Manuel Bouyer wrote:
> On Thu, Apr 03, 2003 at 12:45:59PM +0300, Jarkko Teppo wrote:
> 
>>>It was suggested at the time that the PSU was overloaded, but I do not now
>>>believe that to be the case:  even if the 133 protocol draws significantly
>>>more power than UDMA 66 or 100 (I can't speculate on that), I have since
>>>added a functional 1394 card to the PCI bus as well.
>>>
>>I used to run this machine (AS 600) with four internal Barracudas,
>>three SCSI-buses (ISP + AHA-2940) and three network cards. Now it has
>>a 9GB IBM SCSI + 60GB Maxtor and relevant controllers. I don't think
>>it's the PSU.
>>
>>Is there something I could do to help to debug this ?
> 
> Maybe it's the drive ? Do you have a way to try another drive ?
> Also did you try different PCI slots ?
> If possible, test the adapter+cable+drive configuration in another
> kind of machine (ideally a PC, this is where this hardware is the
> most used)

In my own case I might point to a lurking bug in the driver or a hardware 
limitation:  same drives, same cables, same slots:  Promise UDMA-66 works, 
SIIG UDMA-66 doesn't work, and Promise UDMA-133 doesn't work (where 
'doesn't work' == the same lost interrupt errors Jarkko is having).

I would ask:  1.  Is there any way you can get your hands on an older 
UDMA-66 or -100 card, and 2. is it a Miata or Miata-GL?  It would also be 
useful to hear from other NetBSD Miata / Miata-GL owners who use IDE cards. 
  Both of us.  ;^)

--Dave