Damned Lies Once More
=====================

I conducted a little benchmark - I did run the "arc3d.mac" testing file for the
uFEM software [1] on a bunch of my computers. The pre-processing can take some
time but I only measured the actual run of the solver program. To make the
results (sort of) comparable, the program ran only on one CPU or core.

There are results:

Thinkpad X260      0:16
Thinkpad X61s      0:39
Indy R4400SC/175  36:30
O2 R10000/250     19:24
Compute Stick      1:18
Aquaris M10        2:03
SDF server         0:54

I'm not sure about the hardware at the SDF but it's somewhat modern AMD-based
server. The X260 (2016) is my newest laptop and the X61s (2006) is the old one.
The O2 is from 1998 (this praticular model) and the Indy is from 1994 (this
particular model). Everything is meant as a low-end stuff (not designed for CPU
performance). The Intel Compute Stick's (2015) Atom processor is not a
performer, of course. The M10 table (2015) only shows that it is realy not a
desktop/laptop replacement (but it's result is not that bad when compared ti
the Compute Stick...).

To be honest, I did expected larger difference between a new Thinkpad and the
10 years old one. The difference between 1998's O2 and the 2006's X61s is
obviously huge (I already knew that so I don't thing that this can be a
surprise).

However, I still feel more comfortable with the O2 than with any of the modern
machines.


Reference:

[1] https://github.com/jurabr/ufem