Has anyone played with tape drives on the pmax?

I've got an Exabyte 8500 on my 5900/260.  I'm writing a tape trying to
use 10k blocks with the following:

	mona:~ 18> tar cf - /netbsd | dd of=/dev/nrmt0 bs=10k conv=sync
	tar: Removing leading / from absolute path names in the archive.
	0+645 records in
	645+0 records out
	6604800 bytes transferred in 54 secs (122311 bytes/sec)

but it seems that the tape is only getting written to in 512 byte
blocks.  Using tapex in an Ultrix box to test the tape I just wrote:

	balrog:~ 2# /usr/field/tapex -m -f /dev/nrmt1h

	Testing tape device /dev/nrmt1m   (interface SCSI, controller #0, TZ8MM)
	Wed Sep  1 13:09:18 1999

	----------------------------------------------------------
	Tape disection test.

	This test will read a tape and print out the format of
	what is on the tape.  Information includes the number and
	size of records in each file.  This test will fail if the
	tape contains records of size larger than 64512 bytes.

	File number 1:
		12900 records of size 512 bytes.
		End of file tape mark encountered.

	File number 2:
	Physical end of media has been reached.
	This test is being terminated.

	----------------------------------------------------------
	Wed Sep  1 13:14:36 1999
	Tape exerciser completion.

12900*512 = 645=10240, so the right amount of data is there.  Using gtar on
the same Ultrix box, I get:

	balrog:~ 5# gtar tvf /dev/nrmt1h
	tar: Record size = 1 blocks
	-rwxr-xr-x simonb/admin 2638155 1999-08-31 10:42 netbsd

The time to read the file is awfully slow, backing up the small blocks
size.

I've started looking at tzopen in sys/arch/pmax/dev/tz.c, but thought d
I'ask here in-case anyone knows what's happening.

Simon.