Left 4 Dead 2 Vulkan Performance With Radeon Graphics On Linux

[eydee]

Game was released more than a decade ago, in the dual-core era. The engine itself was designed in the single-core era.

Runs tests on modern ultra high-end 12-core (24T) CPU...

Give it an Athlon X2 or Celeron, that's where the Vulkan renderer is supposed to shine.

[Linuxxx]

Any idea why my vulkan performance is less than 1/3 that of OpenGL ?
https://openbenchmarking.org/result/...IB-L4D2VULKA83

Seriously, that has to be one of the most broken Linux setups I have ever seen!

I mean, just take a look at this:

- OS: Slackware 14.2
- Desktop: KDE 4.14.38
- Graphics: AMD Radeon Frontier Edition 16GB (1350/988MHz)
- Kernel: amdgpu.ppfeaturemask=0xFFFD7FFF amdgpu.vm_fragment_size=9 - Transparent Huge Pages: always
- Processor: Scaling Governor: acpi-cpufreq ondemand (Boost: Disabled)

Just ... WOW!!

[Azpegath]

Seriously, that has to be one of the most broken Linux setups I have ever seen!

I mean, just take a look at this:

- OS: Slackware 14.2
- Desktop: KDE 4.14.38
- Graphics: AMD Radeon Frontier Edition 16GB (1350/988MHz)
- Kernel: amdgpu.ppfeaturemask=0xFFFD7FFF amdgpu.vm_fragment_size=9 - Transparent Huge Pages: always
- Processor: Scaling Governor: acpi-cpufreq ondemand (Boost: Disabled)

Just ... WOW!!

I don't follow, how is it broken based on those stats?

[Azpegath]

No measurable diff with this system.

Could we get the tests with regards to older hardware? It feels a bit uninteresting to go from 350 to 370 fps, but 60 to 70 fps on older hardware is a nice gain.
Expecially if it boosts the lowest framerates and decreases frame-drops.

[willmore]

Could we get the tests with regards to older hardware? It feels a bit uninteresting to go from 350 to 370 fps, but 60 to 70 fps on older hardware is a nice gain.
Expecially if it boosts the lowest framerates and decreases frame-drops.

I second this. When I saw the 'with modern hardware' line, I knew the results wouldn't tell me anything useful. I appreciate the testing, but older GPUs and CPUs would make much more sense for this test. Because no one with modern hardware cares about this performance change. Unless they have an 8K/120 display or something.....

[Flaburgan]

Yep, RX580, and even on Intel Iris (I have friends playing L4D2 without GPU, on intel i5 or Ryzen 5)

[Soul_keeper]

Seriously, that has to be one of the most broken Linux setups I have ever seen!

I mean, just take a look at this:

- OS: Slackware 14.2
- Desktop: KDE 4.14.38
- Graphics: AMD Radeon Frontier Edition 16GB (1350/988MHz)
- Kernel: amdgpu.ppfeaturemask=0xFFFD7FFF amdgpu.vm_fragment_size=9 - Transparent Huge Pages: always
- Processor: Scaling Governor: acpi-cpufreq ondemand (Boost: Disabled)

Just ... WOW!!

I'm actually using xfce and slackware64 current.
Everything is set the way it is on purpose.

[blackshard]

Once you have GPUs with matching results, the CPU is the bottleneck.
The lesser CPU overhead is why the Vulkan results yield higher performance. But the CPU still isn't able to keep up.

CPU is clearly the bottleneck at 1920x1080, but maybe the engine is limited somehow: it does not make sense that you have a CPU-limited condition and minimal increase with Vulkan which is supposed to reduce most of the CPU overhead.

[indepe]

So for 6800 XT with Vulkan, FPS increases with resolution?

Must that be a coincidence within the error margin, or is it possibly a consistent phenomenon?

[schmidtbag]

CPU is clearly the bottleneck at 1920x1080, but maybe the engine is limited somehow: it does not make sense that you have a CPU-limited condition and minimal increase with Vulkan which is supposed to reduce most of the CPU overhead.

The CPU isn't that high when it comes to games. This isn't that graphically demanding of a game either, so to me the results make sense.