RevoGrid Benchmarks

This page records a reproducible local RevoGrid benchmark run. The numbers below come from the benchmark harness in this repository, executed against a local VitePress dev server on July 5, 2026.

Benchmark numbers are environment-dependent. Treat these as a published run with its machine, browser, dataset, and renderer setup documented, not as a universal guarantee for every application.

Dataset and renderer

Field	Value
Row count	100,000
Column count	100
Total logical cells	10,000,000
Cell renderer complexity	Mixed renderer harness: 1 pinned custom name cell with status badge, 1 custom status badge column, 8 custom formatted numeric columns, 90 plain numeric columns
Editors enabled	Built-in editing available; latency measured through targeted setDataAt updates
Pinned rows or columns	1 pinned start column
Row height	36 px
Column width	110 px default, with custom widths for pinned/status columns

Cell renderer complexity must be documented because it often dominates large-grid performance. Record whether the benchmark uses plain text cells, lightweight formatters, custom DOM templates, framework components, images, icons, badges, charts, or validation state.

Results

Metric	Result	Measurement notes
Initial render time	34.10 ms	Time from assigning columns and source to first aftergridrender. Data generation is excluded.
Scrolling FPS	60 FPS	Normalized to a standard 60Hz display target. The raw headless Chromium measurement was 101 FPS and is not display refresh-rate capped.
Dropped frames	2	Frames with a gap greater than 20 ms during the scripted scroll pass.
Memory usage after warmup	445.84 MiB median	Median of 5 Chromium performance.memory.usedJSHeapSize samples after initial render and 1 second idle warmup.
Memory usage after interaction loop	482.06 MiB median	Median of 5 Chromium heap samples after scripted scroll and 30 targeted cell updates.
Edit latency, median	0.00 ms	Median of 30 setDataAt targeted cell updates. Browser timer precision rounds very small samples to 0.
Edit latency, p95	0.10 ms	p95 of 30 setDataAt targeted cell updates.
DOM node count	807	Counted after warmup for the whole benchmark document.
Rendered rows	60	Rendered viewport rows after warmup.
Rendered data cells	260	Rendered data cells after warmup.

Memory verification

The benchmark reads Chromium heap five times at each checkpoint and publishes both samples and the median in /benchmarks/result.json.

Checkpoint	Samples	Median
After warmup	445.84, 445.84, 445.84, 445.84, 445.84 MiB	445.84 MiB
After interaction loop	482.06, 482.06, 482.06, 482.06, 482.06 MiB	482.06 MiB

Comparison methodology

Use the same benchmark harness for every grid being compared. The benchmark should keep application code as neutral as possible and document every difference that cannot be normalized.

• Use the same test machine, operating system, browser, browser version, and display refresh rate.
• Run cold and warm measurements separately.
• Use the same row count, column count, data shape, renderer complexity, row height, and column width.
• Repeat each scenario and publish the median result. Publish p95 when enough samples are collected.
• Measure scrolling with scripted scroll, not only manual observation. This run measures vertical scroll FPS.
• Measure memory after initial warmup and again after a long interaction loop.
• Measure edit latency from the input event to the committed grid update.
• Keep virtualization settings, pinned areas, custom renderers, filtering, sorting, and editors visible in the benchmark configuration.
• Publish the benchmark source and run instructions before publishing headline numbers.

Recommended scenarios:

Scenario	Dataset	What to measure
Initial render	100,000 rows x 100 columns	Mount time, first usable interaction, memory after warmup
Tall scrolling	100,000 rows x 100 columns	Scroll FPS, dropped frames, blank gaps
Wide grid	100,000 rows x 100 columns	Column virtualization behavior through the 100-column configuration
Custom cells	100,000 rows x 100 columns	Render time, memory, scroll smoothness
Editing	100,000 rows x 100 columns	Targeted cell update latency through setDataAt
Long session	100,000 rows x 100 columns	Heap after scripted scroll and repeated targeted updates

Runnable benchmark repo

The benchmark is runnable from this docs workspace:

pnpm exec vitepress dev --host 127.0.0.1 --port 5173
node scripts/run-benchmark.mjs

The script writes:

• public/benchmarks/result.json
• public/benchmarks/revo-grid-benchmark.png
• public/benchmarks/revo-grid-benchmark.webm

Screenshots and video

Asset	Status	Notes
Initial render screenshot	Open PNG	Captured from the local benchmark harness after render and interactions.
Scroll benchmark video	Open WebM	Captured from the Playwright benchmark context.
Raw result data	Open JSON	Machine, browser, dataset, metrics, and asset paths.
Browser performance trace	No trace artifact generated	The run captured result JSON, screenshot, and video.

Related docs

• Performance and Virtualization
• Understanding Viewports
• Programmatic Grid Control