Game: dark grid lines over Hyrule Field's ground on AMD's Windows driver since #1843 (RX 6700 XT; gone with RX_VK_BC_SMALL_LAYERS=1) #2006

Closed
opened 2026-11-21 04:04:32 +00:00 by ballotstomper · 3 comments
Contributor

Longshot version

Not known (the player has no log).

What kind of problem

Graphics

What happened

Dark grid lines across Hyrule Field's ground: thin dark lines in a regular grid over the terrain, most visible on the far
ground (screenshot in the first comment). Reported by a Windows player with an AMD Radeon RX 6700 XT and a Ryzen 5 5600X.

RX_VK_BC_SMALL_LAYERS=1 in longshot_env.txt removes them (the player confirmed). That switch brings back one layer a run for the ASTC-to-BC re-encode of mip levels of 1024 x 1024 or less, which #1843 made eight layers a run through a small scratch. The grid lines are the ones #1843's description names: one layer a run was the fix for Hyrule Field's grid lines on an RX 6600 (ec89f92e), whose cause was never found, and #1843 checked eight layers a run only on RDNA 4 with RADV. Its Still to test listed Windows AMD and RX 6000 cards.

Not seen on Linux on the same GPU family: an RX 6750 XT (Navi 22, as the RX 6700 XT) on Mesa RADV 26.2.7, with #1843 in the build, shows no grid lines in Hyrule Field. So it looks like AMD's Windows driver (at least on RDNA 2) with multi-layer runs.

Related: #1987 (WIP) re-encodes a texture's levels in one run, #1843's idea applied further; worth checking against this before it merges.

Where in the game

Hyrule Field, looking across the ground towards the western slopes; while riding Epona in the screenshot.

How often

Every time (as reported)

Graphics card and driver version

AMD Radeon RX 6700 XT, AMD's Windows driver 26.10.1

CPU, RAM and operating system

AMD Ryzen 5 5600X, 32 GB, Windows (version not known)

Changes from the defaults

None needed to see it. RX_VK_BC_SMALL_LAYERS=1 makes it go away.

Screenshots or video

In the first comment (the player's, posted with their permission).

### Longshot version Not known (the player has no log). ### What kind of problem Graphics ### What happened Dark grid lines across Hyrule Field's ground: thin dark lines in a regular grid over the terrain, most visible on the far ground (screenshot in the first comment). Reported by a Windows player with an AMD Radeon RX 6700 XT and a Ryzen 5 5600X. `RX_VK_BC_SMALL_LAYERS=1` in `longshot_env.txt` removes them (the player confirmed). That switch brings back one layer a run for the ASTC-to-BC re-encode of mip levels of 1024 x 1024 or less, which #1843 made eight layers a run through a small scratch. The grid lines are the ones #1843's description names: one layer a run was the fix for Hyrule Field's grid lines on an RX 6600 (ec89f92e), whose cause was never found, and #1843 checked eight layers a run only on RDNA 4 with RADV. Its Still to test listed Windows AMD and RX 6000 cards. Not seen on Linux on the same GPU family: an RX 6750 XT (Navi 22, as the RX 6700 XT) on Mesa RADV 26.2.7, with #1843 in the build, shows no grid lines in Hyrule Field. So it looks like AMD's Windows driver (at least on RDNA 2) with multi-layer runs. Related: #1987 (WIP) re-encodes a texture's levels in one run, #1843's idea applied further; worth checking against this before it merges. ### Where in the game Hyrule Field, looking across the ground towards the western slopes; while riding Epona in the screenshot. ### How often Every time (as reported) ### Graphics card and driver version AMD Radeon RX 6700 XT, AMD's Windows driver 26.10.1 ### CPU, RAM and operating system AMD Ryzen 5 5600X, 32 GB, Windows (version not known) ### Changes from the defaults None needed to see it. `RX_VK_BC_SMALL_LAYERS=1` makes it go away. ### Screenshots or video In the first comment (the player's, posted with their permission).
Author
Contributor

The player's screenshot (Windows, RX 6700 XT, AMD driver 26.10.1), before setting RX_VK_BC_SMALL_LAYERS=1:

Dark grid lines over Hyrule Field's ground

The player's screenshot (Windows, RX 6700 XT, AMD driver 26.10.1), before setting RX_VK_BC_SMALL_LAYERS=1: ![Dark grid lines over Hyrule Field's ground](https://git.magicbean.dev/attachments/4e047b67-a937-4bae-b39d-eff3be425b81)
Contributor

A data point for narrowing this down. On Windows 11 with NVIDIA (RTX 3080, driver 621.14), main @ a7646537 (with #1843) showed no grid lines in Hyrule Field over three starts at the default 8 layers. I looked at the field from the Castle Town gate and in 8 shots each while running round it. Three more starts with RX_VK_BC_SMALL_LAYERS=1 looked the same: #1843 (comment).

The caveat is that the game's clock had it at night in every run, so the ground was dim. With the RX 6750 XT on RADV also clean, it does point at AMD's Windows driver rather than Windows as such.

Posted by an AI agent (Claude Code) for d4nkn3ss.

A data point for narrowing this down. On Windows 11 with NVIDIA (RTX 3080, driver 621.14), main @ a7646537 (with #1843) showed no grid lines in Hyrule Field over three starts at the default 8 layers. I looked at the field from the Castle Town gate and in 8 shots each while running round it. Three more starts with `RX_VK_BC_SMALL_LAYERS=1` looked the same: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/1843#issuecomment-61700. The caveat is that the game's clock had it at night in every run, so the ground was dim. With the RX 6750 XT on RADV also clean, it does point at AMD's Windows driver rather than Windows as such. *Posted by an AI agent (Claude Code) for d4nkn3ss.*
Member

Possibly related, for context:

  • PR #1843 perf(vk): ASTC levels of 1024 or less re-encode to BC eight layers a run, through a small scratch — Merged change introducing multi-layer runs; explicitly identified grid-line risk and Windows AMD remains untested.
  • PR #1987 WIP: perf(vk): a BC3-kept ASTC texture's levels re-encode in one run, and a GPU-drawn target is not uploaded again for a buffer write (test switches) — Open WIP extends multi-layer re-encoding; its testing plan specifically calls out RDNA 2 grid-line risk.
Possibly related, for context: - PR #1843 perf(vk): ASTC levels of 1024 or less re-encode to BC eight layers a run, through a small scratch — Merged change introducing multi-layer runs; explicitly identified grid-line risk and Windows AMD remains untested. - PR #1987 WIP: perf(vk): a BC3-kept ASTC texture's levels re-encode in one run, and a GPU-drawn target is not uploaded again for a buffer write (test switches) — Open WIP extends multi-layer re-encoding; its testing plan specifically calls out RDNA 2 grid-line risk.
Sign in to join this conversation.
No milestone
No project
No assignees
3 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot#2006
No description provided.