I found I got almost identical results with the eGPUs and the i5 2019 MBP 13 inch with maxed out RAM. Which is almost half the price of my MBP 15 inch...
There are a couple of things: for the eGPU to work, the app has to be tagged for "Prefer external GPU", the app has to use either Open CL/GL or better Metal, and the app has to be taking actually advantage of GPU. I get excellent results, but one has to fiddle a little bit with settings, counterintuitively switching off GPU support sometimes to make the app for instance switch to Metal. And you need to connect your monitors directly to the eGPU and pull your work to those monitors with most apps. You can monitor with iStats Menu what is being used, in combination with timing your workflow.
I had an eGPU running with my old 2014 MBP on 2x PCIe lanes, and it was difference of day and night. However, GPU is not used for many tasks one may think, for instance transcoding is typically done in CPU, whereas filters are applied in GPU. And then there are those apps that do not run with eGPU at all like X-Plane. And finally those apps without named GPU support like Omni Graffle that greatly benefit from eGPU, you will know when you receive large converted documents that end up in millions of vectors. But most productivity apps, including both hypervisors Fusion and Parallels work very well with eGPU.
Largest hassle with eGPU is that the system logs out and back in when you disconnect, which is a real bother if you want to take your computer from your desk to your couch and have lots of apps open that are attached to eGPU. Otherwise, it works very well with my 2019 MBP and 4x PCI 3.1 lanes are plenty fast.
A great side benefit that I noticed when using VMs is when the system sucks down 100+ Watts CPU, plus GPU processing done on eGPU. You get additional performance out of your system which otherwise would heat-throttle the CPU and redirect half of the power to dGPU.
