGPU to see if I can get it to work. Might invest in a mightier NVIDIA card in the future for some ML work if I can get the Ubuntu Linux boot to work.