Migrating from cMP to Mac mini 2018

[Synchro3]

I can’t remember the results but my 3DMark score in Windows using a GTX 1080 in an eGPU was better than on the cMP. Not hugely so but better by a decent margin. More than I thought it would be tbh.

Here the opposite. I benchmarked Starcraft in June in 4K resolution on my Mac Mini with eGPU RX 580: https://forums.macrumors.com/threads/mac-pro-2019-and-gaming.2185465/page-3#post-27468661
- Starcraft II (Metal) = 47 fps @4K

Now with my new EFI RX 580 in the Mac Pro:
- Starcraft II (Metal) = 67 fps @4K

Conclusion:
- PCI-E 2.0 16x is faster than Thunderbolt 3
- an RX 580 with Metal support beats a GTX Titan X (both PCI-E 16x)

 

[orph]

Thunderbolt 3 is PCIE 3 x4 so = PCIE 2 x8 + some overhead

i have seen it mentioned that a GTX 2080TI is only just a tad limited by PCIE 3 x8 (ie vary small gains past PCIE 3 x8 in most tasks)

 

[Squuiid]

Here the opposite. I benchmarked Starcraft in June in 4K resolution on my Mac Mini with eGPU RX 580: https://forums.macrumors.com/threads/mac-pro-2019-and-gaming.2185465/page-3#post-27468661
- Starcraft II (Metal) = 47 fps @4K

Now with my new EFI RX 580 in the Mac Pro:
- Starcraft II (Metal) = 67 fps @4K

Conclusion:
- PCI-E 2.0 16x is faster than Thunderbolt 3
- an RX 580 with Metal support beats a GTX Titan X

Yeah, I think in my case the mini's CPU contributed to the gain. See sig for details.

 

[Synchro3]

Yeah, I think in my case the mini's CPU contributed to the gain. See sig for details.

My Mini has also an i7. I think the difference is Nvidia uses pcie data compression over Thunderbolt, AMD not (in macOS).