Above we plot the PII/266-66 cpu, Tyan 440BX performance with PPSkit 2.0.1.
Note the spike at 40000 due to unix cron (makewhatis, updatedb) activity.
The overall swish and sway are due to temperature changes--the dive
after 10000 was due to opening the window in the machine room, letting
in cool foggy air. The recovery was due to closing that window.
Above we plot the K6-2/400-100 cpu, FIC MVP3 performance with PPSkit 2.0.2.
The excursions around 60000 reflect kernel recompile, and the right hand
side reflects operation with the new kernel. This recompiled version
of 2.4.19 has the NTP debugging disabled.
The overall range of the two setups appears similar by examination
of the plots. The broad smear of the K6-2 is a result of the "interrupt
collision" that occurs when both the atomized NMEA driver, and the PPS
driver, are both running on the same interface, which is the configuration
used during this plotting period.
The uarts are driven by on board 18.432 Mhz RTXO 7400 logic based
LO, knocked down to 1.8432 by a 7490. A (motherboard) chipset based UART setup
would probably have cpu and uart divider chain sharing the same oscillator.