-=[ AZBENCH.TXT: Benchmarks of ARC, LHARC, PAK, PKZIP, and ZOO ]=- Page 1


Written by Eric Burleyson, sysop April 4, 1989
That Programming Place
(301) 474-1702
College Park, Maryland
PCPursuitable as Washington DC area code (202)
300/1200/2400 bps, 24 hours a day
Specializing in C/ASM/Pascal source code and utilities for MS-DOS



Now that the SEA versus PKware lawsuit is behind us, we have only to
contend with the turmoil of its aftermath! With this in mind, I
devised a test to satisfy my curiosity of the relative performance of
the various file compression packages available. I intend to update
this as necessitated by new releases -- you can help by uploaded the
latest releases of all worthy file compression packages to by BBS.

My goal was to do a real-world comparison, not a strictly-controlled
laboratory experiment. I did, however, take reasonable care to produce
fair and accurate results (e.g., all runs were performed in the same
clean environment: no multi-tasking or obtrusive TSR's were active,
and the disk had been de-fragmented recently.) For the insanely
curious I've included the BAT files that I used for the test -- be
forewarned that you'll need the STOPWATCH or similar program, as well
as gobs of a available disk space, to perform the same test that I
did.

All the compressed files consist of the same 7 files (the 2 binary and
5 ascii files from the shareware package TheDraw v3.10) totaling
332,504 bytes. I chose these arbitrarily as a representative mix of
files. Some of the packages maniplulate more than one style of
compressed file -- I ran these once for each type of supported file.
All packages are very recent versions (first quarter 1989.) I ran test
was run on an both an XT clone and an AT clone -- the results of each
test are presented on the next page.


-=[ AZBENCH.TXT: Benchmarks of ARC, LHARC, PAK, PKZIP, and ZOO ]=- Page 2



Here's the performance benchmarks of the current crop of data
compressors. Uncompressed data size was 332504 bytes (a total of
7 ascii and binary files). Notes follow below.

Compress/Extract Time (sec)
Software - Style Size Ratio on XT clone on AT clone
------------------ ------------ ----------- -----------
ARC v6.00 - ARC 183285 55% 100 / 65 34 / 21

LHARC v1.0 - LZH 138998 42% 257 / 103 86 / 33

PAK v1.6 - PAK 156430 47% 154 / 121 46 / 36
- ARC 178439 54% 110 / 98 34 / 30
- ARCsq 174047 52% 104 / 94 32 / 28

PK(UN)ZIP - ZIP 166520 50% 49 / 44 14 / 12
v0.92 - ZIPex 149140 45% 120 / 39 50 / 11

ZOO v2.01 - ZOO 174844 53% 93 / 99 37 / 23

For the sake of historical perspective, here're the figures for
the verisons of ARC and PKARC current around late July 1988, the
the of the great 'shareware lawsuit'. Notes follow below.

Compress/Extract Time (sec)
Software - Style Size Ratio on XT clone on AT clone
------------------ ------------ ----------- -----------
ARC v5.22 - ARC 181061 54% 326 / 201 109 / 68

PK(X)ARC - ARCsq 173498 52% 57 / 42 14 / 12
v3.6 - ARC 178357 54% 53 / 43 16 / 13

Notes
-----
'XT clone': 8088-2 @ 7.44 MHz, standard 20 Mb disk, MS-DOS v3.1

'AT clone': 80286 @ 7.16 MHz, 0 wait states, standard 40 Mb disk,
MS-DOS v3.1

'Style' values indicate the type of compressed file produced. In
general I used the default file extensions generated by the individual
compression programs with these caveats:

ARC indicates the SEA-standard ARC format
ARCsq indicates the ARC variant that includes 'squashing'
ZIPex indicates the ZIP format with maximum compression selected

'Size' is the number of bytes in the compressed file; smaller is
better.

'Ratio' is size of the the uncompressed files divided by the size of
the compressed file; smaller is better.

'Time' is number of seconds of wallclock time the operation took;
smaller is better.