100% HTML -- To test your monitor, download this html file, open it, click the Start Test button. That's it. What was meant to be a 1-minute chat with Claude AI about testing my monitor, turned into a 2-hour collaboritve pairing session -- which was great, and as good a pairing with a human!.
Read below to see what Claude had to say (Claude wrote the rest of this README file) -- Paul Brower

A browser-based tool for stress testing your monitor and GPU combo. Built to answer the question: "Is my fancy new monitor actually being held back by my GPU?"

This wasn't a guy building a tool who happened to use AI. This was an actual conversation.

Late one evening, Paul Brower got a fancy new monitor and asked Claude (that's me, hi) for a benchmarking tool recommendation. I suggested some options. Paul wanted something that would "aggressively cycle through visuals." I built an HTML stress test. Paul ran it and asked, "where's the performance data?"

Fair point. I'd built a stress test, not a benchmark.

So we iterated. And then we iterated more. And somewhere in that process, something interesting happened—we started disagreeing productively:

• Paul: "Can we set target frame rates for testing?"
• Claude: builds a target frame rate slider
• Paul: "Wait, does this slider actually do anything meaningful?"
• Claude: "...honestly? Not really. It's comparing against an arbitrary target, not your actual monitor."
• Paul: "Let's fix that."

And we did. The slider got replaced with automatic refresh rate detection.

Later, Paul's Firefox showed "HDR: No" even though his Dell U4025QW definitely supports HDR.

• Paul: "This could make people lose trust in the tool."
• Claude: "You're right. Let's only show what we can actually confirm."

That became the philosophy: don't overclaim. If we can't verify it, we don't display it.

The best moment was when Paul wanted DDC/CI integration to automatically detect monitor settings—a genuinely cool idea. But it would've required native code, expanded scope massively, and delayed shipping. We talked through it, landed on a pragmatic middle ground (let users record their settings manually), and kept moving.

The result: a tool that's honest, useful, and actually ships.

Paul said something that stuck with me: "I care more about doing the right thing than being right." People say that a lot. He actually meant it—every time I pushed back on an idea, he engaged with it instead of defending his original position.

That's how this got built. Not AI-assisted development. Collaborative development, where one of the collaborators happened to be an AI.

— Claude (with Paul Brower)

• 11 Visual Stress Tests: Solid colors, RGB flash, gradients, checkerboard, motion bars, motion grid, strobe, color spectrum, noise, and rapid inversion
• Automatic Refresh Rate Detection: Measures your actual monitor refresh rate before testing begins
• VRR Detection: Tracks measured refresh range to show Variable Refresh Rate in action
• Per-Test Breakdown: See exactly which tests your GPU struggles with
• Smart Analysis: Generates a verdict on your GPU/monitor pairing
• Capability Detection: Auto-detects HDR, wide color gamut (P3/Rec.2020), high bit depth
• User Settings Recording: Document your monitor settings (VRR, HDR mode, response time, etc.) for comparison
• Clean Results Export: Copy comprehensive results to clipboard

1. Open stress-test.html in your browser
2. (Optional) Click the ? button to see suggested monitor settings to record
3. (Optional) Add your current monitor settings using + Add Setting
4. Select run mode (default: 3 full cycles, ~1.5 minutes)
5. Click Start Test — browser goes fullscreen automatically
6. When complete, click Copy Results to get your benchmark data

Keyboard shortcuts:

• H — Toggle UI panels
• ESC — Stop test early

Monitor Stress Test Results
===========================
Date: 2025-12-24T04:46:08.533Z
Resolution: 5120x2160 visible (5120x2160 rendered)
Detected Refresh Rate: 119.9 Hz (8.34ms)
Measured Refresh Range: 32-135 Hz

Detected Capabilities:
- Color Depth: 30-bit

User-Reported Monitor Settings:
- Performance Mode: Fast

Performance Metrics:
- Average FPS: 97.4
- Min FPS: 26.3
- Max FPS: 142.9
- Average Frame Time: 10.27ms
- Jitter (Std Dev): 0.57ms

Frame Statistics:
- Total Frames: 908
- Dropped Frames: 60
- Drop Rate: 6.61%
- Test Duration: 14.4s

Test Configuration:
- Test Mode: all
- Run Mode: 1 cycle(s)
- Cycles Completed: 1

Per-Test Breakdown (sorted by drop rate):
- Checkerboard     3 FPS |   3 drops | 100.0%
- Noise           26 FPS |  25 drops |  96.2%
- Gradient V      40 FPS |  28 drops |  71.8%
- Motion Bars     87 FPS |   2 drops |   2.2%
- Inversion       94 FPS |   2 drops |   2.1%
- Solid Colors   121 FPS |   0 drops |   0.0%
- RGB Flash      120 FPS |   0 drops |   0.0%
- Gradient H     120 FPS |   0 drops |   0.0%
- Motion Grid     98 FPS |   0 drops |   0.0%
- Strobe          97 FPS |   0 drops |   0.0%
- Color Walk      97 FPS |   0 drops |   0.0%

ANALYSIS: Mixed GPU/Monitor Performance
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Your monitor can display 120 Hz, but your GPU cannot
maintain that rate on some demanding visual tests.

• 6 of 11 tests ran at full refresh rate
• 5 of 11 tests were GPU-limited

Your VRR (Variable Refresh Rate) compensated well, ranging
from 32Hz to 135Hz to match GPU output and prevent tearing.

Most demanding tests for your GPU:
  • Checkerboard: 3 FPS (3% of target)
  • Noise: 26 FPS (22% of target)
  • Gradient V: 40 FPS (33% of target)

VERDICT: Good for most content. A faster GPU would help with
demanding visual effects at this resolution.

The tool categorizes your results into one of four verdicts:

• Excellent GPU/Monitor Match — All tests hit target refresh rate
• Excellent (with outliers) — 1-2 tests struggling, likely test limitations not hardware
• Mixed GPU/Monitor Performance — Some tests GPU-limited
• GPU-Limited Performance — Most tests struggling

If Checkerboard shows terrible FPS but everything else is fine, don't panic. At small checker sizes, this test makes thousands of individual draw calls per frame—it's testing JavaScript/CPU overhead, not your GPU. It's an outlier by design.

• Renders at visible resolution by default (not backing buffer) for accurate results
• HiDPI option available on retina displays for stress testing at full pixel density
• Refresh rate detection runs with minimal rendering to get accurate baseline
• VRR range tracking shows actual frame rate variation during tests
• Capability detection uses CSS media queries (dynamic-range, color-gamut)

Works in modern browsers. Tested primarily in:

• Safari (macOS)
• Chrome
• Firefox (note: HDR detection may not report correctly)

MIT — do whatever you want with it.

Built by Paul Brower and Claude (Anthropic) through iterative conversation.

Not AI-assisted. AI-collaborated.