I have done restores on NT5 from a similar config, and had NO problems.

I agree that autonegotiate is a likely culprit when the throughput is that
slow.  When autonegotiate is the problem, our network guys can see errors
being recorded on the switch; ask for help there.

Another possibility is that the bottleneck is on the NT machine itself,
trying to rebuild the file system.
If you have virus checking software installed, disable it for the time
being.
If this rebuild is being done due to a failure of some sort, did anyone run
CHECKDISK on the drive before starting the restore?

Run from the TSM server command line:  q db f=d
Your cache hit % should ideally be 98% or better; 96% or better works pretty
good, too.
If it's lower than 95%, that may be hurting you some (but not to the tune of
20 hours...)

As a last resort, check your TSM server's AIX errpt log, to check for disk
errors, and run IOSTAT to see if you are getting any bottlenecks on your TSM
DB I/O.