On Thu, May 21, 2020 at 09:26:59AM -0400, cpham2...@comcast.net wrote:
> Hi, I am running linuxptp version 2.0 in software timestamp mode, and I found
> the "master offsets" are all over the places. In ~7000 offset measurements,
> more than 50% is greater than 100 micro-sec. Maximum offsets observed is
> around 300 micro-secs. I wonder anyone ran linuxptp in software timestamp
> mode and what accuracy level was observed.
> My master clock and PTP clients on are a local network with no PTP-support
> switches. There are 4 switches along the path between the master clock and
> the clients. When I ran linuxptp in hardware timestamp mode in the same
> network setting, I achieved great results: 90% of master offsets is within 1
> micro-sec. In software timestamp mode the master offsets are higher than
> expected. Is there anything I can do to improve the accuracy?
There are few things you could try to improve the stability with SW
timestamping:
- disable interrupt coalescing on the NIC with ethtool -C if possible
- disable power saving features (e.g. set the CPU to a constant
frequency, disable idle states)
- disable energy efficient ethernet (EEE) - this helps with HW
timestamping too
To improve accuracy, use the same HW, SW, and configuration on the
master and slave, so the errors in RX/TX timestamping cancel out.
--
Miroslav Lichvar
___
Linuxptp-users mailing list
Linuxptp-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-users
