OpenGL vs. Vulkan For Older NVIDIA Kepler GPUs (Early 2017)

[phoronix]

Phoronix: OpenGL vs. Vulkan For Older NVIDIA Kepler GPUs (Early 2017)

Most often when running any regular NVIDIA Linux benchmarks with Vulkan/OpenGL, it's usually just with the newest Pascal GPUs and then the older Maxwell GPUs for reference. But if you are curious about the OpenGL vs. Vulkan performance for GTX 600/700 "Kepler" graphics processors, I have some fresh results today...

http://www.phoronix.com/scan.php?page=news_item&px=2017-Jan-Kepler-OGL-VLK

[swoorup]

I think we are seeing these results, because nvidia basically optimized their driver for each major AAA games. Talos principle would have been interesting benchmark.

[Kano]

Talos Principle just needs more VRAM with Vulkan. Lower res to 1080p and the benchmark would run. But there are certainly still issues with Vulkan and normal gameplay. With 1 and 2 GB Kepler cards the game crashes in the elevator. A bit useless if you have to switch between OpenGL and Vulkan.

Dota 2 has extreme rendering issues in my test box with lower res than my KDE desktop as it does not really use fullscreen but shows it in the upper left corner (1080p res here), the rest are artefacts.

[bug77]

With every additional piece of evidence it becomes increasingly more obvious the touted lower overhead of Vulkan is just a myth. I mean, I'm sure the overhead is smaller, but that doesn't seem to translate into gains in CPU-bound scenarios
Oh well, interesting times ahead...

[indepe]

With every additional piece of evidence it becomes increasingly more obvious the touted lower overhead of Vulkan is just a myth. I mean, I'm sure the overhead is smaller, but that doesn't seem to translate into gains in CPU-bound scenarios
Oh well, interesting times ahead...

I doubt there is evidence of anything other than the lack of a performant implementation in the game, or the driver, or both. Especially unlikely that there has been any attempt to use multiple cores on the CPU side. This seems to be an engineering-bound scenario.

[johnc]

With every additional piece of evidence it becomes increasingly more obvious the touted lower overhead of Vulkan is just a myth. I mean, I'm sure the overhead is smaller, but that doesn't seem to translate into gains in CPU-bound scenarios
Oh well, interesting times ahead...

I told everyone from the get-go that this whole low-level API craze was nothing more than a big marketing snow-job.

Within a few years developers will be screaming for abstract APIs again.

[indepe]

I told everyone from the get-go that this whole low-level API craze was nothing more than a big marketing snow-job.

Within a few years developers will be screaming for abstract APIs again.

That is expected in the first place. The idea is that abstracted API's will be implemented on top of Vulkan. Vulkan as a low-level API competes mostly with DirectX.

[bug77]

I told everyone from the get-go that this whole low-level API craze was nothing more than a big marketing snow-job.

Within a few years developers will be screaming for abstract APIs again.

That's what I was thinking, too, but I kept it to myself (mostly) until we saw at least some implementations. Developers with big budgets may put Vulkan ot good use, but otherwise it's like everyone went back to write programs in ASM because it gives you access to the hardware and has the lowest overhead. Just like you can use asm routines in C, I expect many things will be written in OpenGL and call Vulkan routines here and there. We'll see...

[Up123]

I prefer stable over max FPS

[indepe]

That's what I was thinking, too, but I kept it to myself (mostly) until we saw at least some implementations. Developers with big budgets may put Vulkan ot good use, but otherwise it's like everyone went back to write programs in ASM because it gives you access to the hardware and has the lowest overhead. Just like you can use asm routines in C, I expect many things will be written in OpenGL and call Vulkan routines here and there. We'll see...

OpenGL doesn't scale to multiple cores, so it's on the way out.