On Wed, Aug 24, 2011 at 7:00 PM, Per-Olov SjC6holm <p...@incedo.org> wrote: > On 23 aug 2011, at 19:30, Tomas Bodzar wrote: >> On Tue, Aug 23, 2011 at 7:21 PM, Per-Olov Sjvholm <p...@incedo.org> wrote: >>> On 23 aug 2011, at 10:54, Patrick Lamaiziere wrote: >>>> Le Mon, 22 Aug 2011 22:49:47 +0200, >>>> Per-Olov SjC6holm <p...@incedo.org> a C)crit : >>>> >>>> Hello, >>>>> Have not tried current, but will try current as soon as I can. >>>>> Also... I will try to do some laborations with CPU speed of the core >>>>> the OpenBSD virtual machine has. This to see how the interrupts and >>>>> throughput is related to the CPU speed of the allocated core. >>>> >>>> It would be nice to know if current is better with Intel em(4) cards. >>>> because of this commit : http://freshbsd.org/2011/04/13/00/19/01 >>>> >>>> Here we reach 400 MBits/s with a CPU rate ~70% but we >>>> run OpenBSD 4.9. >>>> >>>> Regards. >>>> >>> >>> >>> How fast is your CPU ? >>> >>> Yes I can see the 1.254 commit with this came in after the 4.9 release that > I >>> use. I can try to see if I can measure any performance gain with this > update. >>> >>> I will try this from aug 17... >>> http://ftp.sunet.se/pub/os/OpenBSD/snapshots/i386/install50.iso >> >> Can't see that mirror here http://www.openbsd.org/ftp.html , it's >> better to use something more official >> >>> >>> I4ll get back.... >>> >>> [ YES !! More fun tests.... :D ] >>> >>> Regards >>> Per-Olov >>> >>> > > > > Have tried it now... I tried the 5.0 snapshot from aug 17 with the improved em > driver. Also tested with more allocated cores and the SMP kernel. > > Result on 5.0 snapshot with improved em driver: > > - SMP > worse. Really sucks! _Dramatically_ reduced throughput.
Will be fine to see systat ; systat mbufs ; netstat -m ; vmstat -i and compare them with previous version. Including dmesg (if something changed in dmesg) > > - One processor core (as most of my tests have used) > An improvement, but very little. Maybe 10% better As stated in some of links and posts sent to you - SMP doesn't offer better throughput/sped automatically. You need to test on i386 non-SMP/SMP and amd64 non-SMP/SMP to see what's best. > > > /Per-Olov