It's not great science, but it's a real-world benchmark to me, because it's something I did frequently over the years and will therefore benefit from this concretely.
For the last 5 years, I've had a Sun Blade 150. Now it has 650Mhz single core Ultrasparc IIi cpu, 2GB of RAM, and 7200RPM IDE disks, and every version of Solaris 9/10/11 that caught my fancy. Earlier this month I bought half a dozen Sun Ultra 24s. That's a quad-core Intel cpu, 3.0Ghz, with 4GB of RAM and 7200RPM SATA disks. My task: I downloaded opensolaris snv b101. I ran digest -a md5, unzipped the 2 pieces, ran digest -a md5 on the results, cat'ed them together, ran digest -a md5 again. On the Sun Blade 150, the elapsed wall clock time was 18 mins and 38 seconds. On the Sun Ultra 24, the elapsed wall clock time was 5 mins and 20 seconds. I call that a performance improvement of 354%, or stated negatively, the Sun Blade 150 is 28% of the Sun Ultra 24's performance. Lastly I figured what could run in parallel (duh!) and I got the test suite down to 3 min 58 seconds. Wish I could replace my US-IV cpus in my 6800s with some hardcore quadcore action! Pricing on the Sun M8000 sucks! -- This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list [email protected]
