J.Pearce@cs.ucl.ac.uk (02/20/91)
Further to my previous message to comp.unix.aux I have continued to investigate my problem with my Daystar Fast Cache IIci. I left the Daystar demo program running under A/UX all night with no apparent problems - so I find it difficult to believe there is a hardware fault on the board. I have removed the 8-bit Apple Graphics Card and the additional 4M SIMMs so I have a totally standard Mac IIci, but with no luck. However, I have been able to localise the problem. I logged in using console mode and conducting extensive tests launching Mac environments. The mac24 shell never gave any problems, but the mac32 shell seems to suffer from very erratic problems. Between 5-15% of the launches result in the appearance of an empty debugger/dialog box, with its border being continuously redrawn. You can escape via apple-control-e, or wait a couple of minutes and it returns to the console environment leaving a core dump. Just before the box disappears it turns green or purple. Has anyone any idea what is happening ? John Pearce jpearce@uk.ac.ucl.cs Computer Science Department University College London Gower Street London WC1E 6BT ENGLAND Telephone : +44 71 387 7050 ext 3715 Fax : +44 387 1397
brownrigg@kuhub.cc.ukans.edu (02/22/91)
I have no answers but just wish to re-affirm John's problem. I have a DayStar as well, and have observed quirky behavior on login. A difference is that instead of some dialog box being continually redrawn, my system goes through its normal routine - loading inits, etc, and at about the point where it begins "mounting" disks on the desk top, the screen will clear to the background pattern and then poof!, I'm back to the login screen. This happens very rarely - I have always been able to login in immediately afterwards. As an aside, I've not noticed a dramatic performance increase with the install- ation of this card - not like I'd been lead to believe by discussions of cache-memories with friends and experts. I have a ray tracer I've developed, and expected tremedous improvement there because it is so compute-bound. Nope. Rick Brownrigg