wales@valeria.cs.ucla.edu (Rich Wales) (09/07/90)
I have been experiencing a strange problem with the interleave effi- ciency of various disks and controller cards on my system. My computer is a "baby-AT" -- an 8-MHz 286 motherboard with an 8-bit, 4.77-MHz XT-compatible I/O bus. Up until recently, I was using two 32-meg RLL drives (one by Lapine, one by Kalok). These two drives were originally packaged as hard-drive-card assemblies -- one (4 years old) with an SMS/OMTI 5527 controller, the other (one year old) with a Western Digital WD1004-27X controller. I rearranged things so that both drives were running off of a single controller. With the OMTI controller, I was able to use an interleave of 4:1 on the drives. With the WD, on the other hand, the best I could get away with was a 6:1 interleave. (When I say I was "able" to use such-and-so interleave, of course, what I mean is that this was the best interleave that still gave decent per- formance. I have no problem telling the WD to do a low-level format at 4:1 or 5:1 -- but the resulting throughput is atrocious.) This would have been fine -- except that, a few days ago, I got a new drive (a 65-meg Seagate ST277R-1). I intended to install this drive in place of one of my old 32-meg drives. However, to my chagrin I discov- ered that the OMTI controller doesn't have dynamic configuration capa- bilities and insisted on formatting the new drive with 32-meg parameters (615 cylinders, 4 heads). The WD controller does have dynamic configu- ration, of course -- but, as I said, it could only do a 6:1 interleave. There is a bank of several jumpers on the OMTI card. Perhaps some of them select the drive geometry. But there's no way to tell. The hard drive card manual didn't say what the jumpers do.) I tried calling the company which originally made the hard drive card containing the OMTI controller, to see if they had any ideas. Unfortu- nately, they said Scientific Micro Systems (the people who made the OMTI controller in question) had been bought out by some other company -- they didn't know who -- and that they didn't use OMTI controllers in their products any more anyway. So, I'm left with a 4-year-old controller that gets great performance but doesn't know about 65-meg drives -- and a new controller that can talk to just about any disk, but requires a high interleave and gets significantly lower throughput. I bought a brand-new Western Digital WD1004-27X yesterday, just in case there might be some revision level improvement in the past year. No luck; the new WD controller can also do only 6:1 in my system. Everyone I've described this to (including people at both Western Digi- tal and Seagate) thinks I'm crazy. The interleave, they say, is a function of the computer system, not the controller. If one RLL con- troller can do 4:1 in a given computer, then any RLL controller should be able to do the same. To even =ask= whether a given controller is fast enough to achieve a given interleave is to display a fundamental misunderstanding of the problem. The only thing I can think of is that perhaps my OMTI controller isn't "really" doing true RLL. A disk formatted using the OMTI won't work on the WD (or vice versa) without being reformatted. Is it possible that a 4-year-old "RLL" controller might not be the same thing as a modern "RLL" controller? Any ideas or comments? Does anyone know who bought out SMS or OMTI? Maybe =they= can explain this. -- -- Rich Wales <wales@CS.UCLA.EDU> // UCLA Computer Science Department 3531 Boelter Hall // Los Angeles, CA 90024-1596 // +1 (213) 825-5683 "You must not drink the tea. It is deadly to humans."
wales@valeria.cs.ucla.edu (Rich Wales) (09/08/90)
In article <38667@shemp.CS.UCLA.EDU> I asked for any clues as to why I
could get a 4:1 interleave with a four-year-old OMTI disk controller,
while an almost-new Western Digital controller only got 6:1.
Judging from the answers I got, apparently the explanation is that
different controllers do different things in their ROM BIOS code -- and
if this code takes long enough, it can adversely affect the interleave.
Thus, changing the controller really =can= change the interleave you can
get away with; it isn't purely a function of the processor alone.
I got a program called RAMIT from the SIMTEL20 collection (a new version
was just put in tonight, as PD1:<MSDOS.DSKUTL>RAMIT150.ZIP). RAMIT cop-
ies the disk controller's ROM BIOS into RAM, and patches things so that
the system will use the copy in RAM instead of the ROM code. When using
an 8-bit controller in a 286 system (necessary in my case, since it's a
"baby AT" with an 8-bit XT bus), accessing code from a controller's ROM
can be =much= slower than accessing regular memory. That's why lots of
VGA cards come with software to move the video BIOS into RAM for better
performance.
In my case, RAMIT allowed me to change the interleaving on my disks from
6:1 to 4:1. Throughput (as measured by CORETEST) jumped from 115KB/sec
to 175KB/sec.
As for my 1986-vintage OMTI 5527 controller, the drive table can be
accessed by fiddling with a group of four jumpers. (I found this out
for myself by trial and error; I never did manage to find the successors
to Scientific Micro Systems, the OMTI people.) However, none of the
available settings appears to correspond to the geometry of my Seagate
ST-277R-1 drive (820x6). So, the idea of driving my new disk off the
OMTI is out. But given that I can now use 4:1 interleaving on the WD
controller, I don't really mind any more.
As for why a drive which was low-level formatted on the OMTI wouldn't
work on the WD (or vice versa) without reformatting, I now understand
that this incompatibility between different makes of controller (even
if they are both "RLL") is a normal occurrence, and nothing to be sur-
prised about.
--
-- Rich Wales <wales@CS.UCLA.EDU> // UCLA Computer Science Department
3531 Boelter Hall // Los Angeles, CA 90024-1596 // +1 (213) 825-5683
"You must not drink the tea. It is deadly to humans."