v2.0.5: GPU time ~8x higher than v2.0.4 (~10 fps vs ~45), RX 6700 XT #823

Closed

opened 2026-11-08 22:36:04 +00:00 by wheezypulls-hue · 2 comments

wheezypulls-hue commented

2026-11-08 22:36:04 +00:00

(Migrated from github.com)

Hookshot version

v2.0.5

What kind of problem

Graphics (missing, black or wrong-coloured things, lighting, flicker)

What happened

After updating from v2.0.4 to v2.0.5, in-game fps dropped from ~45 to ~10
in the same areas. Menus still run at 60.

From the logs (both attached):

v2.0.4: GPU time ~8-10 ms a frame, 0 shaders failed
v2.0.5: GPU time ~70-140 ms a frame, 165 shaders failed
CPU replay time is about the same in both (~10-12 ms)

RX_VK_CHECKPOINTS is off and frame interpolation is 2x in the v2.0.5 log,
same as v2.0.4.

Where in the game

Kokiri Forest

How often

Every time

Graphics card and driver version

AMD Radeon RX 6700 XT (12 GB), Vulkan 1.4, driver 0x800161, 2560x1440

CPU, RAM and operating system

AMD Ryzen 5 7500F 6-Core Processor (3.70 GHz), RAM: 32 GB, Windows 11 Home, 64-bit

Changes from the defaults

No response

Other software hooked into the game

No, nothing hooked in

hookshot_log.txt

hookshot_log.2.txt

Screenshots or video

No response

Before you send it

I'm on the newest release, or the problem is still there on it.
I searched the open issues for this problem.
I'm not attaching game files, keys, hookshot.exe, the .bin caches or saves.

### Hookshot version v2.0.5 ### What kind of problem Graphics (missing, black or wrong-coloured things, lighting, flicker) ### What happened After updating from v2.0.4 to v2.0.5, in-game fps dropped from ~45 to ~10 in the same areas. Menus still run at 60. From the logs (both attached): - v2.0.4: GPU time ~8-10 ms a frame, 0 shaders failed - v2.0.5: GPU time ~70-140 ms a frame, 165 shaders failed - CPU replay time is about the same in both (~10-12 ms) RX_VK_CHECKPOINTS is off and frame interpolation is 2x in the v2.0.5 log, same as v2.0.4. ### Where in the game Kokiri Forest ### How often Every time ### Graphics card and driver version AMD Radeon RX 6700 XT (12 GB), Vulkan 1.4, driver 0x800161, 2560x1440 ### CPU, RAM and operating system AMD Ryzen 5 7500F 6-Core Processor (3.70 GHz), RAM: 32 GB, Windows 11 Home, 64-bit ### Changes from the defaults _No response_ ### Other software hooked into the game No, nothing hooked in ### hookshot_log.txt [hookshot_log.txt](https://github.com/user-attachments/files/34556586/hookshot_log.txt) [hookshot_log.2.txt](https://github.com/user-attachments/files/34556589/hookshot_log.2.txt) ### Screenshots or video _No response_ ### Before you send it - [x] I'm on the newest release, or the problem is still there on it. - [x] I searched the open issues for this problem. - [x] I'm not attaching game files, keys, hookshot.exe, the .bin caches or saves.

lucasmazzoni45-del commented

2026-11-08 23:49:44 +00:00

(Migrated from github.com)

The first-start prepare also ends with 165 failed here, all does not lower ... unsupported native .... In play the failed count stays at 179 and no draw is skipped, and GPU time averages 8.7 ms (13-14 ms in the heaviest scenes), so the 165 failed programs themselves do not seem to cost GPU time (they look like census extras that are never drawn). I would not look at them first.
Two things I would compare between your v2.0.4 and v2.0.5 logs instead (hypotheses, I have not seen your logs):
1. The vk: VRAM: lines (used vs allowed, and the texture budget). v2.0.5 includes #705, which re-encodes the big terrain texture arrays at an even-halving size (1365 -> 1536) and so uses more VRAM; on a 12 GB card that could tip into paging, which shows up as a very high GPU time with a normal CPU replay time.
2. How often textures retired appears and how many each time. With a working set bigger than the budget it repeats every few seconds.
If the VRAM: line shows used close to allowed, trying a lower texture budget (Quality preset for your VRAM, or RX_VK_VRAM_MB=<mb> in hookshot_env.txt) would tell whether that is it.

Some data points from another AMD card that may help narrow this down (RX 7800 XT 16 GB, Windows 11, v2.0.5, 1440p; my build has a few local changes, listed in #865, so take the numbers as indicative only): - The first-start prepare also ends with `165 failed` here, all `does not lower ... unsupported native ...`. In play the failed count stays at 179 and no draw is skipped, and GPU time averages 8.7 ms (13-14 ms in the heaviest scenes), so the 165 failed programs themselves do not seem to cost GPU time (they look like census extras that are never drawn). I would not look at them first. - Two things I would compare between your v2.0.4 and v2.0.5 logs instead (hypotheses, I have not seen your logs): 1. The `vk: VRAM:` lines (used vs allowed, and the texture budget). v2.0.5 includes #705, which re-encodes the big terrain texture arrays at an even-halving size (1365 -> 1536) and so uses more VRAM; on a 12 GB card that could tip into paging, which shows up as a very high GPU time with a normal CPU replay time. 2. How often `textures retired` appears and how many each time. With a working set bigger than the budget it repeats every few seconds. - If the `VRAM:` line shows used close to allowed, trying a lower texture budget (Quality preset for your VRAM, or `RX_VK_VRAM_MB=<mb>` in `hookshot_env.txt`) would tell whether that is it.

NotRexed commented

2026-11-09 05:23:35 +00:00

(Migrated from github.com)

Closing as fixed in v2.0.6: the terrain re-encoding from v2.0.5 (#705) that raised the RX 6700 XT GPU time from ~10 ms to 70-100 ms is off again (RX_VK_UNEVEN_BC=0 turns it back on for testing). Please open a new issue with hookshot_log.txt if v2.0.6 is still slow on your card.

Commenting is not possible because the repository is archived.