Why the latency is so different
- RStrlcpy
- Offline
- New Member
-
Less
More
- Posts: 1
- Thank you received: 0
20 Dec 2016 17:16 #84465
by RStrlcpy
Why the latency is so different was created by RStrlcpy
Hi.
I'm trying to setup my first installation.
Gigabyte GA-MA785GM + Athlon II X4 620 + 8 GB DDR2-800 + Radeon HD6570 (external)
What I tried and what I have at this time:
1. Debian Wheezy 7.11 + Gnome + RT-Kernel-amd64 from Debian + LinuxCNC 2.7
Max Jitter (servo thread): about 100 us (it is bad according to documentation).
Max Jitter (base thread): ~30 us
After dancing with a tambourine around the BIOS: ~50 us
2. Debian Wheezy 7.11+ XFCE + RT-Kernel-amd64 from Debian + LinuxCNC 2.7
Max Jitter (servo thread): mostly time 14us, sometimes spikes ~30us
Max Jitter (base thread): ~30 us
3. LinuxCNC 2.7 LiveCD (it is based on a 3.4-PAE kernel)
Max Jitter (servo thread): stable 4-5 us
Max Jitter (base thread): 3 us
Could someone explain why the numbers is so different?
I'm trying to setup my first installation.
Gigabyte GA-MA785GM + Athlon II X4 620 + 8 GB DDR2-800 + Radeon HD6570 (external)
What I tried and what I have at this time:
1. Debian Wheezy 7.11 + Gnome + RT-Kernel-amd64 from Debian + LinuxCNC 2.7
Max Jitter (servo thread): about 100 us (it is bad according to documentation).
Max Jitter (base thread): ~30 us
After dancing with a tambourine around the BIOS: ~50 us
2. Debian Wheezy 7.11+ XFCE + RT-Kernel-amd64 from Debian + LinuxCNC 2.7
Max Jitter (servo thread): mostly time 14us, sometimes spikes ~30us
Max Jitter (base thread): ~30 us
3. LinuxCNC 2.7 LiveCD (it is based on a 3.4-PAE kernel)
Max Jitter (servo thread): stable 4-5 us
Max Jitter (base thread): 3 us
Could someone explain why the numbers is so different?
Please Log in or Create an account to join the conversation.
- andypugh
-
- Offline
- Moderator
-
Less
More
- Posts: 19678
- Thank you received: 4559
20 Dec 2016 17:35 #84467
by andypugh
Replied by andypugh on topic Why the latency is so different
One reason that the LiveCD is so much lower is that it uses the RTAI kernel patch for realtime, whereas your other tests are with the PREEMPT-RT kernel (I assume).
RTAI is a more specialised realtime kernel, with some restrictions such as mandating that all realtime code runs at the kernel level. If you are using software stepping then the RTAI kernel should be your choice. If you are using a Mesa ethernet-connected board then you need to use PREEMPT-RT as the kernel-level limitation in RTAI means that the driver can't get to the ethernet port.
RTAI is a more specialised realtime kernel, with some restrictions such as mandating that all realtime code runs at the kernel level. If you are using software stepping then the RTAI kernel should be your choice. If you are using a Mesa ethernet-connected board then you need to use PREEMPT-RT as the kernel-level limitation in RTAI means that the driver can't get to the ethernet port.
Please Log in or Create an account to join the conversation.
Time to create page: 0.221 seconds