MagicBeanDev/Hookshot

Archived

Watch

Fork 0

[Tracking] Frame generation, frame pacing and the frame rate cap #2510

Open

opened 2026-11-11 21:21:47 +00:00 by NotRexed · 2 comments

NotRexed commented

2026-11-11 21:21:47 +00:00

(Migrated from github.com)

The frame generation rework (the game's own rate kept, the shown rate is the game's rate times the multiplier) is in progress.

Reports

#1588: FSR 3.1 frame interpolation at 2x draws the HUD into generated frames: hearts, magic bar and button icons doub
#1650: The FPS cap does not appear to be working correctly
#1652: Frame generation 2x/3x holds the game at 30/20 fps on a 60 Hz display (RTX 2080), 45+ fps with it off
#1702: Frame generation 2x: frames reach the screen 5.7 / 11 ms apart instead of 8.3 (VRR flicker), with cause and a (v2.0.12 (v2.0.13 changed none of this code; the patch applies to v2.0.; RTX 5080, NVIDIA 617.42; Ryzen 7 7800X3D, 32 GB, Windows 11 Pro (26200))
#1797: Frame generation at a render resolution below 100% (4K output, 67% render): generated frames only cover the to (v2.0.13 (v2.0.14 came out today as a hotfix for HDR, the menu and brac; RTX 4090, NVIDIA 621.01 (the system also has the Ryzen's integrated AM; Ryzen 7 9800X3D, 32 GB, Windows 11 Pro 25H2)
#2244: Camera doesn't feel smooth (v2.1.0; Rtx 5070, NVIDIA 616.92; Ryzen 7 7800X3D, 32GB Ram, Win11 25H2)
#2372: the frame pacing and micro stutter make the games dlss ghost more (2.10; RTX 4070 VER621.01; INTELL I7 16 GB WIN 11)
#2389: Frame pacing is not smooth (2.1.0; Nvidia 4080 Super, v621.01, gsync display; 9800X3D, 32GB DDR5 6000+, Windows 11)

These reports were collected here so one fix closes them all and the open list stays readable. Add new details as a comment here (version, card and driver, where in the game, and the hookshot_log.txt from the game folder).

The frame generation rework (the game's own rate kept, the shown rate is the game's rate times the multiplier) is in progress. ## Reports - #1588: FSR 3.1 frame interpolation at 2x draws the HUD into generated frames: hearts, magic bar and button icons doub - #1650: The FPS cap does not appear to be working correctly - #1652: Frame generation 2x/3x holds the game at 30/20 fps on a 60 Hz display (RTX 2080), 45+ fps with it off - #1702: Frame generation 2x: frames reach the screen 5.7 / 11 ms apart instead of 8.3 (VRR flicker), with cause and a (v2.0.12 (v2.0.13 changed none of this code; the patch applies to v2.0.; RTX 5080, NVIDIA 617.42; Ryzen 7 7800X3D, 32 GB, Windows 11 Pro (26200)) - #1797: Frame generation at a render resolution below 100% (4K output, 67% render): generated frames only cover the to (v2.0.13 (v2.0.14 came out today as a hotfix for HDR, the menu and brac; RTX 4090, NVIDIA 621.01 (the system also has the Ryzen's integrated AM; Ryzen 7 9800X3D, 32 GB, Windows 11 Pro 25H2) - #2244: Camera doesn't feel smooth (v2.1.0; Rtx 5070, NVIDIA 616.92; Ryzen 7 7800X3D, 32GB Ram, Win11 25H2) - #2372: the frame pacing and micro stutter make the games dlss ghost more (2.10; RTX 4070 VER621.01; INTELL I7 16 GB WIN 11) - #2389: Frame pacing is not smooth (2.1.0; Nvidia 4080 Super, v621.01, gsync display; 9800X3D, 32GB DDR5 6000+, Windows 11) These reports were collected here so one fix closes them all and the open list stays readable. Add new details as a comment here (version, card and driver, where in the game, and the hookshot_log.txt from the game folder).

denniskonietzko-lab commented

2026-11-12 03:03:11 +00:00

(Migrated from github.com)

FG Auto reduces native FPS on RTX 4060 at 120 Hz (v2.1.0)

Additional measured report for the frame-generation tracking issue, which requests hardware, location and a game log as a comment rather than a duplicate issue.[1]

Summary

Enabling Frame generation Auto reduces the game's native frame rate while the generated-frame output stays close to 120 FPS. The game feels substantially worse with FG enabled, so I normally leave it off.

The 120 FPS output target itself is not the complaint: Windows is set to 120 Hz, and Hookshot logs the same refresh rate. The concern is the loss of native FPS and the worse experience when enabling FG.

Hardware and version

Hookshot: v2.1.0 (source 633eb435), from the settings screenshot and log line 2.
GPU: NVIDIA GeForce RTX 4060, 8 GB VRAM.
Driver: Vulkan raw version 0x9b438000, as logged on lines 35 and 46. NVIDIA version fields decode to 621.14.0.0; the installed package version has not been separately checked in NVIDIA's control panel.
CPU: AMD Ryzen 5 5500 (player supplied).
RAM: 32 GB DDR4-3200 (player supplied).
OS: Windows; exact edition/build was not supplied for this capture.
Display refresh rate: 120 Hz, confirmed by the player in Windows and by Hookshot.

Settings shown in the attached screenshot

Quality preset: Balanced (8GB).
Fullscreen: On.
Game frame rate cap: 60 (the console's).
Frame pacing: On.
NVIDIA Reflex: Off in the settings screenshot.
HDR: Off.
Output resolution setting: 2048 x 1152 (game's).
Render resolution: 100% (2048 x 1152).
Upscaler: Game's own, not DLSS.
Window scaling filter: Catmull-Rom; sharpness 30%.
V-Sync: not visible in the supplied settings screenshot; not confirmed.

The log separately records a 1920 x 1080 window swapchain (lines 57-60); that is distinct from the game's output-resolution setting above.

Reproduction / comparison

Load the Temple of Time as child Link and stand on the circular Triforce platform, facing the camera toward the sealed doorway, as in the screenshots.
With FG Off, note the built-in FPS counter.
Enable FG Auto and compare the built-in total FPS and game FPS at the same position and similar camera angle.

The player supplied the two comparison screenshots and confirmed that the FG-on mode was Auto:

Capture	Native game FPS	Total output FPS
FG Off (`02-FG-off.png`)	51	51
FG Auto (`03-FG-auto.png`)	38	119

These are individual screenshots, not averaged benchmark results.

Log evidence around the FG activation

Line numbers refer to the attached hookshot_log-redacted.txt; privacy redaction preserves the original line count.

Logged metric	Before activation	After activation
Renderer `presents` FPS	54.1 (line 4403)	37.4 (line 4498)
Replay `submit` time	10.41 ms (line 4401)	18.16 ms (line 4496)
GPU time per frame	8.0 ms (line 4406)	9.5 ms (line 4501)

At line 4418, FG starts with:

vk: frame interpolation ready: Auto, a frame for each of the display's refreshes (120 Hz; up to 4 a game frame)

The first full following vi interval reports 37.5 native FPS over 5 seconds (line 4482). The interpolation log separately reports 595 frames shown on time, 0 late (line 4481).

Around the switch, VRAM usage is 5792 MB of a 6614 MB budget (line 4442). The selected before/after samples report 0 shader/pipeline waits and 0 queued translation jobs (lines 4400 and 4495). These observations do not establish the root cause or exclude every memory/synchronization issue.

Expected result and investigation limits

Some FG overhead is expected. I would appreciate investigation of whether this native-FPS loss can be reduced or whether the capture matches an already-known issue. This report does not claim that FG is free, that 120 Hz detection is wrong, or that any particular pending patch fixes this case.

The player has also observed native FPS dropping to roughly 15-20 in other situations, and output tending toward 120 with fixed multipliers as well. Those additional observations are not established by these screenshots or this Auto-mode log, so they should not be treated as separately reproduced here.

This is one player-recorded run, not a repeated Off/On/Off benchmark. The full log includes earlier play, configuration changes and some foreground-window changes; not every interval is a controlled comparison. Mods, overlays/injectors, exact Windows build, and whether Reflex stayed Off throughout the entire run have not been independently confirmed. No separate Reflex regression is claimed.

Attachments and privacy

01-settings.png: supplied graphics-settings screenshot.
02-FG-off.png: 51 FPS comparison screenshot.
03-FG-auto.png: 119 total / 38 game FPS comparison screenshot.
hookshot_log-redacted.txt: full supplied log; the Windows account name in local paths and the save-directory identifier have been replaced with placeholders. Timing and rendering evidence is unchanged.

No game files, generated executables, shader/program caches, saves or pipeline-state binary are included. The pipeline-list contribution is a separate topic.

Prepared with AI assistance from the player's screenshots, log and hardware details; observations and unverified causes are kept separate.

Sources

[1] #2510 — Hookshot frame generation tracking issue

Evidence bundle

Download the report, three screenshots and redacted full log (ZIP)

# FG Auto reduces native FPS on RTX 4060 at 120 Hz (v2.1.0) Additional measured report for the frame-generation tracking issue, which requests hardware, location and a game log as a comment rather than a duplicate issue.[1] ## Summary Enabling Frame generation **Auto** reduces the game's native frame rate while the generated-frame output stays close to 120 FPS. The game feels substantially worse with FG enabled, so I normally leave it off. The 120 FPS output target itself is not the complaint: Windows is set to **120 Hz**, and Hookshot logs the same refresh rate. The concern is the loss of native FPS and the worse experience when enabling FG. ## Hardware and version - Hookshot: **v2.1.0 (source 633eb435)**, from the settings screenshot and log line 2. - GPU: **NVIDIA GeForce RTX 4060, 8 GB VRAM**. - Driver: Vulkan raw version **0x9b438000**, as logged on lines 35 and 46. NVIDIA version fields decode to **621.14.0.0**; the installed package version has not been separately checked in NVIDIA's control panel. - CPU: **AMD Ryzen 5 5500** (player supplied). - RAM: **32 GB DDR4-3200** (player supplied). - OS: **Windows**; exact edition/build was not supplied for this capture. - Display refresh rate: **120 Hz**, confirmed by the player in Windows and by Hookshot. ## Settings shown in the attached screenshot - Quality preset: **Balanced (8GB)**. - Fullscreen: **On**. - Game frame rate cap: **60 (the console's)**. - Frame pacing: **On**. - NVIDIA Reflex: **Off** in the settings screenshot. - HDR: **Off**. - Output resolution setting: **2048 x 1152 (game's)**. - Render resolution: **100% (2048 x 1152)**. - Upscaler: **Game's own**, not DLSS. - Window scaling filter: **Catmull-Rom**; sharpness **30%**. - V-Sync: not visible in the supplied settings screenshot; not confirmed. The log separately records a **1920 x 1080 window swapchain** (lines 57-60); that is distinct from the game's output-resolution setting above. ## Reproduction / comparison 1. Load the Temple of Time as child Link and stand on the circular Triforce platform, facing the camera toward the sealed doorway, as in the screenshots. 2. With FG **Off**, note the built-in FPS counter. 3. Enable FG **Auto** and compare the built-in total FPS and `game` FPS at the same position and similar camera angle. The player supplied the two comparison screenshots and confirmed that the FG-on mode was **Auto**: | Capture | Native game FPS | Total output FPS | |---|---:|---:| | FG Off (`02-FG-off.png`) | 51 | 51 | | FG Auto (`03-FG-auto.png`) | 38 | 119 | These are individual screenshots, not averaged benchmark results. ## Log evidence around the FG activation Line numbers refer to the attached `hookshot_log-redacted.txt`; privacy redaction preserves the original line count. | Logged metric | Before activation | After activation | |---|---:|---:| | Renderer `presents` FPS | 54.1 (line 4403) | 37.4 (line 4498) | | Replay `submit` time | 10.41 ms (line 4401) | 18.16 ms (line 4496) | | GPU time per frame | 8.0 ms (line 4406) | 9.5 ms (line 4501) | At line 4418, FG starts with: ```text vk: frame interpolation ready: Auto, a frame for each of the display's refreshes (120 Hz; up to 4 a game frame) ``` The first full following `vi` interval reports **37.5 native FPS over 5 seconds** (line 4482). The interpolation log separately reports **595 frames shown on time, 0 late** (line 4481). Around the switch, VRAM usage is **5792 MB of a 6614 MB budget** (line 4442). The selected before/after samples report **0 shader/pipeline waits and 0 queued translation jobs** (lines 4400 and 4495). These observations do not establish the root cause or exclude every memory/synchronization issue. ## Expected result and investigation limits Some FG overhead is expected. I would appreciate investigation of whether this native-FPS loss can be reduced or whether the capture matches an already-known issue. This report does **not** claim that FG is free, that 120 Hz detection is wrong, or that any particular pending patch fixes this case. The player has also observed native FPS dropping to roughly 15-20 in other situations, and output tending toward 120 with fixed multipliers as well. Those additional observations are **not established by these screenshots or this Auto-mode log**, so they should not be treated as separately reproduced here. This is one player-recorded run, not a repeated Off/On/Off benchmark. The full log includes earlier play, configuration changes and some foreground-window changes; not every interval is a controlled comparison. Mods, overlays/injectors, exact Windows build, and whether Reflex stayed Off throughout the entire run have not been independently confirmed. No separate Reflex regression is claimed. ## Attachments and privacy - `01-settings.png`: supplied graphics-settings screenshot. - `02-FG-off.png`: 51 FPS comparison screenshot. - `03-FG-auto.png`: 119 total / 38 game FPS comparison screenshot. - `hookshot_log-redacted.txt`: full supplied log; the Windows account name in local paths and the save-directory identifier have been replaced with placeholders. Timing and rendering evidence is unchanged. No game files, generated executables, shader/program caches, saves or pipeline-state binary are included. The pipeline-list contribution is a separate topic. Prepared with AI assistance from the player's screenshots, log and hardware details; observations and unverified causes are kept separate. ## Sources [1] https://git.magicbean.dev/MagicBeanDev/Hookshot/issues/2510 — Hookshot frame generation tracking issue ## Evidence bundle [Download the report, three screenshots and redacted full log (ZIP)](https://github.com/user-attachments/files/34724418/Hookshot-FG-Fehlerbericht.zip)

NotRexed commented

2026-11-12 03:07:28 +00:00

(Migrated from github.com)

More reports of this, collected here:

#2516: Frame generation doesn't work (v2.1.0; RTX 4070, NVIDIA 621.14; Ryzen 5 7600x, 32 GB RAM, Windows 11 26H2)

More reports of this, collected here: - #2516: Frame generation doesn't work (v2.1.0; RTX 4070, NVIDIA 621.14; Ryzen 5 7600x, 32 GB RAM, Windows 11 26H2)