Hi,
my test is very simple:
1) boot on cdrom netbsd v3 amd64,
   exit install and launch shell :
   mount /dev/wd0a /mnt && cd /mnt
   dd if=/dev/zero of=writefile bs=16k count=16k
    29818880 bytes transferred in 24.318 secs (1226206 bytes/sec)
   rm writefile
   dd if=/dev/zero of=writefile bs=64k count=64k
    17170432 bytes transferred in 11.307 secs (1518566 bytes/sec)

2) boot on cdrom netbsd v3 i386,
   exit install and launch shell :
   mount /dev/wd0a /mnt && cd /mnt
   dd if=/dev/zero of=writefile bs=16k count=16k
    268435456 bytes transferred in 4.085 secs (65712473 bytes/sec)
   rm writefile
   dd if=/dev/zero of=writefile bs=64k count=64k
    4294967296 bytes transferred in 71.731 secs (59876027 bytes/sec)

i386 have 10x perf !

same disk, just reboot on cdrom amd64/i386 ...

anyone have same pb ?

dmesg amd64 :
wd0 at atabus4 drive 0: <ST3200021A>
wd0: drive supports 16-sector PIO transfers, LBA48 addressing
wd0: 176 GB, 366247 cyl, 16 head, 63 sec, 512 bytes/sect x 369177306 sectors
wd0: 32-bit data port
wd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 5 (Ultra/100)

dmesg i386 :
wd0 at atabus4 drive 0: <ST3200021A>
wd0: drive supports 16-sector PIO transfers, LBA48 addressing
wd0: 176 GB, 366247 cyl, 16 head, 63 sec, 512 bytes/sect x 369177306 sectors
wd0: 32-bit data port
wd0: drive supports PIO mode 4, DMA mode 2, Ultra-DMA mode 5 (Ultra/100)
wd0(ixpide0:0:0): using PIO mode 4, Ultra-DMA mode 5 (Ultra/100) (using DMA)

dmesg diff amd64/i386 missing last line on amd64 ! (?)
{I have rebooted 10x and same missing last line on amd64}

Best Regards
Rmkml