UBM1 mill with LinuxCNC

Maybe i already have the Preempt-RT kernel? Checked with these commands:
and

No, PREEMPT is normal, PREEMPT-RT is realtime.

Just install LinuxCNC + Debian Linux using this image:
www.linuxcnc.org/testing-stretch-rtpreempt/

After installing, i had very bad latency values > 100000.
After re-insallation Debian jessy 64bit and switching to preempt-rt kernel i had latency values between 50000 and 100000.
Maybe the debian 32bit has better latency performance than debian 64bit. And maybe the rtai kernel has also better latency performance than preempt-rt. With my first installation with rtai kernel, system latency was below 10000 with isolcpus option.
I will try out in the next days.,

It might be that your system will only give adequate performance with RTAI. In that case you would need to use a PCI Mesa card (Or EPP Parallel Pico or Mesa card) as Ethernet and RTAI don't work together.

After playing around with isolcpus command i have acceptable values with preempt-rt kernel.
In my case i have 4 cpu cores and put this into /etc/default/grub:

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash isolcpus=0,1,2"

Now, only one core is present at the system which makes it a bit slower but still acceptable.

Note that getting this good latency is important with parallel ports but if you use external Ethernet hardware you are likely better off leaving ISOLCPUs alone
(You will typically get better Ethernet latency with all cores running)

I tried several installations during the weekend. Good latency seems to be impossible with my board and Preempt-RT kernel.
Switched back to the regular linuxcnc with Debian 7 and rtai kernel.Thats even better.
Think i will go with a 7i43 and parallel port.
Thanks so far for all your Input.

mane wrote: Good latency seems to be impossible with my board and Preempt-RT kernel..

I am confused by this statement, as you seem to have already demonstrated good latency with the preempt-rt kernel.

I had only good results with preempt-rt kernel by using
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash isolcpus=0,1,2"

But as PCW said i should not use isolcpus together with ethernet.

Thats why i go back to parallel port.

"good latency" depends on whether you have external step generation hardware
100K latency is probably fine for an Ethernet stepgen card ( but not for a parallel port interface)