● DirectX Crash Forensics Available: GPU crash analysis (TDR events, VRAM exhaustion, present failures, frame drops) requires --focus-app "application.exe" flag. Run with forensic focus mode to enable DirectX diagnostics (adds 15-30% capture overhead).
● PCIe Slot Configuration
Current Configuration
PCIe 4.0 (16.0 GT/s)
x8 lanes
GPU Maximum
PCIe 4.0 (16.0 GT/s)
x16 lanes
Configuration Issues Detected:
● REDUCED LANE WIDTH: GPU capable of x16 running at x8 (50% bandwidth loss)
Likely cause: PCIe lane bifurcation (M.2 drives sharing lanes with GPU slot) or installed in secondary x8 slot
Common culprit: M.2 expansion cards in lower PCIe slots can force GPU lane reduction AND speed downclocking
Check BIOS: Look for 'PCIe Bifurcation', 'M.2 Mode', or 'PCIe Lane Distribution' settings
Note: PCIe 4.0 (16.0 GT/s) x8 = ~8 GB/s — still faster than PCIe 3.0 x16 (15.75 GB/s)
● Recommended Actions:
- CHECK M.2 CONFIGURATION: Many motherboards share PCIe lanes between GPU and M.2 slots
- M.2 EXPANSION CARDS: If using M.2 adapter cards in lower PCIe slots, they can cause lane contention AND force entire PCIe bus to downclock
- SOLUTION: Remove M.2 expansion cards from lower PCIe slots, use only motherboard-integrated M.2 slots when possible
- In BIOS, check if M.2_1 or M.2_2 slot is stealing 8 lanes from GPU (look for PCIe Bifurcation settings)
- If GPU is in secondary slot, move to primary PCIe x16 slot (topmost, closest to CPU)
- Performance impact: Minimal for most games at 1080p/1440p, ~5-10% at 4K with high VRAM usage
● FPS & Frame Time Timeline — GPU Performance Over Time
Performance Telemetry Overview
Real-time GPU metrics sampled directly from your graphics card and driver telemetry stack:
FPS — DirectX Present timing
GPU Load — Utilization %
Temperature — Thermal state
VRAM — Memory usage
Power — Draw in watts
Clock — GPU frequency
● Use this to spot: Thermal throttling (temp spikes), GPU bottlenecks (100% util), VRAM exhaustion, frame drops
Stable Frame Pacing: Consistent FPS (σ=3.3, avg: 58). Frame delivery looks smooth overall.
● Hover for detailed diagnostics | Click legend to toggle series | Drag to zoom, double-click to reset
● CRITICAL: 5.5%
11 samples >90% VRAM
● WARNING: 94.5%
189 samples 75-90% VRAM
🔎 DirectX Graphics Activity — Forensic App Focus Enabled
Forensic Mode Active: DirectX Graphics Kernel (DxgKrnl) events captured for frame-level analysis. Shows Present() calls, frame timing, DWM composition, and per-process GPU usage. Enabled when using --focus-app with benchmark/game.
✅ Gamer Quick Triage (Do These First)
1) FPS shows 0: Present() events were not captured (fix by running as Admin + --focus-app; update to latest build).
2) Stutter / spikes: Close overlays + GPU background apps (Discord/Chrome HW accel/OBS/NVIDIA Overlay).
3) VRAM > 90%: Lower textures first (biggest VRAM win), then resolution; disable RT if enabled.
4) GPU temp > 85°C: Clean dust, improve airflow, set a more aggressive fan curve (throttling = FPS drops).
5) TDR / Present failures: Revert GPU overclocks, verify PSU + PCIe power cables. Advanced / risky: Clean-install driver with DDU only if standard reinstall fails.
Per-Process GPU Usage
VRAM Usage Over Time
375 samples collected every 2 seconds
9,609 MB avg
Min: 9,455 MB | Max: 9,883 MB
Peak: 80.4% of 12,288 MB
| Process | Present Calls | Context Switches | GPU Share % |
| CrimsonDesert.exe | 2,632 | 1,353 | 73.8% |
| dwm.exe | 452 | 74 | 12.7% |
| csrss.exe | 109 | 6 | 3.1% |
| System | 99 | 1,459 | 2.8% |
| AWCCOverlay.exe | 55 | 122 | 1.5% |
| explorer.exe | 49 | 54 | 1.4% |
| WmiPrvSE.exe | 39 | 269 | 1.1% |
| msedge.exe | 18 | 115 | 0.5% |
| Taskmgr.exe | 15 | 63 | 0.4% |
| MsMpEng.exe | 13 | 57 | 0.4% |
● Understanding This Table:
• Present Calls: Frame submissions to GPU - higher = more rendering activity from this process
• Context Switches: GPU scheduling events - high values indicate GPU multitasking between apps
• GPU Share %: This process's percentage of total frame submissions - shows which app is using the GPU most
📊 DirectX Performance Analysis & Interpretation
Stable Frame Delivery: 3,667 frames presented with 0.00% failure rate. GPU rendering pipeline operating normally.
High GPU Contention: 5,077 context switches (17/sec) - multiple apps competing for GPU. Background apps (Chrome, Discord, OBS) degrading performance.
Perfect VSync Alignment: 1,138 refresh events with 0.00% miss rate - frame delivery synchronized to 60Hz.
High VRAM Usage: Peaked at 9,883 MB (80.4%), averaged 9,609 MB. Approaching capacity limit - reduce texture quality or close background apps.
ℹ️ What These Metrics Mean & How To Improve Them:
• Present() Calls: Frames submitted to GPU. More = higher FPS. Low? Check CPU bottleneck or game settings.
• Context Switches: GPU switching between apps. High (>1000)? Close background apps like Chrome, Discord, OBS.
• DMA Buffers: Command queue latency. Slow (>16ms)? Update GPU drivers, check VRAM usage.
• VSync: Monitor refresh sync. Misses? GPU cannot maintain refresh rate - lower graphics settings or enable G-SYNC/FreeSync.
• TDR: GPU driver crash. Any TDR events? Update drivers immediately, check for overheating, remove GPU overclock.
• DWM: Windows desktop compositor. Frame drops? Disable visual effects in Settings → System → Display → Graphics.
🧪 Data Source: GPU metrics captured via Microsoft-Windows-DxgKrnl ETW provider. For focused app analysis with FPS timeline, use: --focus-app YourApp.exe. System-wide VRAM tracking requires nvidia-smi in PATH (included with NVIDIA drivers).
🩺 GPU Health & Performance Analysis
Comprehensive diagnostics covering thermal state, VRAM allocation, power delivery, driver stability, and performance optimization.
⚠️ WARNING: GPU temperature at 87°C - thermal throttling imminent (clocks reducing to maintain safe temp)
🛠️ Recommended Actions:
- Thermal Analysis: Current temp 87°C approaching throttle threshold (typically 83-87°C for modern GPUs)
- Maintenance: Clean GPU fans/heatsink with compressed air, verify case has adequate ventilation (2+ intake fans)
- Monitoring: Install HWiNFO64 to log 'GPU Temperature' and 'GPU Clock' during gaming - watch for clock speed drops indicating throttling
Results are workload-dependent and not guaranteed. Validate each change with a fresh capture and keep rollback options (restore point/driver rollback) available.