Intel Celeron/Pentium/Core i3/i5/i7 - NVIDIA vs. AMD Linux Gaming Performance

[oooverclocker]

It looks like most games are CPU-bound.

It looks like all tested games are CPU bound The only exception is the GTX 1050 with Unigine Heaven.

on Windows CPU isn't usually the bottleneck.

Tells us two things: Either OpenGL or the current implementation sucks

[Passso]

@Michael: your new graphs are beautiful and readable, congrats!

[Michael]

@Michael: your new graphs are beautiful and readable, congrats!

Thanks, but comparing to what? The only change recently made in PTS7 Git was bolding the title and shortening up the identifier height on multi-way comparisons, or is that what you are referring to?

[coder111]

I was just thinking the same thing.

And it shouldn't be the case- that doesn't seem to be the case on Windows. GPU should have more of an effect on FPS and CPU less. Which leads me to believe there are inefficiencies either in OpenGL drivers, or the way games use OpenGL(engines), or X-windows or somewhere.

And there's probably low hanging fruit that could be optimized away quite easily. Trick would be to find them...

It looks like most games are CPU-bound.

[linuxjacques]

Thanks for this benchmark. I was not expecting those results.

[tomtomme]

If i may whish: all benchmarks again at 4k. Would be interesting what games get gpu bound at 4k.

[wdb974]

Thanks, but comparing to what? The only change recently made in PTS7 Git was bolding the title and shortening up the identifier height on multi-way comparisons, or is that what you are referring to?

I guess it's the colors. That was my reaction as well. Thanks!

[indepe]

I was just thinking the same thing.

And it shouldn't be the case- that doesn't seem to be the case on Windows. GPU should have more of an effect on FPS and CPU less. Which leads me to believe there are inefficiencies either in OpenGL drivers, or the way games use OpenGL(engines), or X-windows or somewhere.

And there's probably low hanging fruit that could be optimized away quite easily. Trick would be to find them...

Do you have a pointer to related Windows tests?

Well done test with surprising results. I'd be interested to find out the explanation. Perhaps at least on Linux, even if it is due to some fix-able problem, the theory that a game/computer combination is CPU-bound or GPU-bound, needs to be re-considered.

If it is a fixable problem, my first question would be if there generally is triple-buffering in the games, and if that works correctly with the graphics stack.

[dungeon]

Do you have a pointer to related Windows tests?

For example this

https://arstechnica.com/gaming/2015/...teamos-gaming/

On mediocre CPU and GPU everything is clear, Linux's best case driver scenario is slower than WIndows in every single tests.

Now if we don't use best drivers, situation is even worse But we can make it up to look better of course, if we use not mediocre hardware to avoid an issue for some time and again and again...

[zenbi]

Michael, are you sure that the horizontal/vertical legends aren't simply reversed? The graphs would make much more sense if the different colors where the different video cards (instead of the CPU.)