#VTR SEARCH CONTINUES!
19-Dec-93 19:45:45
Sb: #72509-#VTR SEARCH CONTINUES!
Fm: Gus Grubba 70673,1605
To: John Tissavary 71054,15
What you're missing is that there is no bus involved when it comes to the PAR
unit. The hard disk is connected directly to the PAR hardware. The disk is
dedicated to the PAR and has its own format (it's not a DOS disk and it is not
connected to any "controller" in your computer). Other than that, everything
you said is right. The MAX has a greater throughput but because it uses SCSI
instead if IDE. It's got nothing to do with EISA or otherwise. Yes, because
SCSI-II disks are a bit faster than IDE, the MAX can use less compression but
the difference isn't that much though. At least not nearly proportional to the
added cost. You should take a very close look before committing the cash to a
MAX system (unless, of course, you have your mind set and don't care).
By the way, you can't possibly be getting a steady 10MB/sec out of your SCSI
disk. You controller (and your BUS!) cannot sustain that. You're more likely in
the 2mb/sec neighborhood at the most. If you want, I can upload a simple test
program that will tell you exactly what your transfer rate is.
The figures you see in the literatures are only true in "optimum" cases and for
very specific amounts of data. Once you start reading a 30 megabyte file, even
if all sectors are contiguous, the rate will drop dramatically. The 10mb figure
is only true for data already loaded into the controller's memory. That's the
absolute maximum rate the controller will pump _that_ block of data to the
computer. As soon as the block spans over another DMA read (just 64k), or
worse, you have to seek the disk for another sector, you're way down. The
30mb/sec figure (burst mode) is only true for very tiny amounts of data. It
can't be even considered in this context.
This gets even more complicated if you really care to know the details. Another
note worth taking is that there are two sets of "rates". One thing is the rate
the controller can sustain between the hard disk and itself. The other rate is
between the controller and the computer. In your EISA system, I can guarantee
you your hard disk spends a lot of time waiting for the controller to transfer
data to the computer.
Another note, EISA is doomed. There us no reason whatsoever to keep it alive in
a VLB and PCI world. Both are orders of magnitude better, cheaper and by far,
more attractive to hardware manufacturers.