|
Goals
1)
Compare top compression formats 7Z, ARC, and
ZPAQ (from PeaZip) with RAR (WinRar) and ZIPX (WinZip) in terms of
maximum possible compression ratio.
2) Evaluate how each format scales in terms
of compression ratio improvements versus compression and extraction
speed reduction, with the increase of compression level.
|
|
Software
settings
Benchmarks
are conducted
on Windows 10 2009 64 bit using 64 bit versions of:
- PeaZip 7.3.1 and 9.5.0 for Addendum
section
- WinRar 5.8
- WinZip 24
All applications are tested using default,
out-of-the-box compression settings for the selected archive format.
No cryptography option is set, since encryption impact on performances
is out of the scope of
this benchmark.
|
|
Hardware
settings
Notebook with Intel Core i7-8565U CPU, 4
physical cores with hyper-threading (8 logical cores), 8 GB RAM
System disk 512 GB PCIe NVMe SSD, NTFS filesystemaaa
|
|
Input data
Benchmark input, same of general purpose compression benchmark,
contains
42 files in 4 directories for total 303 MB
(318.000.857 bytes), composed by:
|
|
|
|
Archive
formats compared in this benchmark
- 7Z file
format
(7-Zip, 1999) Open Source archive format, widely used and well-known
for excellent compression ratio, tested using default LZMA2 algorithm
- ARC file
format
(FreeArc, 2007) Open Source archive format
designed aiming to superior compression. performances
- RAR file
format
(RarLabs, 1995) proprietary archive format quickly building reputation
of being powerful compressor outperforming contemporary zip, bz2, and
ace formats. Latest RAR5 format revision is being tested.
- ZIPX
file format
(WinZip, 2008) designed to improve compression ratio over ZIP with
support for more compression algorithms. At minimum and default
compression level Deflate algorithm is used, at maximum compression
level WinZip tries to determine the most fit algorithm from supported
ones, including BZip2, LZMA and PPMd.
- ZPAQ format (Matt Mahoney, 2009)
Open Source archive format, computing intensive, providing very high
compression ratio,
supersedes previous PAQ format, multiple times Hutter Prize winner.
ADDENDUM
With PeaZip 9.5.0 BR and ZST compression
formats were also tested, at maximum compression settings
- Brotli file
format
(Google, 2013) Open Source pure compression format developed for
maximum compression and decompression speed, but very flexible and
capable of also providing high compression ratio.
- Zstandard file
format
(Yann Collet, Przemysław Skibiński, Facebook, 2015) Open Source pure
compression format, Zstd compressor was developed with same scope of
Brotli, also very
flexible and being capable of deliver high compression ratio in
alternative to impressive compression speed.
|
|
|
Maximum compression
benchmark methods
Benchmark
input data is saved to system disk (PCIe SSD) and compressed to system
disk, same partition, separate directory; the resulting archives are
then extracted to
separate directory on same (system) disk/partition.
Each compression and extraction test is repeated 10 times to get an
average value; size is expressed in MB, time in seconds.
Each archive format is is tested with minimum (fastest non-store
level), default, and maximum
compression settings.
Default compression algorithm and default compression settings, as
pre-set out of the box by file archivers being tested (PeaZip, WinRar,
and WinZip), are employed for each format / level.
For Brotli and Zstandard compression, the input data was consolidated
in a single TAR archive for running the benchmark
Maximum
compression
benchmark results
table, the lower the better for all columns
Utility, format, level
|
Compression
SSD (sec)
|
Archive
size (MB) |
Compression
ratio
|
Extraction
SSD (sec) |
PeaZip,
7Z, fastest
|
3.5
|
92.90
|
30.66%
|
0.9
|
PeaZip,
7Z, default
|
39.6
|
73.60
|
24.29%
|
1.0
|
PeaZip,
7Z, ultra
|
125.0
|
71.40
|
23.56%
|
2.7
|
PeaZip,
ARC, 1
|
1.4
|
109.00
|
35.97%
|
1.9
|
PeaZip,
ARC, 4
|
17.2
|
71.70
|
23.66%
|
7.1
|
PeaZip,
ARC, 9
|
81.0
|
66.10
|
21.82%
|
40.8
|
PeaZip,
ZPAQ, fast
|
6.3
|
102.00
|
33.66%
|
2.1
|
PeaZip,
ZPAQ, normal
|
37.2
|
68.50
|
22.61%
|
26.8
|
PeaZip,
ZPAQ, ultra
|
354.0
|
57.60
|
19.01%
|
356.0
|
PeaZip,
Brotli (max)
|
190.0
|
82.60
|
27.26% |
1.3
|
PeaZip
Zstandard (max)
|
107.0
|
76.80
|
25.335%
|
1.0
|
WinRar,
RAR, fastest
|
2.0
|
106.00
|
34.98%
|
1.0
|
WinRar,
RAR, normal
|
13.8
|
80.40
|
26.53%
|
1.0
|
WinRar,
RAR, best
|
48.0
|
78.90
|
26.04%
|
1.8
|
WinZip,
ZIPX, fastest
|
3.0
|
105.00
|
34.65%
|
4.5
|
WinZip,
ZIPX, enhanced deflate
|
21.0
|
94.00
|
31.02%
|
5.0
|
WinZip,
ZIPX, best method
|
34.7
|
70.70
|
23.33%
|
46.2
|
Maximum compression
ratio
results
ZPAQ reached maximum compression ratio, compressing the 303.00 MB input
down to 57.60 MB (19.01%), followed by ARC with output reduced to 66.10
MB. ZIPX and 7Z were able to reduce the input size to 70.70 MB and
71.40 MB respectively, and RAR scored the most poor compression
performance with 78.90 MB output employing best compression setting.
All format shown a significant improvement in compression ratio
switching from fastest to best compression settings, but for most
formats most of the advantage emerged switching from fastest to default
compression settings.
Especially, 7Z and RAR format shown very small improvements from
default to ultra settings - which, as you can see in following chapter,
comes at cost of much longer compression times.
The exception to this behavior is ZIPX format, providing a mediocre
deflate based compression unless switching to alternate compression
algorithms, with maximum compression, comparable to 7Z ultra, being
attained using "best method" settings.
Even if usually associated with fast compression, both Brotli and Zstd
compressors attained quite high compression ratios when used at maximum
compression settings, with Zstandard being the best of the two and
compressing slightly better than RAR at best compression level (and
Brotli slighly worse).
Both, however, compressed less than 7Z at medium compression level.
Maximum compression
speed
results
Compression times increases for all format with the increase of
compression settings, generally with best returns (in terms of
compression ratio improvements) switching from fastest to default
compression settings, and diminishing returns switching to ultra/best
settings.
At maximum compression level, ZIPX is the fastest format, followed by
RAR, ARC, and 7Z, ZPAQ being the slowest.
Using moderate compression settings, RAR and ARC emerge as the fastest
formats.
Brotli suffered a noticeable performance penality when used at maximum
compression level, being the second slowest compressor.
Zstandard too took a seizable performance hit, but overall remails
noticeably faster than Brotly when both are used at maximum compression
level,
Maximum compression
extraction speed results
RAR and 7Z formats shows a clear advantage in terms of extraction speed
compared to all other formats, with decompression times staying under 3
seconds even at higher compression levels.
ARC, ZIPX, and ZPAQ decompression speed increases significantly at high
compression settings, with ARC being generally comparable or better in
speed than ZIPX, and ZPAQ being the slowest.
It is noteworthy to point out that both Brotli and Zstandard excels in
extraction speed even for archives created at maximum compression
settings, and are amongst the fastest extractors in this benchmark,
with a slight advantage for Zstd in terms of speed.
|
Best format for
maximum possible compression ratio
ZPAQ is the
winner in
terms of maximum attainable compression,
but is slower than other formats.
ZPAQ at maximum compression level reached a 19.01% compression ratio
versus 21.82% reached by ARC at maximum compression level, the second
best result of the benchmark.
Anyway, even ZPAQ at default level can compare favorably (in terms of
compression ratio versus speed tradeoff) with other formats at best
compression settings - only ARC at highest level surpassing its
compression ratio - with a 10x speed advantage of ZPAQ default
vs ZPAQ ultra.
For all formats excluding ZIPX, average compression settings
represented an optimal tradeoff between compression ratio and
compression speed, with diminishing returns for switching to highest
compression levels.
For all format each compression level increase represented an higher
computational cost (with longer compression times), for smaller
improvements in compression ratio.
For this reason it can be recommended to use other methods than
increasing compression level to keep the output below a desired
threshold, in example spanning the outpput
to
multiple volumes of fixed size in order to meet maximum size
constrains, or deduplicate
input data before
compression.
This is even more true when compressing data sets containing
multimedia, or encrypted files, which generally does not compress
well, or does not compress at all regardless employed
algorithm/settings.
Unlike compression speed, which generally scaled uniformly with
increasing of compression level, extraction speed was more correlated
with the archive format nature, with 7Z and RAR decompression times
remain fast (well below 3 seconds) at any compression level.
Benchmark
conclusions
in brief:
What is the most
powerful file compressor?
ZPAQ is clearly the top performing format in this benchmark focused on
maximum attainable compression.
What is the
overall
best compression format?
It depends on user's
need, with compression ratio being only one factor
of the equation.
ZPAQ and ARC are the best compressors, but 7Z and RAR formats has a
clear advantage in terms of decompression speed, faster
than for any other tested format.
7Z vs RAR, which
is
the best compressor?
7Z outperformed RAR in terms of
compression ratio at all compression
levels, but RAR outperformed 7Z in terms of compression speed.
Extraction speed is quite similar, and keeps reasonable for both
formats at all compression levels.
Are Brotli and
Zstandard suited for max compression?
Brotli and Zstd can provide surprisingly good compression ratios, even
if they are designed primarily for fast compression tasks, with
Zstandard being overall the best choice.
It must be noted the performance penality is very relevant in terms of
compression speed, with traditional compression formats like 7Z and RAR
being more fit for this task.
However, decompression speed remains very high even extracting BR and
ZST files compressed at maximum settings.
Is it worth to set
best /
maximum / ultra compression settings?
Increasing compression level decreases performances with diminishing
returns. File spanning and data deduplication are recomended in
alternative to higher compression settings, and solid compression
option is a must to attain highest compression ratio when multiple
similar files are involved - even if single files are not compressible
(in the traditional sense) taken one by one.
Read more about how to create 7Z
files, or alternatively how to
create ARC files, or how
to create PAQ / ZPAQ files if maximum compression is needed. Please
note PeaZip is also capable to create
RAR files if WinRar is installed in the system, and to create ZIPX files following the
new WinZip standard (even if not all allowed algorithms are yet
supported).
Synopsis: Maximum file
compression benchmark. 7Z, ARC, ZIPX versus RAR comparison for best,
most powerful compressor format. What algorithm compresses the most.
What archive format reach highest compression ratio. Which file
archiver utility compresses better. What are the best compression
options and settings.
Topics: maximum compression
benchmark, best file compressor, best archive format for max
compression, 7z vs rar vs zpaq
PeaZip > Compression
benchmark > Maximum compression benchmark: 7z, arc, zpaq vs rar, zipx
|
|
