joe@oregon.uoregon.edu (09/25/90)
********************************************************************* * * * BEFORE YOU BUY A THIRD-PARTY HARD DISK FOR YOUR CUBE, READ THIS! * * * ********************************************************************* Because a number of people have mentioned plans to buy a basic NeXT system and then add cheap third party hard disk drives, etc., let me share the following story with you in the hope that you can avoid the hassles I've encountered. SOME BACKGROUND Here at the University of Oregon Computing Center, we have a cube which I use as my personal workstation, and which other users use for Mathematica and TeX over our campus network. Our current configuration is 16 meg of RAM, with a too-full 330 Meg Maxtor (purchased from NeXT with the cube itself). Because we were running low on free space on the 330, and because we had some money at the end of the fiscal year, we decided to buy an external 660 for the cube. When we contacted Oregon State University (where we'd originally bought the cube), we were told they couldn't get us a genuine NeXT 660. Fine. No problem -- a bunch of people on the net suggested buying a third party external, and provided glowing testimonials about how easy they were to connect and install. DISK NUMBER ONE: THE WREN VI FROM IDS We proceeded to call around, and eventually ended up ordering a Wren VI from IDS (in part because this was the system which was recommended by the folks from Brown University who wrote TheFormatter). The price was good, and the people at IDS sounded like they were more than willing to work to make the sale. Fine. This was near the middle of June. The drive arrived, we connected it to the cube, and rebooted. We promptly obtained a series of SCSI boot errors along the line of: >MAXTOR XT-8380S as sd0 at sc0 target 0 lun 0 } Seems to find the > Disk Label: HD1 } internal MAXTOR just > Disk Capacity: 349 MB, Device Block 1024 bytes } fine... but then: >s0 (2,0) ERROR op: 0x0 sd_state:4 scsi status 0x2 >s0 (2,0) ERROR op: 0x12 sd_state:4 scsi status 0x2 >(repeated for (2,1), (2,2), ... (2,7)) Hmmm... Our hardware tech then tried connecting the drive to a Mac II running SCSIProbe and SCSIAnalyzer. It looked just fine to those utilities, and the drive even worked fine on a Sun 3 workstation. We tried several different cables (including a four footer shipped with the drive, and a short Apple SCSI cable), every combination of terminators known to man, different SCSI target numbers, etc. Nothing made a difference. BLIND ALLEY NUMBER ONE: 53c90 SCSI CONTROLLER CHIP After extensive discussions with NeXT, one of the NeXT techs happened to notice that some problems had been reported with the 53c90 SCSI controller chip used on early NeXT motherboards, and sure enough, that's the SCSI controller chip which was on my old motherboard. One trip to Corvallis later, I was proudly in possession of a new 53c90a- based SCSI controller-holding motherboard which STILL couldn't see the external disk. BLIND ALLEY NUMBER TWO: DRIVE TO CUBE CABLING More discussions with NeXT yielded the information that NeXT "pushed the SCSI interface pretty hard," and that you had to use EXACTLY a 100 ohm cable, or echos and such would confuse the SCSI interface and leave you dead in the water. Okay, that sounded reasonable enough. I've seen enough cable-related problems on PCs and such that I was willing to believe this was a definite possibility. Rrrr--nnn--hhh. Wrong. After finally begging a NeXT-blessed cable from NeXT, I was left with precisely the same error I'd had previously. BLIND ALLEY NUMBER THREE: THE DRIVE ITSELF At this point, more discussions were conducted with IDS, the drive vendor. They agreed to do an RMA and let me return the Wren, which they agreed to swap for a Maxtor 8760S which should be "plug and play." Noticing that this was the same drive NeXT itself uses, we were all excited and enthusiastic once more. Fine. Had a fun couple of hours playing with styrofoam popcorn packing the Wren VI up, and a week or so later, in comes the Maxtor 8760s. With great (naive) hope, we connected it up and sure enough -- we hit the same problem we'd been having with the Wren VI. No apparent difference. BACK TO NEXT: More discussions with NeXT ensued. Andreas Sauter sent along information on the cannonical jumper settings for the Maxtor 8760S: > JP14 IN JP38 IN or OUT (start when power is applied) > > all SCSI terminators in place (R-packs must be oriented > properly) > > JP34 OUT JP41 IN (drive provides terminator power to its own > terminators only) > > or > > JP34 IN JP41 IN (drive provides terminator power to itself, > and to the SCSI bus.) > > JP18 OUT (read/write) > > JP40 IN (enabling parity detection) > > Other jumpers (not user selectable!) must be: > > JP10 IN > JP11 IN > JP15 OUT > JP26 OUT > JP32 IN > JP33 IN > JP39 IN > TXD-E19 IN All of my jumpers were correct per that information. Paula Lorenz at NeXT then managed to get me in touch with Jimmie Quan, NeXT's SCSI disk support guru. He deserves that status! According to Jimmie, all Maxtor 8760S's aren't equal. In fact, the critical number you want to note is the TLA (top level assembly) number associated with the on-disk controller board. According to Jimmie, only certain TLA numbers will work with the NeXT -- naturally, my drive's TLA #1098208 A isn't one of them (Jimmie suggested you'd want TLA #1098388 B1, ideally). There is a possibility that Version 2.0 of the System will somehow overcome this limitation, although I cannot confirm this yet personally. According to Jimmie, Maxtor won't change TLAs on drives once they're out the door. SUMMARY: I am sitting here, rather unhappy to say the least, with a Maxtor 8760S that my NeXT will not see. It is a third party disk, nominally for the NeXT (according to IDS), but I have *NOT* been able to get it to be seen by my cube. I *MAY* be able to get it working on 2.0, but I'm sure not going to count on it. If I were you, I wouldn't buy ANY third party disk unless you can conclusively guarantee you won't experience the hassles I've gone through. I have about fifty mail messages from NeXT and other parties trying to get this resolved, and the whole thing has consumed just an immense amount of my time and patience. All I'm saying is "think twice." This is one nightmare I wouldn't wish on my worst enemy. Joe St Sauver (JOE@OREGON.UOREGON.EDU or JOE@OREGON) Statistical Programmer and Consultant University of Oregon Computing Center