Another realtime delay question - Mesa card

Ok, thank you. Did that and upgrade to 2.9.2 as well.

So now I get a jitter of 87253 and 83654. It's another 30% reduction on the servo thread and same as before on base thread. Is that ok? Or what else should I do?

Thanks so much, this is quite the part that remains super obscure to me.

EDIT: I still get the realtime error message while running the machine

chienMouille wrote:

 quite the part that remains super obscure to me.
 

The complete lack of an "official preferred list of hardware" that is proven to work with linuxcnc (and its several interfaces), is what remains super obscure to me. After all those years the community remains at the mercy of random and costly hardware trials & errors, tweak this and that, it's endless... everything to ensure that the Free Controller remains unpopular and utterly complex for the new comers. /rant_over. 

Now to your issue. 
1. Can you print the output of below for us to see if the system is up and that no gremlins are lurking:
2. Install ethtool
* Replace eth0 with your ethernet interface name. Mine is 'ens33' for instance. 

Mecanix wrote:

The complete lack of an "official preferred list of hardware" that is proven to work with linuxcnc (and its several interfaces), is what remains super obscure to me. After all those years the community remains at the mercy of random and costly hardware trials & errors, tweak this and that, it's endless... everything to ensure that the Free Controller remains unpopular and utterly complex for the new comers. /rant_over. 

True, and that will remain so but not due to lack of effort on our part (our=members of this forum), i will just run through some reasons for the situation:
-Intel has crippled PC development for nearly 20 years, see especially gen 10, 11, 12, 13 and 14, to many f@ck ups to mention here, the least of it is 14900 using 360W of power to run!
-Microsoft is hell bent on not allowing anything except Windows to run on a PC, while it does support a lot of open source projects just so it can take over whenever those projects gain traction, see "open AI" and GitHub.
-Dell, HP, Lenovo, and all PC hardware manufacturers need to bow down to both of the above, so the PC hardware changes so much that it is impossible to follow what is going on, while miss using the fact that normal users do not understand the hardware, so selling cr@ppy PC, see Dell Alienware for $5000 as utterly useless except as a room heater.
-Asus, MSI, Gigabyte, Asrock selling boards and power supplies etc that are miserably crippled by, oooo say disabled current limiting to the processor ending up in burnt CPU's whenever user changes some settings in BIOS, see youtube Gamers Nexus
-etc etc...
I have done tests for over 50 PC's and laptops, there are several topic with results here, and all of that is not of much use as the time limit for finding those PC's is quite limited, necessitating more test consistently.
So far we have resorted to using used HP 8300 and lately G800 (the 65W version, not the 35W version), Dell 980 as the big one has parallel port and runs Mesa cards perfectly, and some Fujitsu.
Just to prove a point about hardware, i have an Acer laptop that runs Mesa cards perfectly, so i found another one, same model, type, everything, it was terrible! Must have changed something in electronics.
Lenovo laptops were utterly useless even for machine testing, now i have 3 different new-ish ones that are quite good.
An Acer Switch 5 tablet that runs the machine through a USB to Ethernet card perfectly, suddenly the NVME failed, an Acer laptop that runs the machine perfectly, one day just decided to not turn on anymore. Hence never buy Acer for machine control, despite the old laptop mentioned up there still working, new ones are not reliable.
I can go on and on, the point is, not easy to have a list of hardware when the same make, model, type, PC is not the same inside.
Latency tests:
forum.linuxcnc.org/18-computer/39371-res...or-use-with-linuxcnc
Hardware tests, older stuff and some new PC's:
forum.linuxcnc.org/18-computer/38838-har...ency-tests-used-pc-s

Your advanced knowledge obviously evolved from the hobby with hardware you have, you are fortunate, Tommy. Problem being; that's not the case with >98% of us (us, the users). We're really just trying to make parts when you think about it and by all means don't want to f around with hardware markets and/or their stocks (or whatever reason they swim their consumers in that sort of thick mud you're making ref too - I agree btw).

Soon could be a good time to launch and advertise the Linuxcnc EasyMotion™ & the Linuxcnc Pro-Gear™ (names made up) i.e. turn-key and supported half-decent hw solutions for the new comer to harvest the performance of that code and innovative interface. Just a suggestion...

In a meantime I look forward seeing member's chienmouille (wetdog?) problem investigated and hopefully solved

the first command is not returning anything, even with the right interface name which is enp1s0 in my case.
outputs from ethtools:




EDIT: Also, if the name of the interface is enp1s0 does the previous change we made to the interfaces file starting with eth1 does anything?

Attached 's a side-by-side compare of your enp1s0 against another that ping avg 0.04ms with 0pkt losses. I'm not saying your interface is the issue, could also well be core/processing related given the less-than-ideal 80k'ish latency figures. 

Fancy getting familiar with [hint] the system's interrupts affinity and policies? Trust this 'tweaking-jazz' is endless and often makes things worst, as much as occasionally yielding surprising performance enhancements. I'd try that before thinking of getting the Linuxcnc Pro-Gear™