Radeon vs. NVIDIA Performance For HITMAN On Linux With 17 GPUs

[puleglot]

Some cards simply doesn't have enough VRAM. I think TONGA with 4Gb should perform better in high quality/high resolution tests.

[philips]

Hmmm real strange, it should be called with bash and I am doing all of my hitman testing on Ubuntu 16.10 too. Did you make any major customozations to your system?

I have the same issue here on SteamOS, I had to manually create those steam-env-vars.sh...

[Ray Ingles]

You could try ln -s bash sh? If it's calling sh but needs bash then a symlink should fix it?

Actually, bash detects that, and runs in sh-compatible mode if invoked that way.

But I finally got things working. I had to add the following to the beginning of the "pre.sh" script:

 if [ ! "$BASH_VERSION" ] ; then     exec /bin/bash "$0" "$@" fi

This forces the script to re-execute itself via bash, even if it's invoked via sh. At that point, the benchmark would run.
Still, I got some unexpected behavior. Running it at 2560x1440@Medium, there was a lot of 'stuttering' in the performance. It looked like dropped frames and single-digit dips. Running at 1920x1080@Medium, it was far smoother - still some jerks here and there, but generally smooth. But these dips didn't seem to be reflected in the displayed FPS, or the average at the end. The 1440p run was about 45fps, but looked much worse. The 1080p run was 58fps, but seemed much smoother. The in-game performance at 1440p doesn't seem to suffer the same hitches. Anyone else see behavior like that?