NullMoth 2.6.3 ships our fixes. v2.6.3 (2026-10-28) credits "watchthelight's NvGfx" for the window drag stutter (#251, #340, #412), the Dock and Control Center blur (#363), the HiDPI "looks like" sizes (#354), Chromium scrolling and faster System Settings panes. The commits behind our open PRs are on main with their original author: c21ea8b (#421), 617adbc (#422), a7aad72 (#420), and the Intel buffer-texture fix 997c6ee. PRs #420 to #422 are still open. The fork should move onto 2.6.3; upstream dropped the fork's marker comments (4c04b7d), so the rebase will conflict on comments. High confidence.
NMIntelFB brightness (41e7e68): panel
brightness through the PCH PWM, with WhateverXe's licence added. NMIntelFB still publishes no VRAM,totalMB, so it
shows the 15 MB framebuffer as VRAM; we fixed that in XeGfxIntelFB 0.6 and have an upstream-ready patch. Medium
confidence.
#424 (issue): frame submission not locked to vblank at
165 Hz on an RTX 4060 Laptop, a drop about every 2 s; the reporter suggests nvhwvbl=1 but could not test it. Our panel
is also 165 Hz. Worth a test on the 5070 before answering. Medium confidence.
#426 (issue): an RTX 5050 Laptop (GB207) user on 26.7.2 hits a translator bail-out on a sampler array of 2 in three.js. Not tested here yet. Medium confidence.
2.6.3 conf precedence: a setting in the user's own block of nvrm610.conf now wins over the shipped value (release notes, with #368 and #399). Our I-085 (knobs lost after guard rollbacks) came from the other order, so this may let us drop the workaround once the fork is on 2.6.3. Medium confidence until tested.
Translator fixes and merged PRs from other contributors with no link to our problems, a small-BAR change, the 1401 installer housekeeping, steelbrain/reims-vgpu docs, and MegaMind2566's IdeaPad S540 (NVIDIA disabled).