I've been trying to measure NetBSD software RAID-0 performance on my 
system, with 2 hard disks arranged for striping, however the numbers 
(bandwidth) don't seem to add up, i.e.

I'm using two Seagate IDE hard disks for RAID-0 /dev/wd0 and /dev/wd1, 
each connected to a seperate IDE controller
I run the following two commands concurrently to see how much data is 
being read from each drive at the same time

# dd if=/dev/rwd0d of=/dev/null bs=64k &
# dd if=/dev/rwd1d of=/dev/null bs=64k &

Running "systat vmstat 1" gives me the following reading in regard to 
data being read:

Disks: seeks xfers bytes %busy               100 pic0 pin 0            
64 fmin
   fd0                                                                 
85 ftarg
   
md0                                                                    itarg
   wd0         643   40M  97.0                                        
217 wired
   wd1         618   39M  
96.0                                            pdfre
   
wd2                                                                    pdscn
 raid0

However running

# dd if=/dev/rraid0d of=/dev/null bs=64k

Gives me the following systat reading:

Disks: seeks xfers bytes %busy                99 pic0 pin 0            
64 fmin
   fd0                                                                 
85 ftarg
   
md0                                                                    itarg
   wd0         606   19M  66.3                                        
205 wired
   wd1         606   19M  
52.5                                            pdfre
   wd2                                                                pdscn
 raid0         606   38M  97.0

My question is, why is there such a huge performance drop on software 
RAID-0 (I was expecting it to transfer at around 80M/sec, but it only 
manages around 40M/sec) when both hard disks and PCI bus are capable of 
higher bandwidth? I have experimented with different sectors per stripe 
settings etc, but it doesn't make much difference, in my case, the 
optimal value is about 64 sectors. Below is my raid0.conf file.

START array
# numRow numCol numSpare
1 2 0

START disks
/dev/wd0e
/dev/wd1e

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level
64 1 1 0

START queue
fifo 100