Mylex SCSI Controller, 16550A U

Karl Denninger karl at ddsw1.MCS.COM
Fri Oct 13 01:30:21 AEST 1989


In article <6700027 at adaptex> neese at adaptex.UUCP writes:
>
><<STUFF DELETED >>
>>The WD series also has better ECC capability demonstrated on the
>>bench, it can recover data from marginal drives that the Adaptec will error
....
>>You know which board I'm going to trust my data to given a choice, don't you?
>
>Well, I don't want to start a flame war here, but here are some facts about
>both controllers.  

>1) By default, the 2372 does zero retries to get data from a
>device.  The WD controller does 8 retries no matter what.  The 2372 does
>not do retries unless the system it is running in requests them.  This alone
>will account for the difference in errors.

Unless retries are disabled.  Which our bench software does.  And it does
not use the BIOS (we have that turned OFF in both boards in production
systems, remember, this in a Unix (actually Xenix) machine!)

>2) The data separator has nothing to do with errors that are witnessed at the
>BIOS level.  Errors at the BIOS level are usually due to ECC failure after
>all retries have been exhausted.  If the system does not setup retries, as
>per the AT specs, then an immediate failure will be reported by the 2372
>after only one ECC pass on the defective data.

See above.  The BIOS is not used in a protected mode operating system after
the boot sequence completes.  In our case we have it jumpered out anyway.

>3) If you would like to take our data separator and bench test it off of the
>controller, you will find it meets the specs.  In meeting the specs it does
>a better job than the comparable specs for the WD data separator.

Ok.  So?  This isn't a "your chip is better than my chip" fight.  It is a
test with real data, on real disk drives, under real-world use conditions.

Advertising and technical claims are puffery.  Real-world experience is what
counts, and we have it with both controller boards in >exhaustive< tests.

I have one drive here, the above-mentioned ST4144R, which is somewhat flaky.
It came out of a machine for exactly that reason -- it wasn't reliable.  The
ACB board can handle it, sure, but it shows more errors on our bench test
set than with the equivalent WD card (1006VSR2).  Whether due to retries or
not is not the issue; what IS at issue is the performance the user sees under
real-world conditions.

>I will not challenge the performance differences as I haven't tested them
>out, but as far as the rest of your claims, they are full of hot air.  

It would be wise of you not to challenge the performance claims.  They are
based on simple tests that anyone with a Xenix system can duplicate.  To
whit:

Test on a quiet system (I just performed it again):
	
	dd if=/dev/rxxxx of=/dev/null bs=4k count=1000
	(xxxx is the device name, "r" for raw device)
	SCO Xenix 2.3.2 Operating System

	Adaptec	ACB2372		WD1006V/SR2
real	39.64s			11.38s
user	.20s			.08s
sys	2.56s			2.46s
KB/sec	103.329			359.929

Both systems identical except for controller and drive.  Both drives
formatted at 1:1 interleave.  The Adaptec doesn't have a track cache, 
while the WD1006 does (I can see the RAM chip part number on the WD 
board, it is obvious)

>			Roy Neese
>			Adaptec Central Field Applications Engineer

I'll also note that Adaptec said "tough beans" when we had problems with
random hangs on the controllers (which turned out to be a problem between
the ACB2372s and the ST4144 series drives -- a compatibility issue that has
not arisen with other controllers -- you need to insert J6 on the ST4144Rs 
IF AND ONLY IF you have two drives connected).  They didn't even want to talk
to us and it took several hours to get through to on their technical support
line.  WD, on the other hand, was reached with 1 call the last time I had a
question.

The >only< advantage I have seen with the ACB2372s is that Adaptec was
thoughful enough to provide a "secondary IRQ" jumper block.  WD doesn't do
that, which has been the subject of complaint by our organization.  We'll
see if they respond to it.

If our ACB2372 boards are defective I'll be happy to exchange them with
Adaptec and realize the "greater" performance capabilities of the card.  
So far that offer has not been made.

--
Karl Denninger (karl at ddsw1.MCS.COM, <well-connected>!ddsw1!karl)
Public Access Data Line: [+1 312 566-8911], Voice: [+1 312 566-8910]
Macro Computer Solutions, Inc.		"Quality Solutions at a Fair Price"



More information about the Comp.unix.i386 mailing list