Set up the hardware service before the Auto-UV dialog opens

The setup dialog reads GPU identity and limits through the daemon, so a
fresh or upgraded install (0.6.x Python-daemon era, stale AUR daemon)
opened it with a misleading generic "NVIDIA GPU" and no limits — the
exact state that scared users away from the one prompt that would fix
it. The install/update/migrate prompt now runs when Setup Auto Undervolt
is clicked, before the dialog; the remaining unavailable-info fallback
text names the service and the fix instead of "NVML read-only info
unavailable".

README: drop the GIF footnote and add the stock RTX 5080 baseline
(2734 MHz at 341 W) next to the tier results for scale.

Live-verified on real hardware: daemon stopped -> Setup click shows the
service prompt first and no dialog; accepting reinstalls the service and
the dialog then lists the real GPU and limits. Focused Qt suites passed
on this change; the full-suite run is deferred to an idle host (known
environmental Qt deadlock, aggravated by a live Auto-UV scan).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: jpietek

README.md

@@ -25,6 +25,10 @@
 - **Balanced** — undervolt and maintain clock. RTX 5080: 850 mV · 2639 MHz · 267 W.
 - **Performance** — undervolt and overclock. RTX 5080: 915 mV · 2980 MHz · 310 W.
 
+For scale: the same RTX 5080 at stock runs 2734 MHz at **341 W** under the
+same load. The factory curve burns 74 W more than Balanced for 4% clock —
+that is the inefficiency Auto-UV removes.
+
 Verified RTX 5080 examples; every GPU differs. Pre-optimized targets are
 included for RTX 30, 40, and 50 series cards.
 
@@ -33,9 +37,6 @@ automatic undervolting & overclocking. Adaptive per-game tuning targets a
 chosen pre-frame-generation FPS; the optional overlay shows PC latency and live
 FPS.
 
-The GIF uses PenguinBurner's real Qt interface with simulated scan data; live
-Auto-UV uses the same UI while testing your GPU.
-
 ## Install
 
 ```bash

tests/test_ui_window_actions.py

@@ -60,6 +60,11 @@ def win(qapp, monkeypatch):
 
 def test_start_scan_cancelled(win) -> None:
     window, monkeypatch = win
+    monkeypatch.setattr(
+        window_mod,
+        "ensure_daemon_ready_for_privileged_action",
+        lambda **_kwargs: True,
+    )
     monkeypatch.setattr(window_mod, "select_scan_tuning", lambda **k: None)
     fake = _FakeController()
     window.scan_controller = fake
@@ -107,29 +112,36 @@ def test_full_scan_shows_tier_progress_but_selected_profile_scan_hides_it(win) -
     assert window.auto_uv_tier_progress.widget.isHidden()
 
 
-def test_start_scan_runs_daemon_migration_gate_before_scan(win) -> None:
+def test_start_scan_runs_daemon_gate_before_the_setup_dialog(win) -> None:
+    # The setup dialog reads GPU identity/limits through the daemon: the
+    # install/update prompt must fire BEFORE the dialog opens, or a fresh
+    # install shows a misleading generic GPU with no limits.
     window, monkeypatch = win
-    gate_calls = []
-    monkeypatch.setattr(window_mod, "select_scan_tuning", lambda **k: {"gpu_index": 0})
-    monkeypatch.setattr(window_mod, "persist_runtime_gpu_index", lambda idx: int(idx))
-    monkeypatch.setattr(window_mod, "scan_command", lambda options: ["echo", "scan"])
+    order = []
 
     def fake_gate(**kwargs):
-        gate_calls.append(kwargs)
+        order.append(("gate", kwargs["action_label"]))
         return True
 
+    def fake_dialog(**_kwargs):
+        order.append(("dialog", None))
+        return {"gpu_index": 0}
+
     monkeypatch.setattr(
         window_mod,
         "ensure_daemon_ready_for_privileged_action",
         fake_gate,
     )
+    monkeypatch.setattr(window_mod, "select_scan_tuning", fake_dialog)
+    monkeypatch.setattr(window_mod, "persist_runtime_gpu_index", lambda idx: int(idx))
+    monkeypatch.setattr(window_mod, "scan_command", lambda options: ["echo", "scan"])
     fake = _FakeController()
     window.scan_controller = fake
 
     window.start_scan()
 
-    assert gate_calls
-    assert gate_calls[0]["action_label"] == "Starting Auto-UV"
+    assert order[0] == ("gate", "Setting up Auto-UV")
+    assert order[1] == ("dialog", None)
     assert fake.started
 
 

ui/features/tuning/tuning.py

@@ -430,9 +430,16 @@ def _power_limit_set_probe_applicable(power: object | None) -> bool:
     )
 
 
+_NVML_INFO_UNAVAILABLE_TEXT = (
+    "GPU limits unavailable: the PenguinBurner background hardware service "
+    "is not responding. Close this dialog and click Setup Auto Undervolt "
+    "again to install or repair it (one admin prompt)."
+)
+
+
 def auto_uv_nvml_info_text(info: AutoUvNvmlInfo | None) -> str:
     if info is None:
-        return "NVML read-only info unavailable"
+        return _NVML_INFO_UNAVAILABLE_TEXT
 
     rows = [
         _power_limit_text(info),
@@ -443,7 +450,7 @@ def auto_uv_nvml_info_text(info: AutoUvNvmlInfo | None) -> str:
         _power_limit_set_text(info.power_limit_set_supported),
     ]
     text = "\n".join(row for row in rows if row)
-    return text or "NVML read-only info unavailable"
+    return text or _NVML_INFO_UNAVAILABLE_TEXT
 
 
 def power_limit_set_supported(gpu_index: int) -> bool:

ui/window.py

@@ -367,6 +367,21 @@ def show_about(self) -> None:
     def start_scan(self) -> None:
         if self._workflow_running():
             return
+        # The setup dialog reads GPU identity and limits through the daemon,
+        # so the service must be installed/updated BEFORE the dialog opens —
+        # otherwise a fresh or upgraded install shows a misleading generic
+        # "NVIDIA GPU" with no limits (the exact state that scares users off
+        # the one action that would fix it).
+        if not ensure_daemon_ready_for_privileged_action(
+            QtWidgets=self.QtWidgets,
+            parent=self.window,
+            log=self.log_view.append,
+            action_label="Setting up Auto-UV",
+            # The scan streams through the daemon; a stale daemon without this
+            # capability must land in the repair prompt, not fail at scan start.
+            required_capabilities=("scan-stream-v1",),
+        ):
+            return
         options = select_scan_tuning(
             QtCore=self.QtCore,
             QtGui=self.QtGui,
@@ -384,16 +399,6 @@ def start_scan(self) -> None:
                 f"Could not save selected GPU index: {exc}",
             )
             return
-        if not ensure_daemon_ready_for_privileged_action(
-            QtWidgets=self.QtWidgets,
-            parent=self.window,
-            log=self.log_view.append,
-            action_label="Starting Auto-UV",
-            # The scan streams through the daemon; a stale daemon without this
-            # capability must land in the repair prompt, not fail at scan start.
-            required_capabilities=("scan-stream-v1",),
-        ):
-            return
         options = {**options, "gpu_index": int(self.gpu_index)}
         # Bring the scan into view: the live runs/curve are on the Auto-UV tab.
         self.tabs.setCurrentIndex(self.auto_uv_tab_index)