[rescue] Various machines for sale

[Dave McGuire]

On 1/8/25 21:59, Joshua Boyd via rescue wrote:
>> Neither one is parallel, right?  As I understand them, they look like
>> SCSI once you get up to the packetized commands and responses, but
>> nothing like it below that.
>>
>> The major reasons I prefer SCSI amount to "simplicity".  Like USB
>> versus real serial ports: I can interface to a serial port using
>> discretes (though a little SSI helps); I can count the transistors
>> used.  It's really hard to speak USB without thousands-to-millions of
>> transistors.  (Admittedly, USB can do more than a serial port.  This is
>> most relevant when trying to do the simple sort of task that a serial
>> port can serve for.  It's a little bit like firing up a full Lisp
>> engine to add 3 and 4.)
>>
>> In SCSI's case, the protocol is complex enough that the comparison
>> isn't so clear-cut, but the basic pattern I see is more or less the
>> same.  Have I misunderstood?
> 
> Were you counting the transistors used for Ultra-160 or Ultra-320 SCSI?
> 
> If you want speed things are going to get more complicated, and the 
> belief was that at a certain point making serial faster is easier than 
> making parallel faster.  If you don't want more speed, old SCSI is still 
> there.

   And that belief was 100% correct.  Making parallel faster is just as 
easy as making serial faster, but the problem is skew between those 
parallel signals.  Transitions don't arrive at close enough to the same 
time to avoid corruption.

   PCI had the same problem, hence the development of PCIe, which is 
essentially the PCI protocol, serialized.

   The engineering world knew at least a decade or maybe two decades 
prior, that things were going to go this way.  It was no surprise.

             -Dave

-- 
Dave McGuire, AK4HZ
New Kensington, PA