On Sun, 30 Mar 1997, Curt Sampson wrote:

> On Mon, 31 Mar 1997, Soren S. Jorvang wrote:
> 
> > What I meant was, I got similar performance from 'dd bs=8k' and
> > 'dd bs=64k', as would be expected with a smaller MAXPHYS.
> 
> Oh, and now you get better. I see. Well, that's fair enough. I
> personally don't think that what appears, off-hand, to be a 10-15%
> speed increase is worth the loss of disk space, but that's obviously
> a personal decision.

I have just tried it again on some of the disks I mentioned.
'dd if=file of=/dev/null bs=64k' (Throughput measured with iostat):

The end of a 5400 rpm disk:

8K/1K:  0.4MB/sec
32K/4K: 1.7MB/sec.
64K/8K: 3.4MB/sec
raw:    3.5MB/sec

The end of a 7200 rpm disk:

8K/1K:  0.7MB/sec
32K/4K: 3MB/sec
64K/8K: 5MB/sec
raw:    5.5MB/sec

This is a silly benchmark of course, but it seems to me that real
performance is similar.

I consider the difference in speed important enough that I am willing to
waste some of the disk to get it.



-- 
Soren