Better Wine Benchmarking This Summer For Windows Programs On Linux

[Michael]

Spectacular. I love to see PTS getting better year-over-year.

Thanks, that's the goal... Now if only the number of commercial customers would scale with the feature set :/

[Michael]

Michael

This titles have benchmark integrated: batman arkham city - dirt 3 - grand theft auto IV - world in conflict - lost planet extreme condition - devil may cry 4 - resident evil 5/6 - ultra street fighter IV - crysis and maybe others

If you would like to help, feel free to post the command line options to run for each of those games for triggering the benchmark modes, or at least documentation to the CLI options and their outputs, so I can check them out before buying the games while saving time.

[aaahaaap]

But this repo does just new 'releases'? So every 2 weeks? I was hoping for a PPA that would be automatically pushing out new builds nightly or at least like weekly, certainly would help in catching regressions quicker.

Yeah, it's just for new releases.
come to think of it, you can also get that from winehq apart from the staging builds.

[darkbasic]

Why do people think that a wrapper can outperform a native driver? It can't. In practice you can expect ~50-70% of Windows performance, in some games that run exceptionally well it can get somewhat close to Windows performance but those wrappers will never be able to match it.

Gallium 9 actually got faster than native in a couple of titles.

[VikingGe]

Gallium 9 actually got faster than native in a couple of titles.

Gallium Nine is as native as it gets. Or did someone secretly rewrite it to use Vulkan instead?

[darkbasic]

Gallium Nine is as native as it gets. Or did someone secretly rewrite it to use Vulkan instead?

Gallium Nine is exactly as VK9: while VK9 runs DirectX over Vulkan Gallium 9 runs DirectX over Gallium. That's exactly the same.

[Weasel]

Why do people think that a wrapper can outperform a native driver? It can't. In practice you can expect ~50-70% of Windows performance, in some games that run exceptionally well it can get somewhat close to Windows performance but those wrappers will never be able to match it.

Isn't 50% a bit harsh? I mean that's what you'd expect from DX->OGL, no?

[VikingGe]

Isn't 50% a bit harsh? I mean that's what you'd expect from DX->OGL, no?

50% is definitely on the lower end, but not unrealistic (take Frostpunk as an example). DXVK isn't necessarily much faster than wine-pba (especially on Nvidia where pba seems to work quite well in general), just happens to work better for a bunch of newer games.

That's exactly the same.

Except that Vulkan is not exactly the same as Gallium. One is an explicit API designed to be used by applications directly, the other is a higher-level, somewhat D3D11-like API designed solely to implement hardware drivers for legacy APIs. Some of the core concepts are fundamentally different:
- Almost entirely static pipeline state (instead of somewhat dynamic state objects)
- Descriptor sets (instead of fixed resource binding slots)
- Render passes (Vulkan-specific)
- Very explicit memory barriers

How much work a Gallium driver needs to do in order to translate that into something the hardware understands obviously depends on the hardware, but with Vulkan you're pretty much guaranteed to do a fair amount of work that's just completely unnecessary in a Gallium-based stack.

[duby229]

Gallium Nine is as native as it gets. Or did someone secretly rewrite it to use Vulkan instead?

You do realize Vulkan is lower level than Gallium right? So VK9 should actually be cl;oser to hardware than Gallium Nine was. And Gallium Nine could actually outperform Windows in many titles!

[creative]

Michael

I am not sure how far this is in line with the subject, but what about a 64bit virtual drive on PlayOnLinux with the Steam client featuring a Doom 2016 installation? I have it running on my Manjaro 17 XFCE installation drive and it is running pretty good with the Vulkan drivers with Wine 64bit 3.9. While I don't think Doom 2016 has a built in benchmark. One could test the approximate average FPS while standing still in a given area with Dooms built in performance metrics. I am running it on a GTX 1070 and core i7 7700 65watt with 16GB of DDR4 and all is maxed at 1080p. About 175-200fps seems to be what it pulls on average given what area I am in.