iverson@bang.uucp (Tim Iverson) (02/09/91)
In article <1459@tsdiag.ccur.com> jeffp@tsdiag.ccur.com (Jeff Pelosi) writes: > I am considering purchasing a Scsi controller. I would appreciate > getting some recommendations. I know one person who has an Always > IN-2000. Last I saw an Always board (6 months ago?), they were engaging in some very shady practices - both engineering and marketing. They advertised that their board had a phenomenal transfer rate. Well, I checked it under DOS with CoreTest and a wimpy little seagate capable of at most 200KB/s. Surprise, CoreTest came back with 1.2MB/s! This is flatly impossible, they must have been doing something fishy. So, I dusted off soft-ice and observed their BIOS firsthand. Aha! Caught in the act - they disable *all* interrupts while transferring data. This means the clock doesn't tick, which means (to CoreTest) that transfers take no time at all. It also means that your serial port can loose data while writing to disk (like in an old PC jr.), and that other programs that depend on clock ticks or other asynchronous interrupts won't work correctly, if at all. Now, unless they have some *very* dumb hardware, there is no reason to do this other than to cause an abnormally high transfer rate. It is obviously not a programming error - interrupts are enabled up until the transfer, where they do a deliberate cli, followed by the transfer, followed by sti. > I like the price point of this controller ($189.00 at Computability). You're kidding. An Adaptec 1542B can't cost much more than this and I know Future Domain TMC830s are much cheaper - a better card for DOS (especially with windows) than the Adaptec. For unix, you will not be able to live without a 1542, not unless you really know what you're doing. >Jeffrey Pelosi Concurrent Computer Co. Fax:201-870-5952 >jeffp@tsdiag.ccur.com 2 Cresent Place MS117 Phone:201-870-4223 - Tim Iverson iverson@xstor.com -/- uunet!xstor!iverson BTW, now you'll be able to tell your friend why her time-of-day clock is so slow all the time :-}.
mike@cimcor.mn.org (Michael Grenier) (02/11/91)
From article <1991Feb09.052143.1873@bang.uucp>, by iverson@bang.uucp (Tim Iverson): > board had a phenomenal transfer rate. Well, I checked it under DOS with > CoreTest and a wimpy little seagate capable of at most 200KB/s. > > Surprise, CoreTest came back with 1.2MB/s! This is flatly impossible, they > must have been doing something fishy. We've also acheived those rates under UNIX. (Strange how this discussion keeps reappearing...) > So, I dusted off soft-ice and > observed their BIOS firsthand. Aha! Caught in the act - they disable > *all* interrupts while transferring data. This means the clock doesn't tick, > which means (to CoreTest) that transfers take no time at all. > That's not the case under UNIX. > For unix, you will not be able to live > without a 1542, not unless you really know what you're doing. BullSh*t -Mike Grenier mike@cimcor.mn.org
bchin@umd5.umd.edu (Bill Chin) (04/09/91)
I'm looking for recommendations on a SCSI controller for a 386 ISA bus clone. Needs to work with both DOS/Windows and OS/2 (but not UNIX), 16 bit ISA bus interface, and can work in conjunction with an Adaptec HD/FD MFM controller. Any suggestions? Don't need on-board cache, or any really fancy features. Just fast, compatible, and cheap. Currently considering the Adaptec 1542B, and boards from DTP, WD, Future Domain, and DTC. Thanks for your help! -- Bill Chin internet: bchin@umd5.umd.edu MS-Windows Programmer NeXTmail: bchin@is-next.umd.edu PCIP, Computer Science Center CompuServe: 74130,2714 University of Maryland, College Park *Standard Disclaimers Apply*