On Mon, 13 Oct 2008, Carlos Moreno wrote:
Another really handy way to gauge memory speed on Linux, if there are
similar kernels installed on each system like your case, is to use
"hdparm -T".
Great tip! I was familiar with the -T switch, but was not clear on the
notion that the figure tells you that much about the overall memory
performance!
I wouldn't go so far as to say it tells you *much* about it, but it does
give a fairly useful comparison figure if the kernels are basically the
same and can help spot gross errors. I used it a bunch when I was
tinkering with DDR speeds and such earlier this year, it correlated fairly
well with other memory bandwidth measurements within the same processor
family (so far my tests suggestion results are much higher per clock on
Intel CPUs). Certainly of no use for comparison if one system has a
32-bit kernel and the other 64, and results will vary depending on general
kernel configuration (I get very different results from seemingly similar
RedHat and Ubuntu kernels on the same system for example).
It sounds like your CPU and memory setup are all fine, which leaves your
mystery open. Please let us know if you find anything interesting out,
normally an 8.3 upgrade would run faster so your situation is a bit
curious. The only other benchmark I'd suggest just as a sanity check is
bonnie++, that's a bit more thorough than what hdparm -t reports.
--
* Greg Smith gsmith@xxxxxxxxxxxxx http://www.gregsmith.com Baltimore, MD
--
Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance
