uhclem@trsvax.UUCP (09/14/89)
<> R4>Why did the 1024 limit start? Well, the MSDOS defined partition table R4>layout only allows 10bit cylinder numbers. So I could see if the Speaking as someone who has been working with the Western Digital controllers for at least a year before IBM even knew what one was, the 1024 limit can be placed squarely on the original controller, the WD1000, which ran as a simulation of the eventual chip using an 8x300 microprocessor. These were actually sold! A lot of them! The 1010 controller that later reduced this 11" x 8" board down to half that size, but kept the 1024 limit. No problem at the time, as the first 70 meg drive (1024 x 8) did not arrive on the scene until 1984. Seeing the limit, Western Digital did produce the 2010, which increased the cylinder count by one bit (now 2048) cylinders. (The 2010 is compatible in every other way with the 1010.) My company received early 2010 samples (spring '84 I think), which were a bit buggy and since the 70 Meg drive seemed to satisfy the more storage-hungry customers, we elected not to use it. The price jump for that extra bit (around $10 more) did not seem reasonable at the time, particularly since no significant number of drives that had over 1024 cylinders were even sampling. Of course, our systems allowed 4 hard drives on one controller, so the customer had more upgrade path than 2-drive PCs would allow. I suspect that when IBM went to Western Digital to have a disk controller adapter developed, WD selected the less-costly and certainly stable 1010. Nowdays, newer WD adapter cards like the WA2 have a 2010 controller on them and it has now become the short-sightedness of the BIOS and O/S writers for not paying attention to the fact that the 2010 was out there and that eventually there could even be a 3010. It was pretty obvious what WD would do to expand. There were certainly enough indications to have made the original AT code so that it would handle the 2010 as an upgrade. Here is the cylinder registers on a 1010: 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 +------------------------+ +------------------------+ 1010 Cntrlr | x x x x x x 9 8 | | 7 6 5 4 3 2 1 0 | +------------------------+ +------------------------+ Cylinder MSB Cylinder LSB x = reserved for future use And on the 2010, available in a stable form since early 1985: 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 +------------------------+ +------------------------+ 2010 Cntrlr | x x x x x 10 9 8 | | 7 6 5 4 3 2 1 0 | +------------------------+ +------------------------+ Cylinder MSB Cylinder LSB Hmm. Wonder how they can possibly expand the number of cylinders the controller supports? :-( Well, someone at IBM/Microsoft (whoever was irresponsible) didn't see the connection. If you could sell an incompatible computer these days, this problem would have been solved the first day. But, no, we are stuck with the mistakes made in 1981 and a few new ones in 1985. Oh, well, so much for innovation. <My opinion, and not that of my Golden Retriever who is delivering papers so he can buy a NeXT, Mac, Amiga, anything that isn't based on a design from 1981.> "Thank you, Uh Clem." Frank Durda IV @ <trsvax!uhclem> ...decvax!microsoft!trsvax!uhclem ...hal6000!trsvax!uhclem
neese@adaptex.UUCP (09/15/89)
I have been heitating saying anything, but I have always suggested to people using our controllers to use the physical parameters instead of translation mode, as it reduces the overhead for disk I/O. I know of many people using our controllers that have >10224 cylinders and have had no troubles. But this has been with SCO 2.3.x. Roy Neese Adaptec Central Field Applications Engineer UUCP @ {texbell,attctc}!cpe!adaptex!neese merch!adaptex!neese
larry@nstar.UUCP (Larry Snyder) (09/16/89)
> I have been heitating saying anything, but I have always suggested to > people using our controllers to use the physical parameters instead > of translation mode, as it reduces the overhead for disk I/O. I know > of many people using our controllers that have >10224 cylinders and > have had no troubles. But this has been with SCO 2.3.x. I am running a 2372B with two 1024 * 8 head drives (Micropolis and Miniscribe) with the drives set up as type 1. I have an option in the AMI BIOS (drive type 47) which is user definable (heads, cyl, precomp, etc..) and when disabling the firmware on the controller and using the drives as drive type 47 I have noticed no increase in IO - am I missing something? -- Larry Snyder SCO Xenix 2.3.2 '386 uucp: iuvax!ndcheg!ndmath!nstar!larry Computone Intelliport AT8 The Northern Star Usenet Distribution Site HST / PEP / V.22 Notre Dame, Indiana USA Home of the fighting Irish!
richard@neabbs.UUCP (RICHARD RONTELTAP) (09/18/89)
[ WD story on cylinder limit ] Wow, so when 10 bits wasn't enough WD gave us one whole extra bit. How generous and visionary! Maybe I can't fully grasp the technical details but is it too dificult or expensive to use 16 bits? Or maybe I'm just beeing to optimistic. When 11 bits ain't enough anymore, WD can sell us a brand new controller with an entirely new, flashing, hot 12th bit! I guess this is the way it was, is, and will be, causing users a lot of headaches. Richard (...!hp4nl!neabbs!richard)
neese@adaptex.UUCP (09/18/89)
I should have been more specific. When disabling the translation in our firmware, on our ESDI controllers, it will improve performance. The RLL controllers, which have the same overhead as our ESDI controllers when it comes to the translation code, will not gain a noticable amount of throughput as the data rates from the drive are less than the ESDI drives so the controller has more time to do the translation anyway. Roy Neese Adaptec Central Field Applications Engineer UUCP @ {texbell,attctc}!cpe!adaptex!neese merch!adaptex!neese