GhostDeck v1.43.0

What's new in 1.43.0 - the full write-up, with screenshots

[1.43.0] - 2026-10-30

Added

• Frame pacing in the game-session report (roadmap #6): under the frametime line the popup now shows the median
  frametime, the time 95 % of frames stayed under, the 0.1% low and a small histogram of where the frames sat (0 to 50+ ms,
  stutter-range bars in red). Status → Gaming shows the same numbers for a picked session, and the JSON/CSV export carries
  them with the histogram. Sessions saved before this simply show the popup as before.

• FPS snapshot hotkey (roadmap #35): a new action under Settings → Hotkeys (Ctrl+Shift+F12 by default). One press
  saves a small text file to the Desktop with the last 30 seconds of the game in front (FPS now, average, 1% low, median and
  95% frametime, stutters) plus CPU/GPU temperature, fans and the active profile, and an on-screen message names the file.
  If FPS was not being measured, the file still gets the hardware part and says how to turn FPS on.

• Low-FPS alert (roadmap #3): Settings → Gaming → Overlay, off by default, thresholds 30 / 45 / 60 / 90 / 120 FPS. When
  the game in front drops below it for 3 seconds after having run above it, a short on-screen message says e.g. "FPS dropped
  to 41 (your threshold 60) · CPU 97 °C". At most once a minute per game, works while FPS is measured, no extra hardware reads.

• GE66 Raider 10SF (1541EMS1, model database 20261118, #303, thanks @chopin312) - the backup WMI
  path layout is on record, so GhostDeck offers the path test on this laptop; the owner's Dragon Center capture matches the recipes.

• Katana 15 HX B14WGK (1587EMS1, #304, #305, thanks @suivan1 and @rudnet054) - the
  RTX 5070 variant joins the tested B14WEK / B14WFK entry. Credit also for @denizkat0 on the Vector 16 HX AI (#306).

• Two-table test on the Fan curve page (In action → diagnostics), next to the fan sweep (#224,
  #293). Every board keeps two fan-curve tables in its controller even with one fan in the chassis, and
  that fan listens to both: on a Cyborg 15 the factory values left in the second table kept the fan from dropping below about
  2700 rpm. The test writes six pairs of flat curves into both tables (25/25, 25/45, 45/25, 45/45, 0/45, 45/0 %, one table
  always keeping the fans turning), holds each pair for 20 s and says which table each fan follows and how the firmware combines
  the two. Mains only, refuses to start above 85 °C CPU / 75 °C GPU (a gaming laptop idles at 70-85 °C, so the gate sits above
  that and the run is kept short instead), stops itself at 92 °C, restores both tables and the fan mode; the run lands in the
  sweep's history marked 2×, with a report ready to paste.

• The diagnostics panel on the In action view remembers whether it was left open.

• "Write to both fan tables" - an experimental, opt-in switch on the Fan curve page for single-fan laptops
  (#293). The owner's two-table test measured that the firmware follows the higher of its two
  fan-curve tables, and that MSI Center writes its one slider into table 1 only, leaving the factory table 2 to hold the
  fan up; with the switch on, GhostDeck writes the one curve into both tables, so the fan follows it alone and can slow
  below the factory minimum. Off by default, a consent card explains what changes before it goes on, and switching it
  off puts the saved table 2 back.

• The two-table test has its own sub-tab on the Report page, the fourth card on the start screen (with what it answers
  and what it is usually for), built like the power test's: the explanation, the six pairs with what each one separates,
  a consent box, Start / Cancel / Start over, the progress, then the outcome, the findings and the table of pairs. The
  report goes to the clipboard and to a file on the Desktop, and Finish & open GitHub opens the new Two-table test
  result issue form with the model, the firmware and the verdict filled in. The run restores both tables and the fan
  mode from the controller's own values on its worker, so closing the window cannot leave a flat curve behind; a finished
  run also shows up in the Fan curve page's history picker.

• docs/TWO-FAN-TABLES.md - the whole story in plain language with pictures: the two tables, the measurement, the test
  and the switch; a matching FAQ entry. The FAQ's "Is there any risk" entry and TECHNICAL §19 now say in so many words
  which writes go beyond what MSI's own software does (the test and the switch, both opt-in) and what they touch.

• "Saved to: …" is a link in all four Report wizards (profiles, fan curve, power test, two-table test): a click opens
  the report file in whatever Windows uses for .txt files.

• Creator 16 AI Studio A1VHG (15F3EMS3, model database 20261117, #294, thanks @bengervais) - tested: the
  power test read every profile back intact with 0 % drift, and the owner hears Silent clearly quieter than Balanced under a
  graphics load; Extreme matches Balanced under a combined load, as on several other boards.

• Stealth 16 Studio A13VG (15F2EMS1, model database 20261117, #279, thanks @XereJay) - fan curve
  verified on both fans.

• Cyborg 15 A12VF / A13VF / A13VE (15K1IMS1, model database 20261117, #293, thanks @DeimosCreator) -
  fan RPM read as a 16-bit pair, so a slow fan no longer shows an impossible number; the owner's two-table test measured
  that the firmware follows the higher of its two fan-curve tables, and his MSI Center capture that the vendor's single slider
  writes table 1 only and leaves table 2 at its factory values.

• Modern 15 B11M (15H3IMS1, model database 20261116, #300, #301,
  #302, thanks @rock3tup) - recognised as an experimental model; fan curve verified on both tables and the
  fan RPM readout on (16-bit pair).

• Crosshair A16 HX (D7W/D8W) (15PLIMS1, model database 20261115, #297, #298, thanks
  @maestory) - a second owner confirmed the fan curve on both fans and ran a clean power test; added to the model's credit.

• Vector 17 HX A14V (17S2IMS2, model database 20261114, #296, thanks @Sombrilla1994) - another
  owner's capture matches the recipes in Balanced, Extreme and Super Battery; added to the model's credit.

• Prestige 16 AI Studio B1VGG (15A2EMS1, model database 20261113, #289, thanks @u9328009) - the RTX 4070
  variant joins the tested B1VFG entry; the owner's capture matches the standard recipes in all four profiles.

• Raider GE68HX 13V (15M2IMS1, model database 20261112, #286, #287,
  #288, thanks @ctimmons25) - tested: the power test measured Silent, Balanced and Extreme on the hardware
  (Silent at 80 % of Balanced's work, every profile read back intact), fan RPM readout switched on (16-bit pairs) and the fan
  curve verified on both fans.

• Creator 16 AI Studio A1VHG (15F3EMS3, model database 20261108, #284, thanks @bengervais) - recognised as an
  experimental model; the owner's capture matches the standard recipes in all four profiles, with a real Silent state.
  Fan curve verified on both fans (database 20261109, #285).

• Cyborg 15 AI A1VFK / A1VEK (15K2EMS1, model database 20261105, #281, #282,
  #283, thanks @Jur15) - tested: all four profiles confirmed on the hardware, fan RPM readout switched
  on (16-bit pair, CPU fan only) and the single fan curve verified. The A1VEK name joins the entry.

Changed

• Status → Gaming, saved games: the picked game now shows its frametime line and how even its frames were over the whole
  game, above its numbers (games saved before this version show the line only). The live part at the top is unchanged.
  The frame chart counts frames on its vertical axis, gives every time in frames per second too (16 ms / 60 fps), keeps a
  bubble at the tallest bar and one with the number of stutters, shows a clicked bar's exact count until it is clicked
  again, and has a "?" card that explains it. A switch, Real counts / All frames visible, squeezes the height so the
  rare slow frames show; the bubbles keep the real numbers. The legend names the game's own stutter line (twice a typical
  frame, at least 25 ms).

• Stutters are one number everywhere: the game-session popup, the session table, the export and the frame chart now
  count the same frames (those in the red bars). The count kept during the game followed the last few seconds and could
  differ by a few frames.

• Fan-curve report on a one-slider laptop names the table the curve was found in ("one test curve in the Fan 1 / CPU table
  @ 0x72") instead of naming it after whichever test values were typed into the single slider, and counts that as a match with
  the shipped map (#293).

• Raider GE78HX 13V / Vector 17 HX A14V (17S1IMS1, 17S2IMS2, model database 20261119): fan RPM is read as
  16-bit pairs (0xC8:0xC9 / 0xCA:0xCB). Measured on the reference GE78HX with a flat 25 % curve: below ~1870 rpm the
  single-byte read showed 6207 or 7468 rpm where HWiNFO64 showed 1435 / 1493, and the pairs read 01:4D / 01:40 = exactly
  those figures. MSI Center's factory curves never took the fans that low, which is why the old read matched for a year.

• Fan curve, "In action" view (#172) - readings whose value labels would overlap now form one group: a
  bigger ringed dot at the newest reading, one label with the fan, the temperature range, the duty and the count
  ("CPU 45-50°/0% (14)"), and the full list of readings in the info row under the chart while the group is hovered or pinned.
  A pile of idle readings in one spot or a row one degree apart along the curve used to paint their labels on top of each
  other. Labels say which fan they belong to and are 9 pt instead of 7.5.

• Models tab, "Your laptop" card (#172) - when several retail models share the firmware, the variant this
  machine is (the model name Windows reports) is drawn in the accent colour inside the entry name. The name itself stays
  whole: one name per firmware, the same in the app and the docs.

• Model database 20261107 - entry names: a slash separates variants of one model line (Raider A18 HX A9WIG / A9WJG),
  a comma separates different lines on the same board (Vector A18 HX A9WHG, Raider A18 HX A9WIG / A9WJG); 13 entries renamed.

• Model database 20261104 - credits: @XereJay on the Stealth 16 Studio A13VG (15F2EMS1, #279, Super
  Battery state confirmed), @HardFlakeCZ on the Katana 15 HX B14WEK / B14WFK (1587EMS1, #280, curve confirmed again).

• Cyborg 15 A12VF / A13VF / A13VE (15K1IMS1, model database 20261103, #224) - the chassis has one fan, so the
  curve editor shows a single curve and the second fan's RPM reading (always zero on this board) is gone.

Fixed

• Report page above 150 % display scale: the step rows of the Profiles and Power test wizards kept a fixed height, so
  their second line was clipped and the rows overlapped at 200 %; the "Addresses the test may write" label sat on top of the
  address list; and the Finish & open GitHub button swallowed its ampersand. Rows and labels now take their height from
  the fonts (#172).

• Keyboard backlight tile off on the Bravo 15 C7V / Katana A15 AI B8VG (158NIMS1, model database 20261114,
  #276) - on the owner's Katana the tile's levels did not change the backlight; MSI Center and the
  Fn key keep controlling it.

• Panic reset button kept its old caption after a language switch (#172) - the feature bricks on the
  Scenarios tab translate themselves on every paint, but the button inside one of them was captioned once at build time.
  The Scenarios tab now re-reads every such caption when the language changes.

• Turkish: the Extreme tile said "maks" (#172) - now "en yüksek · gürültülü".

Full Changelog: v1.42.0...v1.43.0