CompuServe Thread

#3.1 & syncronous SCSI

7 messages in this thread
#185302From: Rafael VicariaFeb 12, 1995 3:40 PM
Hello all! I installed wb3.1 yesterday, and I had only a slight problem. BattMem doc says that "All Quantums support syncronous SCSI transfers" I have a LP52S and a LPS240, which is the boot disk. No problem with the 128ms timeout, but if I set the "Initiate syncronous SCSI on, my A3000 will not boot. Do I have to remove the "EP" jumper? Do I have to do something else? Many thanks in advance. Rafa <Motorola Inside> from Berlin. >…olive oil is the way to go.<
#185606From: Steven D. KapplinFeb 15, 1995 8:40 PM
Rafa, The EP jumper is a parity jumper on the drive, if I remember my Quantum booklet correctly. It should be turned OFF or left alone, but will not have any impact on synchronous transfers. Apparently, you can turn that on, but it doesn't appear to work with the 240. However, you probably prefer to leave it off since synchronous transfer locks out any other device on the buss while the drive transfers data. Although they say transfer rate is faster to the drive, overall it should be slower than using asynchronous. I played with that on my 3000T and frankly found that using synch resulted in problems with the drive booting, no noticeable increase in speed, and overall slowdown in buss reponse with the other devices I had connected. IMHO, leave it alone.
#185647From: Rafael VicariaFeb 16, 1995 9:30 AM
Thanks for your help, Steven. Then, what's the advantage of syncronous SCSI? Rafa <Motorola Inside> from Berlin. >…olive oil is the way to go.<
#185773From: Steven D. KapplinFeb 17, 1995 4:05 PM
Rafa, Really, none, as best I can tell unless you have a system with the single drive which can be written to synchronously by the OS without interrupt. When the drive is operated synchoronously, data can only pass when the drive has signalled for it, thus, it won't be interrupted and can take the data stream at its maximum transfer rate. However, in the Amiga, it appears, based on what I found on my system, that the multi-tasking nature of the system is such that if the drive can take over the datat stream, then any other data transfer to that drive or other devices on the scsi buss must wait. I think it would be comparable to using synchronous transfer during backup to a tape drive where the tape gets to write the data, but the backup program will be locked out of reading data into a buffer during the tape writes. If the tape runs asynch, then while data is written to the tape, the backup program can read data from the hard drive. That's about the only analogy I can make that might explain the difference, Rafa.
#186112From: Rafael VicariaFeb 20, 1995 1:57 PM
Thanks for your comprehensive reply, Steven. I have 2 scsi drives, so syncronous scsi seems to me a silly invention. Never mind. I would really like to help you if I could. Warmest regards Rafa <Motorola Inside> from Berlin. >…olive oil is the way to go.<
#185792From: Tom JonesFeb 17, 1995 6:43 PM
One of my friends had the same problems with his 3000 and syncronous HD's under 3.1, seems to be a bad thing. The slowing down of the buss is really strange however, sync should be faster. alternate: jonest@hfsi.com I never met a man I didn't like…Will Rogers
#186240From: Steven D. KapplinFeb 21, 1995 2:50 PM
Tom, Well, as I understand it it's not that the buss is slowed nor that the transfer rate is slower, but rather that the multi-tasking nature of the OS means that if you do a synched transfer to one device, nothing can happen on any other scsi device until that transfer is done. You get maximum transfer rate to each device, but only one device at a time. That's if I understand the idea of synchronous buss transfers. Given the multi-tasking nature of the AmigaOS, there isn't much to be gained by using synchronous transfers because when one device is in use the others are locked out. I suspect that may be one reason for having boot problems, especially if there's more than one scsi drive on the buss or if the boot drive is partitioned. Anyway, I tried using synched and asynched transfers on one of my non-boot drives and measured transfer rates with Diskspeed 4.1 and didn't find any noticeable difference. Now, part of the reason for that is likely because the drive itself contains a 256K read/write buffer and read/write buffering is on. I also run a separate disk caching program. With all that I think whatever advantages synched transfers might provide are washed away. At least that was my experience with one drive, one time.