Latency. EMC2 (ubuntu v8) vs Linuxcnc 2.7 (ubuntu v10)

Hello, I have a Dell Optiplex GX260: P4, 2GHz, 1GB RAM, stock from the factory. I performed the latency test using linuxcnc v2.7 which returned value of over 100,000nS. I researched possible fixes and discovered the SMI issue. Before tackling that, I decided to see if the test was any different using EMC2 v8 latency test. It was more than ten times better, <10,000nS.
I have a few cnc projects which I would like to utilize linuxcnc. I thought I would start with the GX260 computer and my HFT 7" lathe. If the older version of linuxcnc, EMC2 v8, will run well on the GX260 and the 7" lathe then I can use a more suitable PC and linuxcnc v2.7 for my next project. I welcome your comments and suggestions.
Thanks

forum.linuxcnc.org/18-computer/19125-faq-some-latency-solutions > Which Kernel

You can update the Linuxcnc version up to a certain level, can't remember what that is OTTOMH, something like 2.5 I think

Thanks for the link to the faq. I suppose I am more curious about why the latency test in EMC2 v8 gives a very good 10,000nS but linuxcnc v2.7 gives a terrible 100,000nS on the same PC. Am I to conclude that the older EMC2 will work well on the GX260 but linuxcnc 2.7 will not? If so then the options are use the older distribution or dive into the software guts of the PC and linuxcnc 2.7 to try and create a fix. Other than better features and etc. is there a down side to using the older EMC2 with the little lathe? Of course I have considered a different PC but the wonderful thing about linux(cnc) is the low resource requirements allow perfectly good old(er) PCs to find a new home other than a trash can. I'd like to see if I can do that with this PC and lathe. Again I welcome the comments, insights, and suggestions of the membership.
Thanks

I suppose I am more curious about why the latency test in EMC2 v8 gives a very good 10,000nS but linuxcnc v2.7 gives a terrible 100,000nS on the same PC.

If you read the link I posted, you will have your answer.

It is nothing to do with the latency-test, but using a kernel compiled for a single processor with a single processor machine.

Arceye, I read the information referenced. Now let's see if I understand. The older distribution, EMC2, was compiled for a single processor, such as the P4 and that is why the test results produce low latency number. Correct? The newer distribution of linuxcnc is compiled for a multiCore processor and I am running it on a single core processor, so the latency test numbers are poor. Correct? Though my Linux knowledge is on the low end of the curve, I believe that I understand why it is necessary to set the core number to 0 (zero). I don't understand why this applies to a single core processor that has only one. Your patience is greatly appreciated.

Let's try this again.

Linuxcnc Latency Tests

Computer. Ubuntu v8 Ubuntu v10 Ubuntu v12

---

---

---

---

#1 Dell Optiplex GX260 9,000nS 10,000nS 110,000nS ***
#2 MSI MB w/Athalon 64 4000+ 8,000nS 9,000nS 15,000nS

If the problem is related to the SMP issue then I don't understand given that both processors in computers #1 and #2 are single core. If it is SMP related then this must mean that everyone who wants to use linuxcnc Ubuntu v12 with a single core processor computer must then go through the motions of modifying the standard installation? I am sure that this problem is old and worn but it is new to me.
Thanks

On this litle computer (SX260) and Ubuntu 10, I have the best latency I ever see ~4000ns. Stock, no tuning, 2 GRam.


The post is from 2016