Speed: Measures GPU compute throughput and memory bandwidth. For overnight stability testing, go to 5. Endurance.
` + base
+}
+
+// renderEndurance renders the Endurance page (step 5): long-duration reliability tests.
+// Focuses on Stability and Overnight profiles for multi-hour burn validation.
+// For short load tests, see Load (step 3). For throughput measurement, see Speed (step 4).
+func renderEndurance(opts HandlerOptions) string {
+ base := renderBenchmark(opts)
+ return `
Endurance: Long-duration reliability tests — Stability (several hours) and Overnight (8+ h) profiles. These profiles run hardware at sustained load; results show whether the server holds its performance envelope over time.
+
Use the Stability or Overnight profile in the setup card below. The Standard profile is available too but is better suited for the 4. Speed page.
` + base
+}
diff --git a/audit/internal/webui/page_burn.go b/audit/internal/webui/page_burn.go
index 4d42b9e..f6de5cc 100644
--- a/audit/internal/webui/page_burn.go
+++ b/audit/internal/webui/page_burn.go
@@ -1,8 +1,13 @@
package webui
+// renderLoad renders the Load page (step 3): sustained stress tests.
+// For non-destructive status checks, see Check (step 2).
+// For DCGM targeted diagnostics (targeted_stress, targeted_power, pulse), see Check → Validate mode.
+func renderLoad() string { return renderBurn() }
+
func renderBurn() string {
return `
⚠ Warning: Stress tests on this page run hardware at high load. Repeated or prolonged use may reduce hardware lifespan. Use only when necessary.
-
Scope: Burn exposes sustained GPU compute load recipes. DCGM diagnostics (` + "targeted_stress, targeted_power, pulse_test" + `) and LINPACK remain in Validate → Stress mode; NCCL and NVBandwidth are available directly from Validate.
+
Scope: Load runs sustained GPU compute and CPU/memory stress recipes. DCGM diagnostics (targeted_stress, targeted_power, pulse_test) and NCCL/NVBandwidth are on the 2. Check page. For overnight endurance runs, see 5. Endurance.
Tasks continue in the background — view progress in Tasks.
diff --git a/audit/internal/webui/page_settings.go b/audit/internal/webui/page_settings.go
new file mode 100644
index 0000000..dba08c7
--- /dev/null
+++ b/audit/internal/webui/page_settings.go
@@ -0,0 +1,77 @@
+package webui
+
+import "html"
+
+func renderSettings(opts HandlerOptions) string {
+ version := opts.BuildLabel
+ if version == "" {
+ version = "dev"
+ }
+ return `
+
+
+
Blackbox Logging
+
+
Continuous hardware monitoring that writes a rolling log of sensor readings to the export directory. Useful for capturing thermal or power anomalies during long runs.
+
+
+
+ Loading...
+
+
+
+
+
+
NVIDIA Recovery
+
+
Reset NVIDIA GPU driver state. Use when nvidia-smi reports errors or GPUs appear stuck after a failed test.
Non-destructive: Check tests collect diagnostics only — no writes to disks, no sustained load, no hardware wear counters incremented. For stress testing, go to 3. Load.
+
+
+
+ est. ` + validateTotalStr + gpuNote + `
+
+
+
+` + renderSATCard("cpu", "CPU", "runSAT('cpu')", "", renderValidateCardBody(
+ inv.CPU,
+ `Collects CPU inventory and temperatures, then runs a bounded CPU stress pass.`,
+ `lscpu, sensors, stress-ng`,
+ validateFmtDur(platform.SATEstimatedCPUValidateSec)+` (stress-ng 60 s).`,
+ )) +
+ renderSATCard("memory", "Memory", "runSAT('memory')", "", renderValidateCardBody(
+ inv.Memory,
+ `Runs a RAM validation pass and records memory state around the test.`,
+ `free, memtester`,
+ validateFmtDur(platform.SATEstimatedMemoryValidateSec)+` (256 MB × 1 pass).`,
+ )) +
+ renderSATCard("storage", "Storage", "runSAT('storage')", "", renderValidateCardBody(
+ inv.Storage,
+ `Scans all storage devices and runs the matching health or self-test path for each.`,
+ `lsblk; NVMe: nvme; SATA/SAS: smartctl`,
+ `Seconds (NVMe: instant device query; SATA/SAS: short self-test).`,
+ )) +
+ `
+
+
+
NVIDIA GPU Selection
+
+
` + inv.NVIDIA + `
+
+
+
+
+
+
Loading NVIDIA GPUs...
+
+
Select at least one NVIDIA GPU to enable NVIDIA check tasks.