Performance Report

GitHubEdit on GitHub

This page summarizes key performance data for Univer Pro. Results depend on hardware, dataset size, and configuration.

Load Snapshot from Memory

Workbook CellsTime (s)
1M (100,000 rows, 10 cols)0.274
1M (10,000 rows, 100 cols)0.241
10M (1,000,000 rows, 10 cols)1.520
10M (100,000 rows, 100 cols)0.945

Formula Calculation

Formula ComplexityTime (s)
20,000 formulas, random ranges1.32
20,000 nested formulas1.37
20,000 formulas with VLOOKUP4.73
20,000 formulas with nested SUM1.38

Rendering Performance

Test machine: Windows, 32GB RAM, i9-13900H.

Cell CountScrolling FPS
100k (10k rows, 10 cols)50 ~ 60
200k (20k rows, 10 cols)50 ~ 60
1m (100k rows, 10 cols)50 ~ 60
6m (600k rows, 10 cols)50 ~ 60

Pivot Table Performance

Test machine: macOS, Apple Silicon M2 Pro.

Cell CountRead Time (s)Query Time (s)
500k (1 row dim, 1 col dim, 2 value dim)0.2520.040
1M (1 row dim, 1 col dim, 2 value dim)0.3760.078
2M (1 row dim, 1 col dim, 2 value dim)0.7320.140
5M (1 row dim, 1 col dim, 2 value dim)3.2000.415

Office Format Conversion

Test machine: Linux, 8GB RAM, Intel Xeon Platinum 8369B 4 cores.

Cell CountFile SizeImport .xlsx (s)Export .xlsx (s)
50k (1,000 rows / 25 cols / 2 sheets)0.5M0.040.23
500k (2,500 rows / 50 cols / 4 sheets)5.4M1.222.5
1M (5,000 rows / 25 cols / 8 sheets)11M2.395.1
5M (10,000 rows / 50 cols / 10 sheets)55M11.325.3
10M (200,000 rows / 50 cols / 1 sheet)110M34.454.9

Collaboration Latency

On a 4C8G server, 200 concurrent users achieved ~1.3s collaboration latency, close to industry benchmarks.

Collaboration latency

How is this guide?