Benchmark test of TI-92, 92II and 92+


[Prev][Next][Index][Thread]

Benchmark test of TI-92, 92II and 92+



Since there have been some controversies regarding the speed of TI-92
and its clones TI-92II and TI-92+ I have compiled a list of measured
benchmarks.

As a benchmark test I have chosen inversion of a 10 by 10 matrix in both
"approximate" and "exact"  mode, i.e.,

randmat(10,10)^-1

Note that the choice is not beneficial for TI-92 if it is to be compared
to other calculators since the algorithm for inversion of matrices seems
unnecessary slow.  According to John Edry <jeejohn@aol.com> inversion of
a 30 by 30 matrix takes 1297s on a TI-92 against only 174 s on a TI-85
and 81 s on a HP-48GX. The time for inverting an (N*N) integer matrix on
a TI-92 grows as expected N^3 for approximative calculations but almost
as N^4 for exact calculations. However, the simple benchmark suffices
for indicating the speed differencie between the different TI-92 clones.
Note that the values between units of the same type can vary a few
percent due to variations in battery condition, clock speed and memory
status (e.g. the calculator slows down when the history file grows). All
times are in seconds

Calculator              Exact           Approximative   Measured by
__________________________________________________________
TI-92                   45                      26                              RS
TI-92II                 32                      18                              GJ
TI-92+                  31                      21                              JMF
TI-92ext                32                      18                              JJM
TI-92acc1               17                      10                              RS
TI-92ext/acc2   12                      7                               JJM

ext= home made memory extension, see www-page of Jean-Jacques Michel
 http://www.daewoo.fr/~jjm/ti92.html
acc1=accelerated (C11=10pF), see webpage of Boris Lutz
http://cccsrv.trevano.ch/~blutz/TI92Spec/TIAccelerate.html
acc2=accelerated (C11=8.2pF)

RS=Richard Schatz (myself) <rschatz@ele.kth.se>
GJ=Gareth James <hjames@globalnet.co.uk>
JMF=JM Ferrard <jm-ferrard@infonie.fr>
JJM=Jean-Jacques Michel  <jjm@han.daewoo.fr>

Note that the values for TI-92+ is for a beta-version since it has not
been released yet. The conclusion is that the TI-92+ is a bit faster
than TI-92 and approximately similar to TI-92II. The values for the
"home" extended TI-92 confirms that this speed increase is solely due to
the larger (and better organised) memory which can utilise the 16 bit
bus instead of addressing the internal 8 bit RAM-memory twice.

JM Ferrard has done a comprehensive review (72! screenshots) of TI-92+
that can (in french) be found on the web
http://www.mygale.org/03/ti92plus/document/ferrard/ferrard.htm
For those (like me) who are not familiar with the french language he has
also made an english version which he was kind to mail me. I have
relayed it on
http://www.ele.kth.se/FMI/staff/92+_eng.zip
It is a zipped RTF-file which should be possible to open with e.g. Word.

I want to thank all contributors to this benchtest

Richard Schatz


Follow-Ups: