Blender Asked by Legoman on December 21, 2021
I’m looking to improve my hardware, and I have decided to invest in a GPU (like the TITAN RTX, GeForce RTX 2070 Super, or something similar).
I’d like to use this as an external GPU. My question is, does using an external GPU vs. an internal one have any performance issues in Blender?
I am using Linux (Manjaro KDE).
I am running an eGPU on Linux and I can confirm the speed is on par with other systems. My setup is a Dell XPS 13 (9370) with a Quad Core i7 (8550U) and a Razer Core X eGPU enclosure with a Radeon RX 580. To give you some numbers, the BMW scene renders in a little over 262 seconds. Here is the result of my render:
https://opendata.blender.org/benchmarks/3b842bbd-ed59-4352-a1f1-f6ef10a25864/
That's below the median time for that scene (266 seconds) on systems with the same configuration (BMW + Linux + OpenCL + RX 580 + Blender 2.90):
From the OpenData results we can't distinguish eGPU's from dGPU's, but I think desktop setups are far more common than eGPU setups. On my machine, rendering on the GPU is about 2.5 x faster, which is what I measure most of the time. As a reference, the same scene renders in 663 seconds on the CPU. Here is my result:
https://opendata.blender.org/benchmarks/fffb4aed-f259-40d8-b307-827f1580c919/
I also tested with combined GPU + CPU render but the result is more or less the same as the GPU alone, often even slower. That's because the CPU is more efficient rendering small tiles (64 x 64 px) while the GPU prefers larger tiles. So if you increase the tile size, the CPU will negatively influence the result. While with smaller tile sizes, the GPU will not work at its full capacity.
On the elementary OS StackExchange, I've outlined my setup:
I've never used Manjaro, so I'm not sure how it will translate.
Answered by wout on December 21, 2021
Get help from others!
Recent Questions
Recent Answers
© 2024 TransWikia.com. All rights reserved. Sites we Love: PCI Database, UKBizDB, Menu Kuliner, Sharing RPP