Game: grid pattern on distant (low-res) textures on AMD Windows, fixed by RX_VK_BC_SMALL_LAYERS=1 #2222
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
3 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot#2222
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Longshot version
v3.0.0-176-gd76d202f (main, source 22ef204e)
Channel
Built from source
What kind of problem
Graphics (missing, black or wrong-coloured things, lighting, flicker)
What happened
A regular grid pattern is visible on textures in the distance: ground, paths and hillsides beyond the near draw range (Hyrule Field, the path toward Castle Town, Kokiri Forest's surroundings). Textures close to the camera look correct; the grid shows only on the lower-resolution versions used further away and disappears as you walk up to it.
It is not shadows: changing Shadow resolution makes no visible difference.
Workaround:
RX_VK_BC_SMALL_LAYERS=1in longshot_env.txt removes the grid completely (checked in game at the same spots). So it looks like the small ASTC->BC scratch path for levels of 1024 or less (#1843), possibly the same kind of pitch problem #2015 fixed for the 2048 x 2048 scratch ("Hyrule Field's grid lines on AMD Windows"), but in the small 8-layer scratch. The log says "7379 MB of VRAM allowed: all large ASTC textures kept as BC3".RX_VK_ASTC_RGBA8was not needed.The rc.7 build (
544229a, before #1843) did not show the grid on the same PC.Related: #1771 (square shadow artifacts behind grass and leaves) may be this same bug if that reporter is on AMD.
Note: the build has one small local change to swapchain creation only (an exclusive-fullscreen option); nothing in the texture path is changed.
Where in the game
Hyrule Field and the paths toward Castle Town / Kokiri Forest, looking at ground and hills in the distance
How often
Every time
Graphics card and driver version
AMD Radeon RX 6600 (Navi 23), Adrenalin 26.11.2 (32.0.21045.29003)
CPU, RAM and operating system
AMD FX-8350, 32 GB, Windows 10 Pro 22H2 (19045)
Changes from the defaults
Quality preset Auto (this PC), V-Sync On, fullscreen, no frame generation, Grass distance Game's. No overlays, ReShade or injectors.
longshot_env.txt:
(plus the local exclusive-fullscreen switch)
longshot_log.txt
Relevant lines (full log on request):
Screenshots or video
No response
Before you send it
This looks like the same problem as #2006 (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)), so I'm closing this one to keep the discussion in one place.
Both report the same distant-ground grid artifact on AMD's Windows Vulkan driver, removed by RX_VK_BC_SMALL_LAYERS=1 and caused by multi-layer ASTC-to-BC re-encoding of small mip levels.
If it's different, say how and a maintainer will reopen it.
The close says this is a duplicate of #2006. This report is on
v3.0.0-176-gd76d202f(main, source 22ef204e), a build that containsed602b94(#2015, the commit that closed #2006), and the grid is still there on the default path; onlyRX_VK_BC_SMALL_LAYERS=1removes it.Please reopen; the evidence is on this issue.
Same group: this is now fixed only by the open PR #2228 (stacked on the also-open #2227), not by a merged commit. Please reopen against #2107.