[3790] (54 lines) Spitzer.Multics 11/29/84 0914.2 mst Thu INFO-MICRO Subject: IBM PC AT disk problems Date: Wednesday, 28 November 1984 19:37 est From: "John Levine, P.O.Box 349, Cambridge MA 02238-0349 (617-494-1400)" To: info-micro at BRL cc: somers at RU-BLUE [In response to several queries, here is the text of my AT comments, somewhat updated and extended.] It appears that the PC AT has severe disk reliability problems, and that IBM is not planning to do anything about it soon. We have six PC ATs, and five of them have developed bad spots on their disks. We tried reformatting them with the regular FORMAT program. No dice; it didn't find any more bad spots than it had previously. I talked to my sister, who works for a big company with access to IBM's PC problem data base, so she asked them. Their answer seemed to be that yes, sometimes AT disks flake out, and that reformatting them with the advanced diagnostic formatter usually fixes it. Great, the advanced diagnostics cost $295. My (non-IBM) repairman was kind enough to lend me his formatter, and I reformatted my disk, which included about 5 minutes of surface analysis. Still no dice -- it still didn't find any more bad spots. Today I talked to a friend who is working on some programs that drive the AT's disk with interrupts, using code completely unrelated to the BIOS. Sad to say, he finds the same sorts of problems, suggesting that the problem is in the controller, not BIOS bugs. He noted that the behavior is as though the disk occasionally writes a bad sector, but that when you rewrite the bad sector it is OK. He further noticed that such failures most often occur around cylinder 300. The disk controller has a "write precompensation register" which turns on the "reduced write current" line, and as it happens cylinder 300 is the first one where the reduced current line is turned on. Hmmn. Perhaps it's reduced a little too much. If you notice the same failure, run chkdsk and see if your disk is 1/2 full, i.e. about 10MB used and 10MB free. If so, you may well be seeing the same problem. The cylinder 300 parameter is set in the BIOS, and you can change it -- it might be interesting to try increasing it a little. The AT has other strangenesses, too. About half the time when I hit ctrl/break, the system hangs up and I have to power cycle to reboot. It appears to be halted or hung, because the keyboard indicator lights, which are updated at interrupt time, decline to change. Also, it sometimes gets into a rather inconvenient state where it is doing eternal print screens, wasting paper very noisily. Again, the only fix seems to be power cycling. Pfui. John Levine, Javelin Software, Cambridge MA 617-494-1400 uucp: { decvax!cca | yale | bbncca | allegra | cbosgd | ihnp4 }!ima!johnl Internet: Levine@YALE.ARPA 7