perry@madnix.UUCP (Perry Kivolowitz) (01/30/89)
There was some discussion a little while ago how any decent self-respecting multiserial implementation ought to have a dedicated on-card processor in order to acheive any reasonable performance. Well, we just finished timing a rather stressful transfer on our Twin-X based multiple serial port solution. Stressful in that the baud rate was set to 57600 and both parties to the transfer were based on the same Twin-X board (of course, on the same A2000). To be honest, we weren't expecting much. However, we were pleasantly surprised to acheive the following results: On a total data rate of 115.2KB we acheived an observed character transfer rate of 3690 characters per second (one way) which translates to 7380 characters per second since both sides of the conversation were on the same machine (or about 73KB). In truth, we are pleased with the performance of our ``dumb'' serial ports. -- Perry Kivolowitz, ASDG Inc. ARPA: madnix!perry@cs.wisc.edu {uunet|ncoast}!marque! UUCP: {harvard|rutgers|ucbvax}!uwvax!astroatc!nicmad!madnix!perry
scotty@ziggy.UUCP (Scott Drysdale) (01/31/89)
In article <437@madnix.UUCP> perry@madnix.UUCP (Perry Kivolowitz) writes: >There was some discussion a little while ago how any decent self-respecting >multiserial implementation ought to have a dedicated on-card processor in >order to acheive any reasonable performance. > >Well, we just finished timing a rather stressful transfer on our Twin-X >based multiple serial port solution. Stressful in that the baud rate was >set to 57600 and both parties to the transfer were based on the same Twin-X >board (of course, on the same A2000). > >To be honest, we weren't expecting much. However, we were pleasantly >surprised to acheive the following results: On a total data rate of 115.2KB >we acheived an observed character transfer rate of 3690 characters per >second (one way) which translates to 7380 characters per second since both >sides of the conversation were on the same machine (or about 73KB). > >In truth, we are pleased with the performance of our ``dumb'' serial ports. it's not clear here if both ports were receiving and transmitting simultaneously, and at what actual rates. my test of the 8 port intel board firmware rewrite looked kinda like this: BERT box ---> port1 rx (application program on host takes rx data port1 tx and echoes it to tx) | V port2 rx (same game with application echoing data) port2 tx | V ... repeat for all ports (ours went to 8 ports) ... | BERT box <--------- the BERT was set for the highest baud rate we were supposed to handle (19.2K). the host was a 286 running iRMX at 10Mhz, and the application programs went through the operating system, which in turn talked to the driver which talked to the board. we ran the test for several hours while doing other work on the 286 box, and never lost a byte. could you please do a similar diagram to show how your test was actually performed? >Perry Kivolowitz, ASDG Inc. > >ARPA: madnix!perry@cs.wisc.edu {uunet|ncoast}!marque! >UUCP: {harvard|rutgers|ucbvax}!uwvax!astroatc!nicmad!madnix!perry --Scotty