🔬 Experimental Ada voltage-cap fix. Ada testers wanted.
🎮 Changed
Ada voltage-cap — third iteration. The previous v0.1.2 (hf57) only delivered boost-cap on Ada: frequencies were capped but voltage stayed near stock. v0.1.3 reproduces the wire pattern Afterburner writes to the GPU, decoded from a 4070 Ti probe-apply snapshot: uniform -502 MHz on every VF point plus a single positive spike at sorted lockIdx+1. The spike is what engages the voltage-cap on Ada.
Quick history of the Ada path:
v0.1.1 — hardlocked Ada within 1.3s of game launch (bad wire-pattern collapse)
v0.1.2 — stable, no crashes, but only boost-cap
v0.1.3 — voltage-cap attempt based on probe-decoded AB pattern
Open question: whether the voltage-cap actually engages on real hardware. The probe-decoded pattern is necessary but possibly not sufficient — AB does additional work in its backend we cannot fully observe from a VF-curve probe. More investigation needed if voltage stays at stock after applying.
🧪 Test request
Ada owners (4060 through 4090, including SUPER refresh): apply a tier profile, monitor GPU-Z or HWiNFO under load.
GPU voltage under load should sit near the lock voltage (e.g. Balanced 925 mV → ~0.925 V). If it stays at stock (~1.05-1.10 V), the fix isn't engaging.
PerfCap reason — "VRel" or "Voltage" = good. "Pwr" only = falling back. "Util" = no cap.
Probe Inspector dump if you have the tool, even better.
Reports either way help.
📖 Compat
Existing settings, stabilizer_adjustments.json, and Custom Profiles from v0.1.0-v0.1.2 are preserved.
⬇️ Download
NV-UV-Play-v0.1.3-alpha.zip (portable, ~74 MB). Hash and VirusTotal report in the asset block.
🙏 Credits
Thanks to the @Gothico74
who reported through the v0.1.0 hardlock and stayed engaged through v0.1.2.
The yesterday's side-by-side probes (Stock / Afterburner / Play on a 4070 Ti) made the wire-pattern decode possible.