Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!cbmvax!daveh From: daveh@cbmvax.commodore.com (Dave Haynie) Newsgroups: comp.sys.amiga.datacomm Subject: Re: 19200 baud amiga Message-ID: <19116@cbmvax.commodore.com> Date: 19 Feb 91 20:22:59 GMT References: <2587@tmiuv0.uucp> <978@faatcrl.UUCP> Reply-To: daveh@cbmvax.commodore.com (Dave Haynie) Distribution: comp Organization: Commodore, West Chester, PA Lines: 68 In article <978@faatcrl.UUCP> jprad@faatcrl.UUCP (Jack Radigan) writes: >rick@tmiuv0.uucp writes: >>The Amiga's serial hardware has problems at bauds greater than 9600, I >>think. The 2000/2500's does. I believe the CBMSpeak on it is, "It is >>not reliable." I have ASDG's DSB board in my A2500/30. It works fine >>at 19,200. I've used it in my 3000, too, but I've not tried anything >>faster than 9600 (betwixt Amoeba III and XyClone ). > The official blurb in the RKM on serial I/O is that anything above >19.2kbps is "optimistic". In actuality, any Amiga can handle serial >output. It's serial input that gets stickey depending on the confiuration >of the system being used. The real problem with high speed serial input is buffering. For every character that comes in, you need to interrupt the CPU and have it stuff that character somewhere. 68030 machines work faster than 68000 machines in all things, interrupts included, so they tend to deal nicely with high speed transfers. The terminal program you're using will also have an effect on your maximum practical speed; it needs to know the right way to talk to serial.device, especially keeping in mind that more than one character may be available on every I/O request. And the Amiga is hardly the only one with the problem -- I used to use our VAX system here at 19,200 baud from the A2500 in my office, but dropped down to 9600 as IT kept losing stuff. The Amiga was not the problem; with the A2232 card in the system, I've had two 19,200 baud channels and one 1200 baud channel running at the same time (using ATalk-III, two direct lines, and one modem). The A2232 does make life easier, since it does its own buffering, but it can't go beyond 19,200 baud. On the other end of things, if you're running a packet oriented protocol, rather than single characters, you can get a reliable Amiga-Amiga connection going in the 150KB-200KB range. > - Certain types of HD controller can also lead to data loss, those that > are hogging the bus for DMA transfers or have device drivers that run > at excessively high priorities are the worst offenders. In general, the DMA devices won't be a problem. Even on the A2000/A2091, a typical SCSI device is going to run at about 1/2 the speed of the bus, so DMA transfes wind up in rather small packets. You may find the DMA device gets on the bus for 20-40 cycles, but probably not much longer at one shot. Any DMA device that takes the bus too long can cause problems, since when the DMA device has the bus, the CPU doesn't. Though it would be rather unnatural for a DMA device to take over the bus for long periods of time, unless for some reason it's doing large block buffering. All C= hard disk controllers use FIFOs, which buffer up maybe 16 or 32 bytes at a time. The programmed I/O controller I call "parasitic" controllers, which read or write a SCSI chip as if it were slow memory (therefore running cycles with considerable wait states, thus the term parasitic), are very deadly in combination with hard disk performance. In general, any time one of these is talking to the disk, all CPU time is eaten up. > As for the A3000, I've been able to drive it at 38.4kbps without a burp. You get a big pile of advantages in the A3000 -- faster CPU, faster memory, and especially the hard disk, which typically takes 4%-8% of your CPU time for transfers, rather than as much as 50% on A2000 based DMA devices. > -jack- -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "What works for me might work for you" -Jimmy Buffett