# ThreadPilot v1.7.8-beta.1 β€” Game Boost hardware-test preview **Pre-release, not stable.** This preview contains everything in stable **v1.7.7**, plus the complete local Game Boost implementation and Activity diagnostics through 2026-10-16. If you do not want experimental Game Boost, use [stable v1.7.7](https://github.com/PrimeBuild-pc/ThreadPilot/releases/tag/v1.7.7). The stable/latest release remains v1.7.7. This beta is not published to winget, Chocolatey or the stable automatic-update channel. Install the desired download manually when switching channels; stable and beta use the same application/settings location, not separate side-by-side installations. Export a configuration backup first. ## What's new over stable - **Game Boost (off by default):** topology-aware process CPU Sets for the game and explicitly listed background apps, optional Above Normal priority, respect for existing rules/manual changes, restoration on game exit or disabling Game Boost, and a crash-recovery journal. - **Activity diagnostics:** session start/end, what was applied or left unchanged and why, grouped background-process changes, and restoration results. - **CPU topology fixes:** Intel P-cores are identified by the highest Windows EfficiencyClass; AMD CCD groups use real L3 groups and sizes, including 6+6 layouts. - Full-path entries in the background-app list match correctly. ## What still needs real hardware testing **Game Boost is not yet validated on real dual-CCD Ryzen/X3D or supported Intel hybrid hardware. Comparative gaming benchmarks and real protected-game/anti-cheat testing are still pending. No FPS improvement is claimed.** Earlier QA covered a single-CCD Ryzen 7 5800X3D and a clean Windows 11 VM: UI/settings, session detection, optional priority, respecting rules/manual changes and crash recovery. Native CPU Set apply/read/restore tests used real Windows calls with a **simulated dual-L3 topology**; that does not prove behavior on a real multi-CCD CPU. Small CI runners do not exercise the native Game Boost split. EcoQoS in stable v1.7.7 was separately verified by @SA-Mousavichashmi on an i5-12500H. That result is **not** validation of Game Boost. We especially welcome reports from: - **AMD dual-CCD X3D:** 7900X3D / 7950X3D / 9900X3D / 9950X3D. Expect the game on the V-Cache CCD and listed background apps on the other CCD. Check for conflicts with AMD chipset-driver core parking. - **AMD equal-CCD CPUs:** e.g. 7950X / 9950X. Automatic leaves the game's placement to Windows and moves listed background apps off CCD0; Isolation also selects CCD0 for the game. Compare both modes. - **Intel hybrid:** 12th–14th gen / Core Ultra where supported. Expect game on P-cores, listed background apps on E-cores; check heavily multithreaded games in particular. On unsupported/single-CCD CPUs, incomplete topology or fewer than **six physical cores on the game side**, Game Boost leaves CPU Sets to Windows. Thus a laptop with fewer than six P-cores can correctly report no split. Optional priority remains separate. ## Quick test 1. Export your configuration. Record CPU, SMT on/off, Windows build (`winver`), chipset-driver version and active power plan; on AMD also note whether the `amd3dvcache` service is installed. 2. Open **Tweaks / Ottimizzazioni β†’ Game Boost**. Start in **Automatic**, with **Above Normal off**. Check the background-app list, add the game manually if it is outside a recognized library, enable Game Boost and bring the game to the foreground. 3. Check the topology decision and **Activity** events. Inspect actual game/background **CPU Sets** in ThreadPilot, System Informer or Process Lasso; Task Manager's classic affinity mask does not show soft CPU Sets. With SMT enabled, an 8-core V-Cache CCD normally means 16 game logical CPUs, a 6-core CCD means 12. On Intel compare the selection with P-core threads. 4. Alt+Tab: the session should remain active. Close the game or disable Game Boost: changes owned by Game Boost should be restored; existing rules and later manual changes should be respected. On equal-CCD AMD CPUs repeat in **Isolation**. Do not disable anti-cheat or bypass process protection. 5. For performance, use CapFrameX/PresentMon, the same scene/settings/power plan, and **at least three runs per scenario**: (a) off; (b) on without background load; (c) on with browser video / Steam download; (d) same as (c) with Above Normal. Compare average FPS, 1% lows, P95/P99 frame times and run-to-run variance. ## Report results in Issues Open an [issue](https://github.com/PrimeBuild-pc/ThreadPilot/issues) titled **[1.7.8-beta.1][Game Boost] CPU model β€” result/problem**. Successful tests help too. Send: - Exact app version, CPU, SMT state, Windows build, chipset driver, power plan and AMD `amd3dvcache` service status. - Game/version, mode, Above Normal setting, background load, and relevant saved rules or pre-existing CPU Sets. - Expected versus actual behavior, reproducible steps, whether disabling Game Boost restores normal behavior, and any stutter, crash, access/anti-cheat refusal or performance regression. - Screenshots of the topology decision and actual game/background CPU Sets, relevant Activity entries/export, and benchmark captures or a small comparison table when available. Redact usernames, personal executable paths and unrelated log data before uploading. ## Known limits and safety - Above Normal only raises a game starting at Normal, once per process; other starting priorities are left unchanged. - A manual selection exactly matching Game Boost's own CPU Sets cannot be distinguished from its selection and may be cleared at session end unless saved in a rule. - Game detection needs a recognized library plus foreground activity, or a manual entry. Add Battle.net/custom-install games when needed. - Applications can override QoS/CPU Sets; protected processes may refuse access. ThreadPilot does not bypass protection. - Artifacts are **unsigned** while code-signing enrollment is pending. SmartScreen may warn; verify SHA256. Upgrades preserve settings; uninstalling removes ThreadPilot-owned user data.