Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!sdd.hp.com!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!ira.uka.de!smurf!gopnbg!tmpmbx!geminix!gemini From: gemini@geminix.mbx.sub.org (Uwe Doering) Newsgroups: comp.unix.i386 Subject: Re: ESIX- Additional Serial Ports? Keywords: ESIX Serial Dumb-cheap-AT-serial-cards? Message-ID: Date: 16 Aug 90 18:34:28 GMT References: <569@eds1.UUCP> <538@everex.UUCP> <1990Aug11.115912.10530@pegasus.com> <2011@jwt.UUCP> Organization: Private UNIX Site Lines: 31 john@jwt.UUCP (John Temples) writes: >In article <1990Aug11.115912.10530@pegasus.com> richard@pegasus.com (Richard Foulk) writes: >>> [...] BTW if you have Rev D of ESIX it will >>>make use of the 16550 FIFO chip if you have it. >>How does it compare to the FAS driver? > >I can't say how it compares with the FAS driver, but I did a quick >benchmark and found that the ESIX driver runs *much* faster with a >16550. I just installed a DigiBoard DigiCHANNEL PC/4 board on my >ESIX system. This is a dumb four-port card with socketed UARTs. >It's directly supported by ESIX's asy driver. I replaced one of the >16450's with a 16550A. I connected a Wyse 50 terminal to the 16450 >port, logged in at 38400 baud, and did 'cat /etc/termcap'. u386mon >running on the console, with the system otherwise idle, showed 40% to >45% CPU utilization on my 33 MHz 386. Doing the same test on the 16550 >port showed 6% to 10% CPU utilization. I think I'll be buying some >more 16550's... I tried this with FAS 2.07 beta (soon to be released) and NS16550A chips (actually on ISC 386/ix, but this doesn't matter in this case) and got 4% to 5%. But note that I used a 25 MHz cache board. With 33 MHz the result will be even lower. Therefor, still no way to avoid FAS. ;-) Uwe -- Uwe Doering | USA : gemini@geminix.mbx.sub.org Berlin | World : gemini%geminix@tmpmbx.UUCP West Germany | Bangpath : ...!{backbone}!tmpmbx!geminix!gemini