2012 15" HP Envy DV6 (GT630M) [3rd,4C,Q] + RX 580 @ 4Gbps-mPCIe2 (EXP GDC 8.4) + Win10 [Golden Ratio]  

Hello,

I would like to show you two fully functional setups with the same notebook, but different eGPU's. First a small hint, my written knowledge of english is not very good. I wrote this text with the help of an online translator! 🙂

System specs
2012 15" HP ENVY dv6-7240sg

• Windows 10, Version 1809, 64-Bit
• Intel Core i7-3630QM
• 8 GB DDR3 1600
• TOLUD of 2,5 GB (9FA00000)
• 500 GB Samsung SSD 580 EVO
• Intel HD Graphics 4000 (iGPU)
• NVIDIA GeForce GT 630M, 2 GB (dGPU) @x8 2.0
• 1366 x 768 screen

   

eGPU hardware

• EXP GDC 8.4d, mounted in a simple metal enclosure
• Dell DA-2 220 Watt PSU
• KFA² GeForce GTX 1060 OC, 6 GB  (eGPU1) @x1 2.0 mPCIe2
• ASUS AREZ Dual Radeon RX 580 OC, 8 GB (eGPU2) @x1 2.0 mPCIe2
• Dell UltraSharp U2311H 23", 1920 x 1080 screen

Hardware pictures

Device Manager for Setup with GTX 1060 (eGPU1) on the left and RX 580 (eGPU2) on the right	AIDA64 GPGPU Benchmark with RX 580 (eGPU2)
Setup with GTX 1060 (eGPU1)	Setup with RX 580 (eGPU2)

Installation steps

First, I installed the GTX 1060 (eGPU1):

1. Uninstalling dGPU driver (GeForce 391.35) with DDU in Safe Mode
2. Disabling dGPU in device manager after restart
3. Installing latest eGPU1 driver (GeForce 416.94) and restart again
4. Fix error 43 in eGPU1 with Nando's script "nvidia-error43-fixer"
5. Done! 🙂

For the RX 580 (eGPU2) i have repeated steps 1-3 as described above (by uninstalling eGPU1 instead of dGPU).
The installation of the RX 580 itself with the newest driver (AMD adrenalin 18.11.2) passed without recieving any error after restart. So i could proceed immediately with Battlefield and 3DMark as can be seen in the screenshots.

Benchmarks

So lets have a quick look on the benchmark results. Due to lack of time i used for testing the Battlefield 4 Test Range (DirectX 11) and 3DMark's Fire Strike (DirectX 11) and Time Spy (DirectX 12) on an external FHD display.

• Battlefield 4

Preset was 1920 x 1080 with details on Ultra.
As one can see, the framerate increases by the factor of 5 by moving from GT 630M (~12) to GTX 1060 and RX 580 (66~70). This is really impressive for an 6 year old laptop. While the FPS on both eGPU's were almost equal, the RX 580 without undervolting took almost twice (~120 W) as much Power as the GTX 1060 (~70 W). I couldn't get any power consumption data for the dedicated graphics.

Battlefield 4 Test Range with GT 630M, GTX 1060 and RX 580 (from upper left to lower right)

Another interesting fact about my RX 580 is that Battlefield 5 keeps crashing after joining an active session when DirectX 12 is activated. Unfortunately i didn't test it with the GTX 1060.
Update: DirectX 12 is disabled in Battlefield 5 on mobile devices, which i didn't know at the time of testing.

• 3DMark

Lets focus on 3DMark scores. On Fire Strike (DirectX 11) the scores starts with 842 for the GT 630M, and goes up to 8378 (factor 10) for the GTX 1060 and even more 9790 (factor 12) for the RX 580. On Time Spy (DirectX 12) the Radeon is only marginally better than its companion.

Fire Strike (DirectX 11) with GT 630M, GTX 1060 and RX 580

Time Spy (DirectX 12) with GTX 1060 and RX 580

Even if the AMD results in my case are slightly better in both 3DMark benchmarks, i couldn't notice any big differences ingame between the two cards, which is probably due to the bottleneck of the data bus.

Cheers!