Desktops tend to run faster than laptops; and more expensive PCs tend to run faster.
But how much faster are bigger or more expensive PCs? Are all "i7" cpus equally fast? The
table below gives some answers. We tested 4 PCs:
2 Pro is a "convertible tablet" with fold-back keyboard. Bought 2014.
the "Jerie PC" is a fairly standard business desktop. Bought 2014.
the "Horridge PC" is a more expensive desktop or "workstation", using a specialized
Sandy Bridge-E cpu and motherboard. Bought 2013. Hyperthreading
was disabled in the BIOS.
Note that all 4 PCs were fairly high-end: all had SSD drives and an "I7" cpu. "I5" and
especially "I3" CPUs would run slower.
Two recursive-dynamic jobs were run with RunDynam with all 4 PCs:
Job1 is small. It solves a 1-region, 62-sector SAM-based model for 40 periods (120
Job2 is larger. It solves an 8-region, 58-industry, 63-commodity model for 26 periods
Both jobs ran base, rerun, and policy simulations for each period. RunDynam allows for
either 2 or 3 simulations (sims) to be run at once: both options were tested. You could
download the tests to run on your own PC.
For the smaller Job1, all PCs performed similarly, and there was little time saving
by running 3 sims at once rather than 2.[This is mainly due to some details of the way RunDynam works: it is not particularly efficient with very small models.]
For the larger Job2, differences emerge. However, the differences are not too large:
at worst the little Yoga 2 Pro takes 50% longer than the beefy workstation.
The cheaper desktop performs similarly to the workstation -- in some cases the
desktop goes faster! [See the row labelled Speed.]
The final row shows the percent time saving by running 3 sims at once rather than 2
(the theoretical maximum is 26% [first policy shock was 5 years in]). Here the
workstation draws ahead, probably due to its Sandy Bridge-E architecture and bigger cache
(both helping with multi-processing).