These types of tests should be run against the raw device. You are proving nothing with this test.
On Mon, Aug 25, 2008 at 06:58:30PM -0700, Neko wrote: > Hi all, > > having a 250 GB drive on a PATA strip using lowest PIO mode (without dma if > possible), drive specs show a 8 MB buffer , > > 2 cases are : on same pata strip, one on each strip, > > important to note that the booted drive is the if's straight up to my slave. > (yea i know ill have to run fsck afterwards since itll think it still mounted) > and having no user but su himself. > > i had ran mine at 4mb block space thinking ill use the 16mb bus transfer > divided at most in 4, per second, but i achieved that in a minute instead. > this is really poor performance, 3 days for 250gb transfer at 4mb bs > > dd if=/dev/wd0c of=/dev/wd1c bs=4m > > so what if i was to use > > dd if=/dev/wd0c of=/dev/wd1c bs=4000m > > or more, or reflecting number sector divided by lets say 16... would in > the end i get the same effect, what are the plateau i should top of, > a mb is way too cheezy > > i have 512mb of ram, would 512mb be the most ? the machine is idling either > way. > > > neko