Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!usc!elroy.jpl.nasa.gov!ames!vsi1!zorch!sjsca4!poffen From: poffen@sj.ate.slb.com (Russ Poffenberger) Newsgroups: comp.dcom.modems Subject: Re: Questions on High speed UARTS Message-ID: <1990Nov21.191157.11004@sj.ate.slb.com> Date: 21 Nov 90 19:11:57 GMT References: <4889.274A0A5B@zswamp.fidonet.org> Reply-To: poffen@sj.ate.slb.com (Russ Poffenberger) Organization: Schlumberger Technologies, San Jose, CA. Lines: 40 In article <4889.274A0A5B@zswamp.fidonet.org> root@zswamp.fidonet.org (Geoffrey Welsh) writes: >Robert Nee (nee@cf_su14.Salomon.Com ) wrote: > > >1 I read in PC Magazine that in order to use a high speed > > modem at peak efficiency you must have the 16550 UART > > chip installed in your computer. (not an 8250 or 16450) > > The 16550 simply offers a 16-byte FIFO buffer in stead of a transmit hold >register and receive data register. This can be importent if (a) you're using >an XT which can't process the data quickly enough, and which loses bytes >occasionally because of it (the 16550 stores up to 16 bytes, to be fetched >whenever the XT gets around to it); this usually happens at 19,200 or 38,400 >bps. (b) you're using extended memory under MS-DOS on a 286; interrupts may be >lost when a 286 switches from protected mode to real mode and the 16550 >prevents loss of characters along with it. (c) You're using a Perstor ARLL >controller, whose long DMA cycles can prevent the CPU from fetching a byte >from the UART before the next arrives. > > I'm sure there are more situations in which a 16550 would be helpful, but >these are the main ones I've encountered. > > The bottom line is that the 16550 is ideal for making sure that your system >isn't losing characters. If your system already isn't, then you will never >notice the difference between a 16550 and whatever you replaced with it. > > I always have a hard time convincing people that throughput won't improve >with a 16550. What it boils down to is this: the 16550 helps your CPU keep up >with the data. If your CPU isn't keeping up with the data right now, you'd >KNOW it. > > Also note that I heard that these "features" of the 16550 are TURNED OFF by default at power on, and must be specifically enabled. If you don't have a driver or comm program that can do this, you will not notice any benefit. Russ Poffenberger DOMAIN: poffen@sj.ate.slb.com Schlumberger Technologies UUCP: {uunet,decwrl,amdahl}!sjsca4!poffen 1601 Technology Drive CIS: 72401,276 San Jose, Ca. 95110 (408)437-5254