2015 15" MacBook Pro (R9 M370X) [4th,4C,H] + RX 580 @ 16Gbps-TB2>TB3 (AKiTiO Node) + macOS 10.13.6 [xpcyl8]
Functionally worked perfectly, including eject eGPU. Performance expected more though.
2015 MacBookPro 15" 2.5ghz i7
AMD Radeon R9 M370X
MacOS 10.13.6 High Sierra
Apple Thunderbolt2/3 adaptor
Zotac GTX 1070 Mini
I previously had this setup with Zotac GTX1070, but I was eager to find out how well RX580 compared because it's supported natively by Apple, so I sourced a Sapphire Nitro+ RX580 with 8GB GDDR5, thinking this would give me 80% or higher performance of what I got from 1070, as everyone on the forum seem to say go with AMD on the Mac's.
Functionally it worked perfectly, no drivers no nothing, it just worked.
With everything powered on and connected, HighSierra recognized eGPU and turned on external display as if an external display was plugged in. Everything worked as usual. UI and responses were smooth, so I thought this might be the new direction going forward. Until I ran benchmark...
Again, I used Heaven and not Valley because that's what I used before on my previous system and wanted to compare apple to apple.
The same Heaven setting resulted a paltry 23fps ( my reference of same system running GTX1070 was 40fps)
What? nearly half of what I got from GTX1070?! I thought AMD suppose to perform more efficiently on the Macs?! I re-ran the benchmark 2 more times, with a system reboot in between just to make sure i had a clean system. No change on 23fps.
Comments (eg: how has the eGPU improved your workflow or gaming)
This is my primary X-plane 11 system. ( See my other build guides ) I have to compare XP11 performance differences. Again, at same setting ( aircraft, scenery, weather, settings etc. ps. I can do this because my XP11 is in external SSD so I can be sure no setting is different by launching from the same installed code base)
The result? 13-14fps vs 27-30 fps.
So I am a bit surprised of low performance from RX580, perhaps I am over optimistic about that card to start with? Anyways, with this I can say that one should probably go with vega chips if target app is XP11.
Conclusion: it worked flawlessly and fluidly as expected with MacOS HighSierra, I could even eject eGPU using the menu icon. It's probably a success for anything else. Activity Monitor shows the eGPU correctly with load. It just didn't have the performance I hoped for.
It sounds like you don't quite fully understand how graphics cards perform. 🙂
While a 580 will be easier to get running with native support in macOS the GTX 1070 is a more powerful card overall. A Vega 56 would perform the same as a 1070 where as your 580 just about beats a 1060 which is a fair bit less powerful than the 1070/Vega.
Here are some random benchmarks that show the difference...