Skip to content

Releases: jpietek/PenguinBurner

PenguinBurner 0.5.2

22 Jun 21:35

Choose a tag to compare

PenguinBurner 0.5.2

Changes

  • Large internal cleanup and module split.
  • Auto-UV final-choice, CLI, and systemd flow fixes.

PenguinBurner 0.5.1

20 Jun 21:49

Choose a tag to compare

PenguinBurner 0.5.1

Changes

  • Expanded Auto-UV target tables for newer and Ampere RTX cards.
  • Added an Auto-UV power-limit control that follows the selected GPU's NVML range and is saved with scan options.
  • Tightened the Overlay tab Adaptive UV Target FPS control so it displays as an integer and no longer leaves extra space before the arrows.

PenguinBurner 0.5

20 Jun 21:01

Choose a tag to compare

PenguinBurner 0.5

Changes

  • Removed nvidia-smi shell calls in favor of direct NVML API helpers.
  • Added true headless Q2RTX tuned for PenguinBurner stability benchmarking.
  • Included multiple minor Auto-UV, UI, fan, profile, and cleanup fixes.

PenguinBurner 0.4.7

15 Jun 07:28

Choose a tag to compare

PenguinBurner 0.4.7

Fixes

  • Q2RTX verification crash on systems without OpenSSL 1.1 (e.g. Bazzite) —
    the bundled Q2RTX binary needs libssl.so.1.1 / libcrypto.so.1.1 but ships a
    non-relocatable RUNPATH (/mnt/q2rtx/., NVIDIA's build path), so on distros
    that don't provide OpenSSL 1.1 it died mid-run with error while loading shared libraries: libssl.so.1.1, most visibly during the final verification pass when
    launched through gamescope. PenguinBurner now stages OpenSSL 1.1 next to the
    binary and rewrites its RUNPATH to $ORIGIN, so the libraries resolve from the
    binary's own directory on every launch — independent of the system libraries
    and of how the process is wrapped (a privilege-dropped, gamescope-wrapped
    AT_SECURE launch strips LD_LIBRARY_PATH).
  • Transient launcher failures no longer fail a run or blacklist a voltage
    a Q2RTX launch that dies before producing stable metrics (a flaky headless
    gamescope startup, or a loader error) is now retried in place a few times, and
    an unrecovered launch is reported as an environment error instead of being
    recorded as an unsafe undervolt.

PenguinBurner 0.4.6

14 Jun 12:02

Choose a tag to compare

PenguinBurner 0.4.6

New

  • --dump-latency-data (advanced) — a daemon flag (equivalently
    PENGUIN_BURNER_DUMP_LATENCY_DATA=1) that dumps verbose latency internals to
    the daemon log: swapchain present mode and queue depth, plus Reflex
    sleep-mode (boost / FPS-cap) and recovery transitions. Off by default;
    for debugging display/VRR and frame-generation behaviour. Reflex sleep-mode
    lines are logged on change only, so the per-frame set calls do not flood the
    log.

Docs

  • New Advanced / hacky flags section in the README documenting the
    power-user CLI flags and latency-layer environment variables.
  • New docs/dev/penguin-burner-with-gamescope.md — how to run a game under
    gamescope with the PB latency layer, why the wrapper ordering matters
    (gamescope … -- PENGUIN_BURNER %command%), and the display-latency findings
    that motivate it.

Fixes

  • Update the packaging-metadata test to match the merged AUR build change (cmake
    is a system build dependency, not a Python build-system requirement).

PenguinBurner 0.4.5

14 Jun 11:14

Choose a tag to compare

PenguinBurner 0.4.5

Bug fixes

  • Overlay: fix phantom frame-generation detection. Titles without a Reflex
    marker stream (e.g. a simple game going from a 30fps menu to 120fps gameplay)
    no longer get the genuine framerate increase mis-read as 4x frame generation.
    The displayed/base "deinterlace" now only applies when a concurrent marker
    stream is present, so the overlay reports the true present rate instead of a
    phantom base + "FG" rate.

Packaging and metadata

  • Package the Fedora overlay entry points.
  • Fix the Ubuntu PPA debhelper build.
  • Restore GPL-or-later package metadata and fix the GPL license badge.
  • Mark the project as Beta (Development Status :: 4 - Beta).

Docs

  • Mark power-limit control as planned and list the Steam import features.

PenguinBurner v0.4.4

13 Jun 22:34

Choose a tag to compare

First tagged release of the 0.4 line (since v0.2.6). The experimental latency stack is dormant by default — a plain launch shows present-FPS with zero trace overhead; latency features require explicit opt-in.

Auto-UV & Auto-OC

  • Carry the Auto-OC clock offset through to the final/saved verification row; the Undervolting runs table now shows applied/limit MHz correctly, and non-OC rows read instead of a misleading 0/0.
  • Distinct efficiency / balanced / performance VF-curve tail shapes.

In-game overlay, latency & adaptive (opt-in)

  • New in-game Vulkan overlay (native present-path layer) showing present-FPS, clocks, voltage, and power.
  • Opt-in PC latency telemetry (--experimental-ingame-latency) and opt-in display (present→scanout) latency folded into the LAT number. All latency features default off.
  • Adaptive Auto-UV: switches between efficiency/balanced/performance profiles from achieved FPS vs an Adaptive UV Target; holds the startup tier when no FPS feed is present.

UI

  • Reordered tabs to Auto-UV / Profiles / Overlay.
  • Compacted the Undervolting runs table; renamed the Decision column to Result.
  • Renamed the overlay Target FPS control to Adaptive UV Target.
  • Default overlay update interval of 1s for new users.

Packaging & docs

  • Documented gamescope for headless/server Q2RTX stability runs; a lighter Wayland-free Xvfb path is planned.
  • Large README/feature-docs overhaul; repo restructured into focused packages (common/, runtime_support/, nvidia_driver/, native latency-layer split).

PenguinBurner 0.2.6

03 Jun 19:37

Choose a tag to compare

PenguinBurner 0.2.6 Release Notes

GitHub Release Notes

PenguinBurner 0.2.6 improves managed Q2RTX dependency installation on systems
where the OpenSSL 1.1 compatibility fallback has trouble reaching the primary
CentOS Stream package index.

Q2RTX Dependency Downloads

  • Q2RTX release metadata now falls back from the GitHub API to the official
    GitHub releases page before failing.
  • Q2RTX release downloads now use a mirror-aware retry path, keeping the
    official GitHub release URL as the current source.
  • OpenSSL 1.1 compatibility RPM lookup now retries multiple CentOS Stream
    AppStream package indexes.
  • OpenSSL compatibility RPM downloads now retry the resolved RPM across all
    configured CentOS Stream mirrors.
  • Download timeout and failure messages now include the URL that failed, making
    install reports easier to diagnose.

PyPI Release Summary

PenguinBurner 0.2.6 improves Q2RTX dependency installation reliability by adding
fallback lookup/download sources for the OpenSSL 1.1 compatibility RPM and
clearer URL-specific timeout diagnostics.

PenguinBurner 0.2.5

31 May 19:22

Choose a tag to compare

PenguinBurner 0.2.5 Release Notes

GitHub Release Notes

PenguinBurner 0.2.5 fixes LACT export of saved Auto-UV profiles.

LACT Export

  • LACT exports now respect LACT's Nvidia V/F offset behavior by clamping
    exported clocks to base_mhz + 1000 MHz by default and reporting a warning
    when a generated point is reduced.
  • --export-lact-config now honors --auto-uv-profile, so the CLI can export a
    selected saved Auto-UV profile by profile id, candidate id, JSON path,
    active, or latest.
  • --list-auto-uv-profiles is now visible in CLI help so users can list saved
    profiles before exporting one to LACT.
  • --lact-max-vf-offset-mhz allows overriding the default LACT Nvidia V/F
    offset ceiling for systems where the driver reports a different limit.

PyPI Release Summary

PenguinBurner 0.2.5 fixes LACT export for selected Auto-UV profiles and clamps
exported Nvidia V/F clocks to LACT's default offset limit before writing
gpu_vf_curve.

PenguinBurner 0.2.4

31 May 19:04

Choose a tag to compare

PenguinBurner 0.2.4 Release Notes

GitHub Release Notes

PenguinBurner 0.2.4 fixes Auto-UV startup on headless and SSH/X11-forwarded
NVIDIA systems where LACT can read live voltage but PenguinBurner rejected the
NVAPI voltage reader during preflight.

NVAPI GPU Selection

  • Hidden NVAPI voltage and V/F readers now match the selected GPU by PCI bus
    ID, following LACT's handle selection approach, instead of assuming NVAPI
    physical GPU order matches the nvidia-smi index.
  • This avoids reading voltage from the wrong physical GPU on multi-GPU systems
    and systems where NVAPI enumeration order differs from NVML.

Auto-UV Preflight

  • An implausible idle-time NVAPI voltage sample is now reported as a warning
    with the raw sampled value instead of stopping Auto-UV immediately.
  • Auto-UV continues into the workload phase and attempts to collect voltage
    telemetry again under load.

SSH/X11 Forwarding

  • The privileged GUI launcher now passes a default user ~/.Xauthority path
    when X11 forwarding sets DISPLAY but does not export XAUTHORITY.

PyPI Release Summary

PenguinBurner 0.2.4 fixes NVAPI GPU handle matching for live voltage reads,
allows Auto-UV to continue past implausible idle voltage samples, and improves
SSH/X11 forwarded GUI launches.