Repository navigation
feat: the clock profiles measure stock for five minutes too - #23
Merged
Merged
Conversation
…hutdown NVML counts the calls that are inside it. A call that an access violation throws out (and our guard catches) is never counted out again, and nvmlShutdown then waits for it for ever while it holds the lock nvmlInit needs. On 2026-10-09 that froze the window during an All in one run: the window's thread sat in nvmlShutdown, the run's thread in nvmlInit. After such a fault NVML is no longer shut down or unloaded in that process (Nvml::Abandon, set by guarded_driver around every NVML and NVAPI call), and the tray app restarts itself once no run is under way and no result is on the screen. The new copy keeps the tune applied if the old one did and starts from the driver resets the old one had counted, so the watchdog still gives up after four in an hour. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Check 80 records the freeze of 2026-10-09, what the dump showed, the stand-alone reproduction and the unelevated run of the fix. Check 81 is the elevated half, not run yet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…re written to boot.log A caught access violation left no trace: it took a dump of a frozen process to learn that one had happened. The fault is now written down with where it happened (the faulting instruction and its callers, module+offset), and a reconnect that gave up says what its last attempt failed on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The four clock profiles compared a five-minute soak with a 30 s baseline: a card that was not heated through, so the result looked warmer and louder than stock. Their run now starts with the same five-minute stock measurement as All in one and Undervolt, fans left to the driver, and the result table shows that as the "before". The search keeps its own 30 s baseline as the reference for its thresholds. All in one hands its stock measurement to the overclock step, so nothing is measured twice. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…clock profiles Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Oct 10, 2026
Rovey
marked this pull request as ready for review
October 11, 2026 08:01
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Builds on #22 and contains its commits; merge that one first, or merge this one for both.
What this changes
The four clock profiles compared a five-minute soak with a 30 s baseline. A card is not heated through after 30 s, so the result looked warmer and louder than stock (61 °C and 52 % fan "before" against 70 °C and 63 % "after" in the run that prompted this).
measure_stock, the fans left to the NVIDIA driver), and the result table andgao --optimizeshow that as the "before".A clock profile's run takes about five minutes longer.
Verification
cmake --workflow --preset ci:core_testsandapp_testspassNot verified
docs/hardware-checks.mdon the branch of fix: no install at start, and the logon task without schtasks.exe #24🤖 Generated with Claude Code