🚀 go-pugleaf

In article <Bu5vq5.3wF@iat.holonet.net>, bwilliam@iat.holonet.net (Bill Williams) writes:

I wrote :
|> > These generate a lot of interrupts for each SCSI transaction 
|> 
|> I Believe that NO COMPUTER could efficiently use individual interrupts
|> during the data tranfer phase of a SCSI transaction.

You have misunderstood.  On the 5390 (and other simple SCSI chips) an
interrupt is generated for each SCSI phase change.  Processor intervention
is required as a decision has to be made how to proceed depending on the
state.  The 53700 (et al) effectively replace intelligent host adapators (such
as Adaptec) and run simple software (scripts) which allows them to make these
decisions on their own.  The 5390 typically generates 4 or 5 interrupts per
SCSI operation, 53700 1 or 2 and the 53720 1.  Depending on the size of transfer, typically a block of 4 or 8K in a Unix filesystem (BBC), these interrupts can represent a significant overhead (as well as slowing down the system in general).

|> The NeXT is fine. Quite bellyaching about SCSI Transfer...

Any computer which uses the disk as backing storage can do to have faster 
peripheral I/O, especially when it's as memory hungry as the NeXT.

|> you probably don't have either:

I probably do, but luckily I don't need to connect a NeXT to any of them.

David.