[Regression] Need for Speed: Rivals [BLUS31201] Squid Ink

[Darkhost1999]

The issue was a regression from #9848 showcasing that black splotches took up most of the screen in NFS Rivals.
Supposedly a Turing or newer graphics card issue.

NFS.Rivals.1-2-2022.mp4

RPCS3.log.gz

Some things to note was that the issue was only present at 100% scale not when upscaling. But performance is/was too slow to desire upscaling that when 100% scaling was resolved performance improved a lot.

[illusion0001]

Architecture memes, "eu" images were taken with pascal card.

[kd-11]

I've added to the backlog, though I do not have anything Turing+ on the NVIDIA side.

[kd-11]

I did some testing. Unfortunately I cannot reproduce this bug as reported.
Tested with default settings + WCB + vulkan.

1. AMD proprietary - technically works, but has a driver bug that is messing up the tiles and they don't render.
2. AMD mesa (RADV) - I played for like 30 minutes. Visuals are mostly correct, terrain loads in as expected. The only issue is image is blurry and performance is poor,
3. NVIDIA (pascal) - Renders correctly like RADV for the most part.
   I cannot see any obvious issues with the PR in the report for the moment.

[Darkhost1999]

That's OK, 150%+ resolution scaling provides a clear image.
I just don't know why the 100% resolution scale is broken for Turing and Ampere.

[kd-11]

I committed a workaround for the AMD proprietary driver's point rasterization issue. It was a regression due to some odd interaction between line and point state in those drivers.
I still have no way to reproduce the squid ink, I'll have to revisit if I get a newer nvidia card at some point.

[Darkhost1999]

I have singled out reverting eb01b88 resolves this issue.

Edit: November 24th:
Confirmed to fix it on Linux for me as well.

[kd-11]

Thanks.