[R] Hardware oddity - linux/windows

michaell taylor pols1oh at bestweb.net
Tue Sep 9 14:55:18 CEST 2003


Hi,

I am trying to figure out why a new machine is running my R script so
slowly.  

The script was developed on a Linux , dual 2.8 xeon processor machine
with 4GB ram.  On this machine, the script runs in about an hour (it
creates lots of multilevel simulations). While running, 100% of a single
processor (50% of dual processor) is used consistently during the entire
run, about 800 MB of ram is utilized.  Xeon hyperthread condition
(on/off) seems not to matter to performance.  While the script runs, I
am using the "other" processor for miscellaneous emailing, editing, etc.

Moving the script to a windows XP machine with nearly identical hardware
configuration (2.8 dual xeon, 4.5 GB ram) the script takes nearly 2
hours.  Oddly, the maximum cpu load on this machine is 50% of a single
processor (25% of dual processor) during the run - which seems
consistent with the doubling of total processing time. While running
this script, this machine does no other tasks whatsoever. 

Both machines have at least double the memory required for the task,
both utilize fast scsi drives and all data is stored locally on the
drive (no data on network drives).  Moreover, being a simulated problem,
very little data is read/written to drives during the process.

Anyone have ideas as to why the Windows machine is so slow?

Thanks in advance.

Michaell Taylor




More information about the R-help mailing list