On 10/30/07, Jan Engelhardt <[EMAIL PROTECTED]> wrote: > > On Oct 30 2007 13:46, Greg Freemyer wrote: > > > >Surprisingly (to me at least) we are seeing a speed improvement with a > >specialized version of dd (dcfldd) going against raw disks. > > > >ie. dcfldd if=/dev/sdb of=/dev/sdc bs=4k and dcfldd if=/dev/sdc > >of=/dev/null bs=4k > > Bottleneck: Disk. You cannot really use something like dd or variants > to benchmark your CPU. What you can do is comparing the encoding time > of oggenc or so (but be sure to use the SSE2 mode in the 32-bit > compilation mode before testing, because x64 uses SSE2 by default).
I fully expected the disk to be the only bottleneck. My surprise was that 32/64 bit compile issues had any impact at all. Let alone 25% faster with full 64-bit. FYI: I'm using dcfldd as my benchmark tool because it _is_ my workload!! Indeed we have 4 computers dedicated to running it with a few simple scripts automating the process. We are in the process of replacing the PCs and upgrading to 10.3 as our OS. (We call the process "Forensic Imaging".) Greg -- Greg Freemyer Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]