neese@adaptx1.UUCP (06/11/91)
>>>SCSI is easier to integrate and (with some drivers) more reliable. > >> I've had complaints from Windows and Desqview over DMA-intensive >> devices like my scanner and DMA-heavy activities like Fastback >> (backup software). Will this be a problem with bus-mastering >> SCSI devices? > >The only DMA problem I've heard of with the Adaptec board has to do with >operating the floppy and the transfering from/to the scsi-bus at the time; >if the bus on/off time is not set properly, the floppy will starve for data. > >[Roy, if you're listening, chime in - I know you know the realy story, 'cuz > you're the one who told me about it!] >STUFF DELETED< Tim's right. The default bus on time for the 154x is 11 microseconds. This is long enough to cause a floppy data underrun condition when overlapped disk I/O is performed with floppy I/O. The bus on time needs to be set to 7 microseconds to cure this. I wrote SETSCSI just for this purpose. I put it in a batch file before calling FASTBACK and then call SETSCSI with no arguments afterwards to reset the 1542 back to defaults. Cures the problem like so: setscsi -n:7 fastback (or your favorite program) setscsi Roy Neese Adaptec Senior SCSI Applications Engineer UUCP @ neese@adaptex uunet!cs.utexas.edu!utacfd!merch!adaptex!neese
press@venice.SEDD.TRW.COM (Barry Press) (06/13/91)
In article <283400137@adaptx1> neese@adaptx1.UUCP writes: >I wrote SETSCSI just for this purpose. I put it in a batch file before >calling FASTBACK and then call SETSCSI with no arguments afterwards to reset Roy, can you check me on some empirical observations re SETSCSI? 1. It looks like (I disassembled it) it uses spin loops to check for timeout when talking to the adapter. This seems to be a problem on a 25MHz 486, although there may be an interaction with the next issues. 2. It looks like SETSCSI conflicts with ASPI4DOS.SYS -- it simply wouldn't run reliably unless I removed the device driver. I suspect that the driver was fielding an interrupt that SETSCSI wanted. Between the two, I can't find a way to make SETSCSI run in my configuration. Even if I slow the CPU down and patch the timeout loop to run longer, the lack of the device driver is a problem for running Windows, Smartdrive loaded high under DOS 5, etc. Is there another way to solve these problems I've overlooked? Thanks. -- Barry Press Internet: press@venice.sedd.trw.com
herbing@netmbx.UUCP (Uwe Herbing) (06/14/91)
press@venice.SEDD.TRW.COM (Barry Press) writes: >In article <283400137@adaptx1> neese@adaptx1.UUCP writes: >>I wrote SETSCSI just for this purpose. I put it in a batch file before >>calling FASTBACK and then call SETSCSI with no arguments afterwards to reset [stuff deleted] >2. It looks like SETSCSI conflicts with ASPI4DOS.SYS -- it simply wouldn't >run reliably unless I removed the device driver. I suspect that the driver >was fielding an interrupt that SETSCSI wanted. [stuff deleted] >Is there another way to solve these problems I've overlooked? According to my manual for ASPI4DOS Version 2.1 Rev B, you can give the ASPI4DOS.SYS driver a /N switch to set the bus on time, e.g. DEVICE=ASPI4DOS.SYS /N7 should solve your problem. Greetings Kai -- ------------------------------------------------------------------------------- Kai-Uwe Herbing | phone 49-30-3126145 | I don't claim any Weimarer Strasse 28 | | warranty for tips D-1000 Berlin 12 | herbing@netmbx | tricks, hints and Fed.Rep.of Germany | herbing%netmbx@db0tui6.bitnet | suggestions given -------------------------------------------------------------------------------
neese@adaptx1.UUCP (06/14/91)
>/* ---------- "Re: <12788@xstor.com>" ---------- */ >In article <283400137@adaptx1> neese@adaptx1.UUCP writes: >>I wrote SETSCSI just for this purpose. I put it in a batch file before >>calling FASTBACK and then call SETSCSI with no arguments afterwards to reset > >Roy, can you check me on some empirical observations re SETSCSI? > >1. It looks like (I disassembled it) it uses spin loops to check for timeout >when talking to the adapter. This seems to be a problem on a 25MHz 486, >although there may be an interaction with the next issues. > >2. It looks like SETSCSI conflicts with ASPI4DOS.SYS -- it simply wouldn't >run reliably unless I removed the device driver. I suspect that the driver >was fielding an interrupt that SETSCSI wanted. > >Between the two, I can't find a way to make SETSCSI run in my configuration. >Even if I slow the CPU down and patch the timeout loop to run longer, the >lack of the device driver is a problem for running Windows, Smartdrive loaded >high under DOS 5, etc. > >Is there another way to solve these problems I've overlooked? Your observations are correct. SETSCSI was written quite some time ago (4 years). In my current reincarnation of all of my programs, I am allowing them to use the ASPI interface as an option, but there is no way for SETSCSI to control the hardware (bus on/off) through ASPI. ASPI is a generic SCSI command interface. However, not all is lost. If you are using the ASPI manager version 2.0 or later (and it may apply to the older versions as well - I am not a DOS expert), you can use the command line flag to set the bus on time to 7 micro-seconds (/N7). Granted, that will set it that way always, but there is virtually no detectable performance degradation, unless you are using a very high performance SCSI drive. SETSCSI is being re-written to use timers instead of loops, but will not use the ASPI interface. It is only useable if you do not have the need to use the ASPI4DOS driver. Roy Neese Adaptec Senior SCSI Applications Engineer UUCP @ neese@adaptex uunet!cs.utexas.edu!utacfd!merch!adaptex!neese
press@venice.SEDD.TRW.COM (Barry Press) (06/14/91)
In article <2400@netmbx.UUCP> herbing@netmbx.UUCP (Uwe Herbing) writes: >According to my manual for ASPI4DOS Version 2.1 Rev B, you can give the >ASPI4DOS.SYS driver a /N switch to set the bus on time, e.g. >DEVICE=ASPI4DOS.SYS /N7 should solve your problem. Yup. But the bus on time (per Roy's messages) is only a problem when doing overlapped I/O with both the hard disk and the floppy -- i.e., when doing a backup. On the assumption that the disk transfers are faster with N:11, I'd like to only slow things down for the duration of the backup. Hence my query. Side note: Wouldn't it be nice if there were a place to get the Adaptec driver updates? Northgate is still shipping ASPI4DOS v 2.0; I managed to wangle 2.1 (a beta copy, though), and apparently (from your message), there is a revision to the official 2.1 release. -- Barry Press Internet: press@venice.sedd.trw.com
neese@adaptx1.UUCP (06/16/91)
>STUFF DELETED< >Side note: Wouldn't it be nice if there were a place to get the Adaptec >driver updates? Northgate is still shipping ASPI4DOS v 2.0; I managed to >wangle 2.1 (a beta copy, though), and apparently (from your message), there >is a revision to the official 2.1 release. Okay guys and gals, here's the deal. If you will mail me a message that contains (verbatim) the information the ASPI4DOS driver displays at run time (which can be very lengthy), I will E-mail you the latest ASPI4DOS driver. The current driver is V2.1. Expect a slight lag time on this as I will be on the road the next 3-4 weeks. I don't mind upgrading someone who has an official copy of the software. Roy Neese Adaptec Senior SCSI Applications Engineer UUCP @ neese@adaptex uunet!cs.utexas.edu!utacfd!merch!adaptex!neese
feustel@netcom.COM (David Feustel) (06/19/91)
Of course, all of us suckers who bought the Adaptec board not knowing the drivers didn't come with it and who can't buy the drivers now are just out of luck. Serves us right, I guess. -- David Feustel, 1930 Curdes Ave, Fort Wayne, IN 46805, (219) 482-9631 EMAIL: feustel@netcom.com or feustel@cvax.ipfw.indiana.edu
neese@adaptx1.UUCP (06/20/91)
>/* ---------- "Re: <12788@xstor.com>" ---------- */ >Of course, all of us suckers who bought the Adaptec board >not knowing the drivers didn't come with it and who can't buy the >drivers now are just out of luck. Serves us right, I guess. Our adapters can be had either way. There are kits and there are bare boards. The official distributors also sell the software separately. If you want I can post the offical list of distributors and 800 numbers for them. About the only time I would buy a bare board would be; for use under some UNIX, for development purposes, or if you only intend to use the adapter under DOS without any virtual memory utilities/programs and only want to have 2 disk drives. Roy Neese Adaptec Senior SCSI Applications Engineer UUCP @ neese@adaptex uunet!cs.utexas.edu!utacfd!merch!adaptex!neese
press@venice.SEDD.TRW.COM (Barry Press) (06/24/91)
In article <283400155@adaptx1> neese@adaptx1.UUCP writes: >I can post the offical list of distributors and 800 numbers for them. Yes! Especially if they can answer the odd question like why error correction might need to be disabled or not, etc, as well as provide updates. -- Barry Press Internet: press@venice.sedd.trw.com