Vulkan vs. OpenGL Linux Game CPU Core Scaling

[phoronix]

Phoronix: Vulkan vs. OpenGL Linux Game CPU Core Scaling

After carrying out the P-State/CPUFreq governor comparison with a focus on OpenGL and Vulkan Linux games, next I ran some fresh numbers seeing how well modern OpenGL/Vulkan Linux games are scaling across multiple CPU cores.

http://www.phoronix.com/vr.php?view=24839

[boxie]

looks like Vulkan is doing what it is supposed to be

[LeJimster]

Very interesting. You would expect Vulkan games to scale with more cores. I wonder if this is an limitation of radv right now.

[andrei_me]

Nice how Vulkan achieves higher FPS with fewer threads, probably the reason why everyone asks for vulkan test on weaker CPUs, maybe in a Carrizo or Pentium system

[Xicronic]

What the hell is with Michael using AMD cards exclusively for benchmarks recently. I know not everyone is a fan of Nvidia blobs, but it is no secret they perform & scale better. This is especially nasty for an OpenGL vs Vulkan comparison where RADV is missing Vulkan extensions or has slow codepaths that make Vulkan look worse than it is (and sometimes vice versa, as DOW3 just recently became playable on Mesa GL at all). At least also show an equivalent Nvidia card.

[bosjc]

What the hell is with Michael using AMD cards exclusively for benchmarks recently. I know not everyone is a fan of Nvidia blobs, but it is no secret they perform & scale better. This is especially nasty for an OpenGL vs Vulkan comparison where RADV is missing Vulkan extensions or has slow codepaths that make Vulkan look worse than it is (and sometimes vice versa, as DOW3 just recently became playable on Mesa GL at all). At least also show an equivalent Nvidia card.

Yeah, seriously for Vulkan testing using RADV/AMD makes little sense. Its nice to see how the driver is progressing, but NVIDIA at this point has pretty much the gold standard Vulkan driver that would much better show off what Vulkan can do.

[dkasak]

People with nVidia cards who don't mind binary blobs. Wouldn't they be using Windows? Disclaimer ... half-arsed troll ...

[Shevchen]

What the hell is with Michael using AMD cards exclusively for benchmarks recently. I know not everyone is a fan of Nvidia blobs, but it is no secret they perform & scale better. This is especially nasty for an OpenGL vs Vulkan comparison where RADV is missing Vulkan extensions or has slow codepaths that make Vulkan look worse than it is (and sometimes vice versa, as DOW3 just recently became playable on Mesa GL at all). At least also show an equivalent Nvidia card.

Testing AMD GPUs and finding a bug/flaw: Developers can fix it.
Testing Nvidia GPUs and finding a flaw/bug: Developers have to accept it.


Thats why testing Open Source Software is good: You see a bug, open up your preferred editor, get into the code and can fix it. On closed source: You see a bug, light up the candle and pray that Nvidia will fix it.

As I'm an atheist, I prefer option 1.

[Pentarctagon]

On the other hand, benchmarks showing most games still performing better on OpenGL than Vulkan probably isn't going to help Vulkan's adoption any.

[bug77]

What the hell is with Michael using AMD cards exclusively for benchmarks recently. I know not everyone is a fan of Nvidia blobs, but it is no secret they perform & scale better. This is especially nasty for an OpenGL vs Vulkan comparison where RADV is missing Vulkan extensions or has slow codepaths that make Vulkan look worse than it is (and sometimes vice versa, as DOW3 just recently became playable on Mesa GL at all). At least also show an equivalent Nvidia card.

He always tests "out of the box" if possible. It requires less time.
Whether that makes sense for each and every benchmark is another story. But at least he's predictable.

On the other hand, I want to say "ha!" to everybody that was posting about how OpenGL is going to die because it doesn't multithread well. These forums were littered with those posts just because Vulkan's launch.