13-Way NVIDIA OpenGL vs. Vulkan Dota 2 Ubuntu Linux Benchmarks

[Julius]

Michael, could you add FTEQW (with Afterquake addon) to the vulkan benchmarks? It's a quake derived engine with a pretty nice vultan renderer. After quake includes advanced graphics and a common demo for benchmarking, see: https://qexpo2016.com/afterquake-updated/

[sabun]

Where am I logging Vulkan frames? or I am not sure what you're after. With the Dota 2 testing, it's just using the built-in exposed info: http://openbenchmarking.org/innhold/...54b026f6b29981

Oh! I had no idea within Dota 2 it could log frames itself. I just assumed it was a universal tool you were using to gather this info. Thanks for the info

[humbug]

For AMD i think the same, only Fury/X/Nano on 4K - others might be not clear sometimes so vague, leading to false conclusions of 4K

I think for AMD cards Vulkan will probably be universally faster even at 4K.

Even though it is now very solid and performant, the Vulkan renderer in Source2 is still considered beta and you don't get it by default. See below Dan Ginsburg's comments on how they plan to make it even faster

https://youtu.be/DWLkA6-wzj0?t=40m1s

[efikkan]

I think it's much simpler. The 1070 and 1080 are fast enough for Dota2 to become CPU bound again at 4k.

You totally missed the point; the claim from andre30correia was that the scaling should be fixed by the driver vendor. This game is written using a "Direct3D 11"-like abstraction over the "low level" API Vulkan, so it will most definitely run into bottlenecks in the engine long before the driver.

[Michael]

Michael, I'm curious, what tool are you using to log the Vulkan frames? Is it something you developed in-house or is there a public tool available for fps and frametime logging?

I can't seem to find anything via Google or the Vulkan webpage, which is why I ask.

Where am I logging Vulkan frames? or I am not sure what you're after. With the Dota 2 testing, it's just using the built-in exposed info: http://openbenchmarking.org/innhold/...54b026f6b29981

[Ehvis]

No, drivers don't work differently in different resolutions.
Clearly the games doesn't scale very well, and given it's Vulkan through an "Direct3D 11"-like wrapper I'm not surprised...

I think it's much simpler. The 1070 and 1080 are fast enough for Dota2 to become CPU bound again at 4k.

[efikkan]

probaly they need to work much more in vulkan drivers in 4k.

No, drivers don't work differently in different resolutions.
Clearly the games doesn't scale very well, and given it's Vulkan through an "Direct3D 11"-like wrapper I'm not surprised...

[sabun]

Michael, I'm curious, what tool are you using to log the Vulkan frames? Is it something you developed in-house or is there a public tool available for fps and frametime logging?

I can't seem to find anything via Google or the Vulkan webpage, which is why I ask.

[elvisn]

I would be happy only seeing all of those GPUs on the table in the first picture :'-)

But it got even better with all of the benchmarks! Thanks Michael for all of your hard work making these types of articles (the benchmark ones especially)!

[dungeon]

probaly they need to work much more in vulkan drivers in 4k.

Of course they will work but for future cards. Let alone GPU bound, but 4K also use more VRAM and VK version itself uses more VRAM, so it is harder for driver memory managment on low end, etc... also older cards lacks perf advantage features if let say game wanna use async compute.

You see there what is enough for 4K only GTX 1080 Only on that one there is sort of same VK advantage as on 1080p.