MagicBeanDev/Hookshot

Archived

Watch

Fork 0

Game presentation freezes, gameplay and audio continue. Toggling Fullscreen fixes it. #2185

Closed

opened 2026-11-11 06:48:52 +00:00 by jedinjapan · 5 comments

jedinjapan commented

2026-11-11 06:48:52 +00:00

(Migrated from github.com)

Hookshot version

v2.1.0

What kind of problem

Crash, freeze or black screen

What happened

Render thread freezes while combat/gameplay continues in background. Inputs and sound continue, and the game has not crashed. Render pipeline seems to become unstuck by toggling Fullscreen, which has been effective each time so far.

Where in the game

Hyrule Field, Kakriko Village, when entering and exiting interior to exterior doorways quickly.

How often

Sometimes

Graphics card and driver version

9070XT 26.10.2

CPU, RAM and operating system

Ryzen 7 9800X3D, 32GB, Windows 11

Changes from the defaults

Using FSR 3.1 with auto scaling, and Framegen at 2x, with sharpening at 50%.

Other software hooked into the game

No, nothing hooked in

hookshot_log.txt

hookshot_log.1.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 v2.1.0 ### What kind of problem Crash, freeze or black screen ### What happened Render thread freezes while combat/gameplay continues in background. Inputs and sound continue, and the game has not crashed. Render pipeline seems to become unstuck by toggling Fullscreen, which has been effective each time so far. ### Where in the game Hyrule Field, Kakriko Village, when entering and exiting interior to exterior doorways quickly. ### How often Sometimes ### Graphics card and driver version 9070XT 26.10.2 ### CPU, RAM and operating system Ryzen 7 9800X3D, 32GB, Windows 11 ### Changes from the defaults Using FSR 3.1 with auto scaling, and Framegen at 2x, with sharpening at 50%. ### Other software hooked into the game No, nothing hooked in ### hookshot_log.txt [hookshot_log.1.txt](https://github.com/user-attachments/files/34674268/hookshot_log.1.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.

bunbeans2-arch commented

2026-11-11 07:24:42 +00:00

(Migrated from github.com)

This looks like what we reported in #2180, on the same card: RX 9070 XT, v2.1.0, Windows 11, 3840x2160 HDR. Adding what our log shows in case it helps narrow it down.

The picture froze during steady 2x frame generation, 79 s after the last change in the Insert menu (the options file's save time against the time the picture stopped). The "frame interpolation" reports show 2x right up to the end, for example 504 frames shown in a report at about 50 game fps.
AMD FSR 3.1 had been the upscaler earlier in that run and was switched back to the game's own before the freeze.
After the last "presents" line the log has no more lines from the Presentation thread or from sead::MainThread; the rumble thread and the worker threads keep logging. No device-lost line.
VRAM was close to what the driver allows just before: "VRAM: 13024 MB used of 14706 MB allowed" two reports before the end, then "write-back diff: VRAM 14706 of 14706 MB: the shadows given back".

We did not know about the fullscreen toggle, so we ended the process. Next time we will try it first, and we are now running with RX_VK_HANG_DUMP set so a repeat gives the thread stacks.

@jedinjapan, does the tail of your log also show the VRAM line near its allowed figure before the freeze, or was there plenty of room?

This looks like what we reported in #2180, on the same card: RX 9070 XT, v2.1.0, Windows 11, 3840x2160 HDR. Adding what our log shows in case it helps narrow it down. - The picture froze during steady 2x frame generation, 79 s after the last change in the Insert menu (the options file's save time against the time the picture stopped). The "frame interpolation" reports show 2x right up to the end, for example 504 frames shown in a report at about 50 game fps. - AMD FSR 3.1 had been the upscaler earlier in that run and was switched back to the game's own before the freeze. - After the last "presents" line the log has no more lines from the Presentation thread or from sead::MainThread; the rumble thread and the worker threads keep logging. No device-lost line. - VRAM was close to what the driver allows just before: "VRAM: 13024 MB used of 14706 MB allowed" two reports before the end, then "write-back diff: VRAM 14706 of 14706 MB: the shadows given back". We did not know about the fullscreen toggle, so we ended the process. Next time we will try it first, and we are now running with `RX_VK_HANG_DUMP` set so a repeat gives the thread stacks. @jedinjapan, does the tail of your log also show the VRAM line near its allowed figure before the freeze, or was there plenty of room?

bunbeans2-arch commented

2026-11-11 11:28:02 +00:00

(Migrated from github.com)

If you can still reproduce it: does RX_VK_PRESENT_AHEAD=0 change it? On our 9070 XT a stop like this needed present ahead together with RX_VK_PRESENT_QUEUE=1; numbers in #1702: #1702

If you can still reproduce it: does `RX_VK_PRESENT_AHEAD=0` change it? On our 9070 XT a stop like this needed present ahead together with `RX_VK_PRESENT_QUEUE=1`; numbers in #1702: https://git.magicbean.dev/MagicBeanDev/Hookshot/issues/1702

jedinjapan commented

2026-11-11 18:04:27 +00:00

(Migrated from github.com)

I will attempt to reproduce this again this afternoon after updating to the latest build. I didn't know what event to skim the log for, and as it was a long play session the log was over 4mb.

Separate unrelated note: is the log writing directly for every event? Or can this be batched later. Thinking of SSD writes down the road...

If I encounter the freeze next session, I'll attempt the full screen toggle and if it works, I'll exit after and upload the log here.

I will attempt to reproduce this again this afternoon after updating to the latest build. I didn't know what event to skim the log for, and as it was a long play session the log was over 4mb. Separate unrelated note: is the log writing directly for every event? Or can this be batched later. Thinking of SSD writes down the road... If I encounter the freeze next session, I'll attempt the full screen toggle and if it works, I'll exit after and upload the log here.

jedinjapan commented

2026-11-11 18:23:39 +00:00

(Migrated from github.com)

If you can still reproduce it: does RX_VK_PRESENT_AHEAD=0 change it? On our 9070 XT a stop like this needed present ahead together with RX_VK_PRESENT_QUEUE=1; numbers in #1702: #1702 (comment)

After reading through that comment thread, I don't believe I have present ahead enabled, I had frame pacing disabled, as I had frame generation turned rather high and had more than enough GPU and CPU overhead to allow for that. So I wasn't worried about the frame pacing.

> If you can still reproduce it: does `RX_VK_PRESENT_AHEAD=0` change it? On our 9070 XT a stop like this needed present ahead together with `RX_VK_PRESENT_QUEUE=1`; numbers in [#1702](https://git.magicbean.dev/MagicBeanDev/Hookshot/issues/1702): [#1702 (comment)](https://git.magicbean.dev/MagicBeanDev/Hookshot/issues/1702) After reading through that comment thread, I don't believe I have present ahead enabled, I had frame pacing disabled, as I had frame generation turned rather high and had more than enough GPU and CPU overhead to allow for that. So I wasn't worried about the frame pacing.

NotRexed commented

2026-11-11 21:26:14 +00:00

(Migrated from github.com)

Thanks for the report. Reports of the same problem are now collected in #996, which lists this one with your details and links back here, so one fix closes them together and the open list stays readable. Follow #996 for the fix, and add new details there (or here). If this turns out to be a separate problem it will be reopened.

Commenting is not possible because the repository is archived.