[time-nuts] PCI-E Serial Card For Windows NTP?
johnlaur at gmail.com
Fri Jun 26 14:29:59 UTC 2015
> However, it has been noted that the Raspberry Pi has some limitations for this purpose, not the least of which is having a USB connected ethernet port.
Windows 7 in particular offers it's own set of trouble. Based on my
experiences with both platforms, it will be harder to extract
consistent and predictable NTP performance from Win7 than Raspberry
Pi. Many others have suggested that you experiment with BeagleBone
Black; is there a particular reason that you prefer to take the long
route? The out-of-box performance is already very good, and there is
plenty of remaining challenge and accuracy available from engaging the
hardware timers, using a RTOS, and even replacing the main oscillator.
I believe the ethernet controller is even capable of PTP should you
wish to go beyond NTP (never tried it myself).
On Thu, Jun 25, 2015 at 8:55 PM, Brian M <brayniac at gmail.com> wrote:
> Trying not to go offtopic. If there are specific lists for this type of
> nuttery, contact me off list - would love to learn and discuss more.
> That said - if you're going to test the impact on latency for interrupt
> coalescing, I'd suggest using sockperf ping pong test:
> Should reveal a bit more than a simple icmp ping test can. Use the --pps
> flag to test a variety of packet rates. Should help show the effect of
> - Brian
> On Thu, Jun 25, 2015 at 13:40 Hal Murray <hmurray at megapathdsl.net> wrote:
>> I know next-to-nothing about Windows.
>> Serial ports are pretty standard. I'd get the cheapest card. I'd probably
>> pay a bit more for a second port. It might be handy tomorrow.
>> One difference between chips is the depth of the FIFO.
>> PCI card now come in two heights. Make sure you get the right one. The
>> short ones don't have room for a second connector. (Some cards come with
>> face plates. You can swap in/out the other one if you can use a
>> The other thing to keep an eye on is interrupts from the Gigabit ethernet.
>> With a lot of short packets, you can get in trouble spending all your CPU
>> time in the interrupt handler. Some hardware is setup to batch interrupts.
>> The idea is to delay an interrupt for a while in hopes that more packets
>> arrive and get processed as a batch. You may want to turn that off. It's
>> tradeoff between latency and CPU usage. You may be able to measure it with
>> something like ping.
>> These are my opinions. I hate spam.
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to
>> and follow the instructions there.
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the Time-nuts_lists.febo.com