In order to find out exactly how much is the connection bandwidth interfering with the performance, the following setup was used:
-Asus XG Station Pro with Geforce 1080 Ti
-15" MPB 2018 with 80cm Apple TB3 cable
-13" MPB 2015 with 100cm TB2 cable and Apple TB2->TB3 adapter
-Superposition benchmark, because it is very demanding for GPU compared to CPU.
The results yield following conclusions:
1) The more demanding the test, the less is the difference between TB3 and TB2 (external display). For smaller FPS the difference in bandwidth does not play a role at all. For 1440p Extreme there is hardly a difference between TB3 and TB2.
2) There is a upper limit in FPS when using TB2/3. This effect is clearly notable for TB2 as the score is not changing a lot between the 720p and 1080p tests. For TB3, which has a higher bandwidth the number are much more different.
3) It's best for an eGPU on external displays to use the highest settings possible. Most details and a highest resolution that are over 60 FPS.
In this test this would be 1440p High for eGPU with external Display.
4) For the loopback scores another factor is coming in: The resolution. As the data has be sent back again through Thunderbolt, the resolution plays an important role (other than for external displays). That's why there is a notable difference in the 1440p test even though the FPS are quite low, but hardly no difference in the 720p Test.
5)For an eGPU on on internal display a medium resolution (1080p, FHD) with high details is recommended.
In this test this would be 1080p High for eGPU with internal Display (loopback).
Superposition Preset/Connection Type | TB3 external Display | TB2 external Display | TB2 internal display (loopback) |
720p Low |
|
|
|
1080p Medium |
|
|
|
1080p High |
|
|
|
1440p High |
|
|
|
1440p Extreme |
|
|
|
2018 15" MBP & 2015 13" MBP connected to RTX2080Ti GTX1080Ti GTX1080 Vega56 RX580 R9-290 GTX680
2018 15" MacBook Pro (RP560X) [8th,6C,H] + RX 5700 XT @ 32Gbps-TB3 (ASUS XG Station Pro) + Win10 & macOS 10.15.4 // Navi vs Radeon VII vs GTX 1080 Ti [build link]
Finally I got my internal acceleration for the TB3 running, so I got the benchmarks for the TB3-internal-display-loopback:
Those relation are safe for every preset:
1) TB3 external > TB2 external > TB2 internal
2) TB3 external > TB3 internal > TB2 internal
But what about TB3 internal and TB2 external? Which one performs better? Well, that depends a lot of the resolution, as suggested before. For 720p and 1080p TB3-internal has the upper hand, while for 1440p TB2 external performs better:
Superposition Preset/Connection Type | TB3 internal (loopback) | TB2 external |
720p Low |
|
|
1080p Medium |
|
|
1080p High |
|
|
1440p High |
|
|
1440p Extreme |
|
|
2018 15" MBP & 2015 13" MBP connected to RTX2080Ti GTX1080Ti GTX1080 Vega56 RX580 R9-290 GTX680
2018 15" MacBook Pro (RP560X) [8th,6C,H] + RX 5700 XT @ 32Gbps-TB3 (ASUS XG Station Pro) + Win10 & macOS 10.15.4 // Navi vs Radeon VII vs GTX 1080 Ti [build link]
Retrospectively I must say, this is a very fine analysis and answers a lot of questions about bandwidth usage.
2018 15" MBP & 2015 13" MBP connected to RTX2080Ti GTX1080Ti GTX1080 Vega56 RX580 R9-290 GTX680
2018 15" MacBook Pro (RP560X) [8th,6C,H] + RX 5700 XT @ 32Gbps-TB3 (ASUS XG Station Pro) + Win10 & macOS 10.15.4 // Navi vs Radeon VII vs GTX 1080 Ti [build link]
it's awesome!Thanks for your effort!
To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.
.Hey, do you think switching to an external display going to yield a significant performance gain? I have the same macbook 13" 2015 but powered by RX 570 egpu and I even struggle to reach 30 fps playing GTA V on 1366x768, which seems too underpowered for an RX 570
To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.
.