On Mon, 19 Jan 2009 00:28:22 -0600 in comp.protocols.time.ntp, hal-use...@ip-64-139-1-69.sjc.megapath.net (Hal Murray) wrote:
>>> Linux seme to be having a real real problem with its time calibration >>> routines. It's drift rate jumps on reboot by up to 50PPM from one >>> reboot to the next. > >>Really? I don't recall ever seeing that. > >I thought it was well known. It's been discussed here several/many >times. I've seen jumps much bigger than 50 ppm. > >The problem is the TSC calibration routine in recent kernels. > >It doesn't get the same answer. It's not just temperature. At >one point, I hacked the kernel to call it several times and print >all the answers. It matched the spread below. > >It's close enough so that nobody but a time geek would notice. > > >Here are some handy samples from the line that gets printed out >at boot time: > >Jan 1 14:23:16 shuksan kernel: Detected 2793.159 MHz processor. >Dec 13 10:14:27 shuksan kernel: Detected 2793.118 MHz processor. >Nov 13 02:15:22 shuksan kernel: Detected 2793.049 MHz processor. >Oct 2 03:29:00 shuksan kernel: Detected 2793.117 MHz processor. >Oct 13 12:26:08 shuksan kernel: Detected 2793.236 MHz processor. > >If I did the math right, that's 66 ppm peak-to-peak. I'm pretty >sure I've seen much worse. Is spread specturm clock signal generation (EMI reduction) disabled for the CPU and buses in the firmware? That can cause these variations. ISTR it being mentioned in the FAQ. -- Thanks. Take care, Brian Inglis Calgary, Alberta, Canada brian.ing...@csi.com (Brian[dot]Inglis{at}SystematicSW[dot]ab[dot]ca) fake address use address above to reply _______________________________________________ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions