Benchmark

Spreadsheet write benchmark of openxml-office v4.x against a comparison library for each language.

How to read this benchmark

The charts compare v4.x with the other library at the same number of cells. A ratio of 1.00× means they are equal; above 1.00× is a v4.x gain, and below 1.00× is a v4.x loss. Green marks a gain and red marks a loss. Speed and allocation are shown separately because a library can be faster while allocating more memory, or vice versa. The exact ratios are printed beside each bar.

C#

• Generated: 2026-10-20 01:33:09Z
• Machine: Common (8 logical CPUs)
• Runtime: 10.0.12 on Ubuntu 26.04.1 LTS
• Layout: cells written in rows of up to 100 columns.

Notes:

• Alloc = total managed bytes allocated during the run (churn), Managed Δ = managed heap retained afterwards.
• WS = OS working set, Priv = private memory. CPU % is normalised across all logical processors (100% = one fully busy core-equivalent).
• µs/cell = wall time per cell, Cells/s = throughput.

Summary

Metric (across all sizes)	v4.x	v2.9.0
Total managed allocation churn (MB)	23.64	37210.09
Total GC collections (Gen0+1+2)	1	2510
Total GC pause time (ms)	0.78	2093.19

• v4.x is 291.5× faster on average (range 47.9×–563.9×).
• v4.x allocates 950.5× less managed memory on average (range 126.3×–1599.4×).
• v4.x triggered 1 GC collections vs 2510 for the legacy package.

Performance at a glance

The speed ratios below compare elapsed time at each workload size. Bar lengths are scaled within each chart; the printed ratios are the exact comparison.

Speed ratio

10 cells

563.86×

100 cells

494.35×

1,000 cells

279.43×

10,000 cells

71.96×

100,000 cells

47.90×

Less allocation ratio

10 cells

126.29×

100 cells

368.76×

1,000 cells

1,117.94×

10,000 cells

1,540.09×

100,000 cells

1,599.44×

Raw metrics

Library	Cells	Time (ms)	µs/cell	Cells/s	Alloc (MB)	Managed Δ (MB)	WS Δ (MB)	Peak WS (MB)	Peak Priv (MB)	Avg CPU (%)	Peak CPU (%)	GC 0/1/2	File (KB)
v4.x (project reference)	10	8.73	873.12	1145	0.12	0.14	0.06	122.31	182.80	13.7	13.7	0/0/0	4.21
v2.9.0	10	4923.14	492314.17	2	14.55	14.62	23.57	138.26	194.88	1.7	46.1	0/0/0	4.40
v4.x (project reference)	100	9.19	91.93	10878	0.12	0.13	0.00	132.37	189.00	12.4	12.3	0/0/0	4.60
v2.9.0	100	4544.28	45442.83	22	44.01	14.53	14.87	143.50	198.01	2.6	22.5	2/1/0	4.79
v4.x (project reference)	1000	16.83	16.83	59426	0.31	0.33	0.00	140.23	194.74	12.7	47.1	0/0/0	7.25
v2.9.0	1000	4702.12	4702.12	213	344.20	16.94	11.65	148.33	202.00	1.8	30.5	22/1/0	7.67
v4.x (project reference)	10000	108.89	10.89	91833	2.18	2.20	1.88	156.49	210.48	12.7	21.8	0/0/0	31.64
v2.9.0	10000	7836.23	783.62	1276	3351.02	6.58	18.54	161.24	228.22	6.3	31.6	224/3/1	34.75
v4.x (project reference)	100000	1119.14	11.19	89355	20.92	6.04	19.35	314.28	366.96	13.3	26.6	1/0/0	280.27
v2.9.0	100000	53607.02	536.07	1865	33456.32	64.75	82.73	258.66	358.36	11.7	41.6	2239/14/3	306.24

GC / memory stress

Highlights managed allocation churn and garbage collector pressure. High Gen0/1/2 counts, GC pause time and allocation rate indicate heavier memory stress.

Library	Cells	Alloc (MB)	Alloc/cell (bytes)	Alloc rate (MB/s)	Gen0	Gen1	Gen2	GC pause (ms)	GC pause (%)	Heap (MB)	Fragmented (MB)	Committed (MB)
v4.x (project reference)	10	0.12	12078	13.19	0	0	0	0.00	0.00	3.30	0.85	15.54
v2.9.0	10	14.55	1525422	2.95	0	0	0	0.00	0.00	3.44	0.98	7.93
v4.x (project reference)	100	0.12	1251	12.98	0	0	0	0.00	0.00	3.22	0.76	15.86
v2.9.0	100	44.01	461458	9.68	2	1	0	2.18	0.05	3.68	0.23	22.86
v4.x (project reference)	1000	0.31	323	18.30	0	0	0	0.00	0.00	3.27	0.77	19.59
v2.9.0	1000	344.20	360915	73.20	22	1	0	13.52	0.29	5.07	0.45	23.56
v4.x (project reference)	10000	2.18	228	19.98	0	0	0	0.00	0.00	3.29	0.70	16.05
v2.9.0	10000	3351.02	351380	427.63	224	3	1	177.32	2.26	10.57	1.47	33.71
v4.x (project reference)	100000	20.92	219	18.69	1	0	0	0.78	0.07	3.38	0.79	27.60
v2.9.0	100000	33456.32	350815	624.10	2239	14	3	1900.17	3.54	94.98	27.84	112.82

Head-to-head: v4.x vs comparison library

Each cell is a multiplier comparing the two libraries at the same cell count. Higher is always better for v4.x, and 1.00× means they are equal.

For example, 500.00× under Faster means v4.x finished the work 500 times faster than v2.9.0; 120.00× under Less RAM churn means v4.x allocated 120 times less managed memory.

Cells	Faster (time)	Higher throughput	Less RAM churn (alloc)	Less GC pause	Lower peak RAM	Lower peak private RAM	Smaller file
10	563.86×	563.86×	126.29×	n/a	1.13×	1.07×	1.04×
100	494.35×	494.35×	368.76×	n/a	1.08×	1.05×	1.04×
1000	279.43×	279.43×	1117.94×	n/a	1.06×	1.04×	1.06×
10000	71.96×	71.96×	1540.09×	n/a	1.03×	1.08×	1.10×
100000	47.90×	47.90×	1599.44×	2442.38×	0.82×	0.98×	1.09×

Go

• Generated: 2026-10-20T01:35:59Z
• Machine: Common (8 logical CPUs)
• Runtime: go1.26.4 on linux/amd64
• Layout: cells written in rows of up to 100 columns.

Notes:

• Alloc = total heap bytes allocated during the run (churn), Heap Δ = heap retained afterwards.
• Peak Heap = peak in-use heap, Peak Sys = peak memory reserved from the OS. CPU % is normalised across all logical processors (100% = one fully busy core-equivalent).
• µs/cell = wall time per cell, Cells/s = throughput.

Summary

Metric (across all sizes)	v4.x	excelize v2
Total heap allocation churn (MB)	20.51	114.81
Total GC collections	7	15
Total GC pause time (ms)	0.42	0.71

• v4.x is 0.1× faster on average (range 0.1×–0.1×).
• v4.x allocates 54.8× less heap memory on average (range 5.4×–195.3×).
• v4.x triggered 7 GC collections vs 15 for excelize.

Performance at a glance

The speed ratios below compare elapsed time at each workload size; allocation ratios show how much less heap v4.x used. Bar lengths are scaled within each chart; the printed ratios are the exact comparison. Green marks a v4.x gain and red marks a loss.

Speed ratio

10 cells

0.13×

100 cells

0.10×

1,000 cells

0.14×

10,000 cells

0.13×

100,000 cells

0.12×

Less allocation ratio

10 cells

195.25×

100 cells

56.26×

1,000 cells

11.06×

10,000 cells

6.01×

100,000 cells

5.39×

Raw metrics

Library	Cells	Time (ms)	µs/cell	Cells/s	Alloc (MB)	Heap Δ (MB)	Peak Heap (MB)	Peak Sys (MB)	Avg CPU (%)	Peak CPU (%)	GC	File (KB)
v4.x (module reference)	10	7.41	740.92	1350	0.01	0.01	2.84	13.03	12.7	0.0	0	4.22
excelize v2	10	0.94	94.21	10615	1.08	1.08	2.05	13.40	13.2	0.0	0	5.59
v4.x (module reference)	100	8.90	88.97	11240	0.02	0.02	2.84	13.66	12.7	0.0	0	4.60
excelize v2	100	0.87	8.67	115390	1.19	1.19	2.05	13.66	13.1	0.0	0	5.98
v4.x (module reference)	1000	16.34	16.34	61182	0.19	0.19	3.02	13.66	12.7	11.1	0	7.25
excelize v2	1000	2.36	2.36	423401	2.05	2.05	2.05	13.92	13.4	0.0	0	8.62
v4.x (module reference)	10000	103.89	10.39	96253	1.83	1.83	4.69	13.92	12.6	13.5	0	31.65
excelize v2	10000	13.91	1.39	718978	11.02	1.20	2.05	13.92	14.9	0.0	4	32.62
v4.x (module reference)	100000	1084.88	10.85	92176	18.46	-0.41	4.47	17.94	12.7	15.6	7	280.28
excelize v2	100000	128.64	1.29	777366	99.47	4.61	48.68	62.33	15.0	19.2	11	279.39

GC / memory stress

Highlights heap allocation churn and garbage collector pressure. High GC counts, GC pause time and allocation rate indicate heavier memory stress.

Library	Cells	Alloc (MB)	Alloc/cell (bytes)	Alloc rate (MB/s)	GC	GC pause (ms)	GC pause (%)	Heap Sys (MB)
v4.x (module reference)	10	0.01	578	0.74	0	0.00	0.00	7.59
excelize v2	10	1.08	112934	1143.26	0	0.00	0.00	7.56
v4.x (module reference)	100	0.02	222	2.38	0	0.00	0.00	7.50
excelize v2	100	1.19	12508	1376.39	0	0.00	0.00	7.53
v4.x (module reference)	1000	0.19	195	11.36	0	0.00	0.00	7.53
excelize v2	1000	2.05	2154	869.95	0	0.00	0.00	7.50
v4.x (module reference)	10000	1.83	192	17.66	0	0.00	0.00	7.50
excelize v2	10000	11.02	1156	792.50	4	0.18	1.30	11.50
v4.x (module reference)	100000	18.46	194	17.01	7	0.42	0.04	11.50
excelize v2	100000	99.47	1043	773.21	11	0.53	0.41	55.50

Head-to-head: v4.x vs excelize v2

Each cell is a multiplier comparing the two libraries at the same cell count. Higher is always better for v4.x, and 1.00× means they are equal.

Cells	Faster (time)	Higher throughput	Less RAM churn (alloc)	Less GC pause	Lower peak heap	Lower peak sys RAM	Smaller file
10	0.13×	0.13×	195.25×	n/a	0.72×	1.03×	1.33×
100	0.10×	0.10×	56.26×	n/a	0.72×	1.00×	1.30×
1000	0.14×	0.14×	11.06×	n/a	0.68×	1.02×	1.19×
10000	0.13×	0.13×	6.01×	n/a	0.44×	1.00×	1.03×
100000	0.12×	0.12×	5.39×	1.27×	10.89×	3.47×	1.00×

Other

Under construction

Benchmarks for additional languages are coming soon.