30 to 60 FPS: Claude Code Found About 40 Fixes in Oblivion Remastered
- Level Advanced
- Setup1–3 hrs (estimated)
- Evidence Some evidence
- CostNot stated in source
- No-codeNo
What they did
A player whose Oblivion Remastered stuttered at 20-30 FPS in cities asked Claude why. It found the game rebuilding a scene's lighting every frame on a single CPU thread. Over two weeks, Claude Code built benchmarks it could run while he was at work and found about 40 improvements: cities went from about 30 to a steady 60 FPS and open-world FPS rose 30-40%.
How it went
A shareable performance mod plus a benchmark log showing which fixes helped.
Worth knowing
- Took two weeks and a bigger Claude plan. It's a beta mod tested mainly on AMD hardware; one commenter's AI code review found minor concurrency bugs and said the FPS numbers include frame generation. Back up the game and check the game's mod rules first.
Prompt
Cleaned version rewritten from the source — not a direct quote.
My PC game [game] runs badly ([FPS] in [place]) on [CPU / GPU / RAM]. Help me find out why. First build a repeatable benchmark for one bad scene that you can run on its own and measure. Then profile it, list the biggest wastes (like work redone every frame or everything stuck on one CPU thread), and try one fix at a time, keeping only changes that beat the benchmark. Log every change with before and after numbers. Don't touch my save files, and ask before changing anything outside a separate mod folder.
Recommended
Try it in Grok Bot
- Copy the prompt (button below).
- Open Grok Bot chat and paste it. Replace any [bracketed] placeholders.
- Connect the listed tools (as needed) as connectors when Grok Bot asks.
- If the task repeats, ask: "Make this a daily routine."
Alternatives: the same prompt works in Claude Code, Claude, ChatGPT Agent, OpenClaw.
Set it up
Stack: Claude Code (Opus 5.5 + sub-agents)
- Read the original source for full context.
- Connect the tools your agent needs.
- Paste the prompt and replace [bracketed] placeholders.
- Run once manually and check the output before scheduling it.
Generic suggested steps, not from the source.
Evidence
Detailed first-person Reddit post with method (benchmark harness, /goal, 2 sub-agents) and results; code shared on GitHub; an independent commenter's AI review called it real but rough
Source: Reddit · 2026-10-17 · Verified by us: not yet · Report broken or outdated