Quoting: mylkai tried the demo a few month agon on windows 7 and with proton + AOC... no flickering

I know that it used to work, but I explicitly said currently because it currently doesn't. Regressions do happen.

FWIW the issue affects the Windows version with both vkd3d and dxvk as well, so this isn't a problem with the port, just an ACO bug.

View this comment - View article - View full comments

Some quick Linux vs Windows benchmarks, settings shown in the screenshots:

Spoiler, click me

Linux:

To protect your privacy, external media requires approval to load.

Source: i.imgur.com

View cookie preferences.
Accept & Show Accept All & Don't show this again   Direct Link

Windows D3D12:

To protect your privacy, external media requires approval to load.

Source: i.imgur.com

View cookie preferences.
Accept & Show Accept All & Don't show this again   Direct Link

Edit: Fixed screenshot order.

Extremely good results I would say. Only things that are missing are HBAO+ (as usual) and RTX shadows (but those wouldn't be supported on my GPU anyway).

Note to AMD users: ACO currently causes particles to flicker.

View this comment - View article - View full comments

Quoting: x_wingI know that I said this before but: I wish we had something like the Mesa matrix page in order to know how far we are for a complete implementation of D3D9/11

For DX11: https://github.com/doitsujin/dxvk/wiki [External Link]

Shared Resources are straight-up impossible to support without additional Wine work, Class Linkage is a huge mess that nobody uses (except for one single Microsoft demo), Predication is technically implemented but broken on most drivers for some reason, Target-independent rasterization is impossible to support properly becuause no Vulkan driver supports 16xMSAA, multi-planar formats are exceptionally hard to support because of D3D<>Vk differences (but again, nobody uses those), and the optional features are... well, optional, and require very recent hardware to work even on Windows, and nobody uses those.

In other words, all features that games actually use are supported, except the border color thing, but we need a Vulkan extension for that.

I think D9VK has more or less feature parity with typical D3D9 drivers as well, except border colors (again) and some depth bias quirks (might need Vulkan extension).

View this comment - View article - View full comments

Quoting: subSpecifically for Navi, would you suggest using AMDVLK or RADV?

I don't have any personal experience with Navi, but from what I'm hearing from users, it doesn't seem to be stable at all yet, on any driver; the kernel driver is apparently also causing major headache (read: system hangs and fun stuff like that). I'd give it a few more months.

View this comment - View article - View full comments

Quoting: 1xokWould you mind to elaborate a bit about AMD's Vulkan driver state on Windows?
What's so bad about it?

It's fairly buggy even compared to the open-source AMDVLK driver, which would indicate issues in the proprietary shader compiler.

It wouldn't be so bad if things were at least consistently broken, but they aren't. One driver update might randomly fix an issue, the next update might fix something else but break the very same thing they fixed in the earlier update. It's a rollercoaster, and honestly I just can't be arsed to report bugs against it.

AMD's Windows driver has been a broken mess ever since Navi launched anyway, with D3D9 being completely broken in some cases and other major issues, but that's a different story.

View this comment - View article - View full comments

Quoting: Comandante ÑoñardoIt is enabled.

Then there's nothing you can do besides maybe trying wined3d. D9VK and DXVK do need more memory, this is a known problem with 32-bit apps.

View this comment - View article - View full comments

Then nevermind, it just happens because it's a 32-bit game and runs out of address space. Please mention such things in the future.

Try PROTON_FORCE_LARGE_ADDRESS_AWARE=1.

View this comment - View article - View full comments

It means that you're running into this [External Link] problem with the Nvidia driver. No official fix available at this time, although they know what's causing it.

View this comment - View article - View full comments

Quoting: LinuxwarperIs it primarily harder because there aren't solutions in place or because the way D3D12 works? Like if Vulkan were to have feature parity with D3D12, if that's right way to put it, would that make mapping to Vulkan easier?

Both. There are some extensions in the works that will solve some problems, but others are caused by fundamental differences in the API - Vulkan is a lot more explicit, and that's where things get really nasty. Right now, vkd3d technically violates the Vulkan spec quite badly because of that, and fixing it will be very hard and potentially come at a cost.

View this comment - View article - View full comments

Quoting: TheRiddickI would like some coding to be done to allow better freesync or VRR support, maybe allowing it to function regardless of monitor(s) setups.

All of that is down to the Linux graphics stack, games/dxvk/whatever can't really do anything at all to improve the situation.

Quoting: 1xokWhat are the advantages of DX12 over Vulkan for developers? I can only think of the Xbox support.

Well, for starters, D3D12 works on all Windows 10 systems and has excellent driver support from all vendors.

Windows drivers shipping with Windows Update sometimes don't come with Vulkan support -> games straigt-up don't work. Driver quality is also way worse (AMD is pretty bad, Intel seems to be especially bad, only Nvidia is acceptable these days - and that also was a whole different story ~2 years ago).

Tools are fine these days, but developers who are used to the Microsoft ecosystem will want to use Microsoft's tools, which obviously don't support Vulkan either.

For end users, Vulkan games are always annoying because stuff like OBS doesn't support Vulkan properly.

There's a ton of reasons why we probably won't see many Vulkan games on Windows, but the API itself isn't really one of them (I've seen people rant about how much better designed D3D12 is, but I'd wholeheartedly disagree there).

Quoting: XpanderWorld of Warcraft Classic started having heavy freezes after reaching to new areas or completing quests, like 1-2 sec freezes.

Sounds like your average shader compiler stutter. 1.4.3 does change some shader code and WoW doesn't really make use of the state cache as well as it could.

There have been no other changes in the code that could cause this.

Quoting: TheRiddickWoW Classis uses DirectX9 doesn't it? is that not D9VK?

WoW classic only supports D3D11.

View this comment - View article - View full comments