Date:        Thu,  7 Oct 1999 09:06:40 +1000 (EST)
    From:        Daniel Carosone <dan@geek.com.au>
    Message-ID:  <14331.54309.765384.180345@passion.geek.com.au>

  | Please do try -current and see if it gets slower.

Doesn't seem to have.   I installed from the 19990926 snapshot:

NetBSD  1.4K NetBSD 1.4K (GENERIC) #7: Sun Sep 26 21:44:47 NZST 1999     dan@anarres:/home/NetBSD/src/sys/arch/alpha/compile/GENERIC alpha

With that I get ...

cd2: 652279808 bytes transferred in 121 secs (5390742 bytes/sec)  (ide ASUS 50x)
cd0: 652279808 bytes transferred in 181 secs (3603755 bytes/sec)  (scsi TEAC 32x)

and ...

cd3: 398655488 bytes transferred in 109 secs (3657389 bytes/sec)   (ide Delta 44x)

Those are all with the same disc (also the same one I used the other day)...
The lower number of bytes from cd3 are because the transfer terminated with ...

cd3(pciide0:1:1):  Check Condition on CDB: 0x28 00 00 02 f8 00 00 00 10 00
    SENSE KEY:  Illegal Request
     ASC/ASCQ:  Illegal Mode For This Track

which doesn't make a lot of sense to me (and suggests some kind of driver
bug I think, but I don't know anything about ide/atapi drivers).

All these are from dd if=/dev/rcd${N}a of=/dev/null bs=32k

These tests were done in single user mode using the GENERIC kernel (I have just
finished the install - didn't even get to set a root passwd yet...) unlike the
ones I reported the other day, which were done multi-user using a tailored
kernel.

kre

ps: this new kernel keeps bitching ...

isp0: skipping update of target 8 on bus 0
isp0: skipping update of target 9 on bus 0
[...]
isp0: skipping update of target 15 on bus 0

when a scsi device is accessed, and various other times.   I believe my
qlogic controller (1020) is 100% narrow - target 8 (etc) cannot exist (though
the previous kernel also reported 16 targets (on scsibus0 attached to isp0),
It didn't used to make all this noise however.)