github Kizzuwatnaa/DLSS5-Autopilot v2.0.1
DLSS 5 Autopilot v2.0.1

7 hours ago

DLSS 5 Autopilot 2.0.1

What people hit in 2.0.0's first day, the removals that were missing, and
one game that could not be installed at all.

Updating from 2.0.0 with update now downloads the new build, checks its
SHA-256 against the SHA256SUMS.txt published with the release, and offers
a one-click restart. Your settings, library, hidden games and profiles carry
over.

New

  • A vr page. A VR game on OpenXR draws the headset's picture through the
    OpenXR runtime; a proxy DLL or the Vulkan layer only reaches the mirror on
    your desktop, so the headset sees nothing. ReShade's OpenXR layer is the
    one that reaches the eyes, and the new page is where that lives: whether a
    layer is registered for your user and whether it is ours, which games in
    your library carry openxr_loader.dll (the only evidence read from disk
    that a game goes through OpenXR at all), and the button that takes our
    registration out again. Installing it is still part of a game's install -
    the VR headset (OpenXR layer) box in its settings. Nobody here has a
    headset: the page says exactly that, and asks for the report that would
    settle it. The detail
  • What other people found, in numbers. The block under a game's actions
    now says how many shared results there are and how many worked, how it
    went on the driver this PC has, and - where three people measured it - the
    frame rate and work area behind that route. With one line that the counts
    are counts: a route that worked for somebody else can still fail here, and
    one that failed for them can work.

Fixed

  • A question that never drew, and could not be closed (#263). Press
    autopilot, or more on the driver line, and the screen went dim with
    nothing on it: the card the question is drawn on was made while the button
    press was still running, so Windows mapped it and never painted it. Dragging
    the window brought it up, because moving the window repaints it. It is
    painted before the press returns now.
  • Closing the window answers the question. Quitting from the tray, or
    Windows shutting down, while a question was up left the program running
    with no window - and every later launch then said it was already open.
  • The same card could sit behind the window (#263). The card that opens on more,
    and every question the tool asks, is its own borderless window. Nothing
    held it in front of the main window, so anything that took the focus left
    it behind - with its click-catcher still on, swallowing the clicks aimed
    at it. Clicking the dim area around it was meant to close it and never
    did: the catcher sat on the card, so Tk dropped those clicks before any
    handler saw them. Now the dim and the card stay in front of the window
    they belong to, the catcher covers both, and ok, Esc (from the
    card or from the main window) and a click on the dim all close it. Coming
    back from the taskbar brings the card back in front, with the keyboard.
  • Uninstall was hard to find (#259). In 2.0.0 the only way out of an
    install was a link inside the game page's settings. It is on the
    game's page again, in the row with 'what will happen?' and 'open folder',
    and it is in a game card's right-click menu in the library, so an install
    can be taken out without opening the page. It still asks first, and the
    game's own files still come back.
  • The row of links under a game's details wraps instead of running off the
    page when there is no room for all of them.
  • A Remix mod can be taken out again. The page installs a mod into a
    game; taking it out was written but never had a control on it. An installed
    mod now has remove beside it: the files this tool wrote go, anything
    they replaced comes back, and a DLSS 5 install in the same folder is left
    alone.
  • The uninstall link inside a game's settings is gone: it is the one
    on the page now, not two of the same link a few rows apart.
  • The row of buttons on a game's page wraps instead of running off the right
    edge: at 250% scaling the last of them - uninstall - hung off the side with
    no way to scroll to it.
  • Our own files are never read as the game's own DLSS. A removed or
    half-finished install leaves the nvngx_* runtimes behind and takes its
    record with it; a 2015 game with no DLSS at all was then treated as one
    that ships it, which is a different route entirely.
  • The autopilot tries the routes in the order other people's results put
    them in
    , and the log says what put each one there ("9 of 12 in this
    game"). A route that has never worked in the game goes below one nobody
    has tried, and two results out of two do not outrank nine out of twelve.
    What the tool recommends first is still what the rules say - the shared
    numbers order what comes after it.
  • A game that needs DXVK could not be installed at all. Metal Gear Solid
    V closes itself the moment ReShade hooks Direct3D 11, so it reaches DLSS 5
    only through DXVK, with ReShade as a Vulkan layer. Setting the game's API
    by hand - to Vulkan, which is what that route produces, or to anything else
    – turned DXVK off without a word: the box was not even drawn, and the
    install wrote the ReShade proxy that makes the game quit. For the games
    known to need DXVK the decision now follows the detected API, whatever has
    been set by hand, and the box is drawn and ticked. On drivers 616.64 and
    newer the same game was also being steered to the standalone route, which
    does not go through DXVK - so that steering now leaves those games alone.
  • Nothing the tool's own checks write reaches your autopilot.log any more:
    a verification run used to leave its deliberate errors in the file the
    diagnosis reads and that you are asked to send with a report.

Don't miss a new DLSS5-Autopilot release

NewReleases is sending notifications on new releases.