On Thursday 21 December 2000 08:45, Terry 'Mongoose' Hendrix II wrote: > Russell Coker wrote: > > That's a nice theory. But when I use CPU intensive programs like gzip on > > SPARC and Intel machines I find that I get about the same amount of work > > done MHz for MHz on Intel and SPARC. > > In my tests a P2 @ 400MHz is faster than a SPARC at 250MHz. > > gzip is I/O intensive, so duh - faster disk is more important than a > faster cpu. If you compare cpus, then use high cpu intensive, low I/O > intensive software - otherwise your results will only show overall > system performance. ;)
When the machine reports IO wait near 0 on Solaris (and the process is hardly ever in D state on Linux) and near 100%/num-cpus as the amount of CPU time in use by the process then the process in question is not IO bottlenecked. Here's the results of timing gzip -9 on my P3-650. It's got a 10G IBM IDE drive (and it's a Thinkpad which means an extra-slow disk by PC standards). [EMAIL PROTECTED]:/home/ftp/kern-src/2.4$time gzip -9 < linux-2.4.0-test12.tar.bz2 > /dev/null real 0m9.461s user 0m8.410s sys 0m0.230s As you can see 90% of the CPU time was used by gzip (and it was the only program running). Even the slow hard drives that Sun deigns to sell my clients are faster than my Thinkpad's drive, so the Sun's are unlikely to have IO bottlenecks running gzip either. Trust me, I know what I am doing and don't make claims about these issues without having test results to back them up. BTW If you think you know how to benchmark computers better than I do then please review my Bonnie++ and Postal programs and give me some suggestions for improvements. > > What do you mean "lag refresh"? Each pixel suddenly has it's value set > > to black. > > Cause the redraws to lag, thus making black spots and/or artifacts all > over the screen while waiting for the redraw to occur. It's more of a > problem with a laptop that's got a high load. It can also becaused by > poor screen blanking. Well that's nothing like my situation. I have the entire screen go blank for no good cause when it is under absolutely no load. I have had it happen when the only running process was vi! -- http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark http://www.coker.com.au/postal/ Postal SMTP/POP benchmark http://www.coker.com.au/projects.html Projects I am working on http://www.coker.com.au/~russell/ My home page
