Folding Forum

Hi all,

i've been running World Community Grid for many years and have currently two Ryzen 1700's running 24/7 on it. I now wanted to add a GPU to these and chose Folding@Home as the best project to support science.
I started with a GTX 750 Ti (I know it might not be the most efficient GPU anymore, but I first want to gather some experience before I possibly upgrade to a modern GPU).

So far so good, I first tried folding on Windows 10, worked without issues. After completing 1 WU, I wanted to switch back to Lubuntu my crunchers normally run on. I installed the NVidia drivers as well as the FAH client. The GPU seems to be recognized, however the GPU slot always says "ready" and never starts to fold.
I searched the web, played around a bit but nothing works. Here is a screenshot:



Does anyone have an idea what the cause may be and what I could try next to get it running?

Welcome to the folding support forum.

Information from your log file, especially the first 100 or so lines, would be useful in figuring out your system's issue with folding on the 750ti. Directions on how to locate and post your log are in the Welcome topic - viewtopic.php?f=106&t=26036.

Without seeing that, the most common problem is lack of OpenCL support for the GPU. GPU folding on Linux requires use of the nVidia proprietary drivers and installation of OpenCL support along with them. Depending on where you download the drivers from, they may be complete or require separate downloads for the components.

Thanks Joe!
After experimenting some more, I figured setting opencl-index to 0 instead of -1 fixes the problem. Now everything is running as it should.

Sheridon wrote:I started with a GTX 750 Ti (I know it might not be the most efficient GPU anymore, but I first want to gather some experience before I possibly upgrade to a modern GPU).

Good plan.

I'm still runninig a 750Ti that I've had for a number of years in a Linux system along with one "modern GPU" As long as you've got room for the PCIe connections (plus enough cooling) you can do both. It is (was?) the most efficient of the GTX700 series -- a Kepler on a 28 nm process.