Need some assistance understanding my Latency Test result

More
27 Jul 2022 19:58 #248452 by PCW
1. I would test with just a servo thread since with external hardware,
the base thread will not be used. (use the --nobase latency-histogram option)

2. You should enable all cores

3. I still suspect there is some power saving option enabled (whether or not
it can be disabled in the BIOS is an open question however)
The following user(s) said Thank You: Sray69

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 20:35 #248461 by Sray69
Here are two more histograms. One is nobase. The other has both. This time I enabled all cores and I found a few other things (PCI-E/SATA/etc management) to turn off in the bios. The latency test looks pretty much the same as before, right 10M jitter.

 
Attachments:

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 20:38 #248462 by Sray69
I noticed the servo thread histograms look quite different from each other. The only difference is the nobase one I had a browser open with gmail where I did not with the other one.

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 20:45 #248463 by Todd Zuercher

So are you saying I should run it in a virtual machine and that it will not run on this PC otherwise? Can you explain a little more as to why? Is it just the MB configuration is bad?

I will turn the cores back on. Thanks for clearing that up for me.

You say "I think it may need to say rt-preempt or something". So my kernel says "SMP PREEMPT RT Debian 4.19.194-2". This is somehow different? I guess I am not understanding.

Thanks
 


No I was trying to ask if you were trying to run this in a VM because your latency results are that bad, and that your results are worse than I see in a VM. But A VM will never have good enough latency to run a machine. (No you should not try it. VMs are only good for simulations.)

Like Peter said there must be some power saving feature (such as C states) or some other latency destroying feature enabled in your bios that needs to be turned off.
The following user(s) said Thank You: Sray69

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 20:58 #248465 by Sray69
Oh OK. I was hoping not to use a VM. Thanks for explaining.

Yeah the BIOS is fairly limited. I will dig through it again and see what I can find.

I did think of a couple things that may contribute to my problem.
  • The memory in this PC are not matched. In fact 2 sticks are 4GB each and the other two sticks are 8GB each (completely different brands), but I cannot get the 8GB sticks to register as 8GB. They only show up as 4GB each. BTW, this also happens when I only have the two 8GB sticks in by themselves. Not sure why this happens. So I decided to just throw the other two in with them (since this MB allows for mismatched memory) to get to 16GB.
  • I am using a BT mouse on this PC.
Not sure how much these things would affect latency?

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 21:45 #248472 by arvidb

Also from the terminal while histogram was running:
"msg=can't use non-numeric floating-point value as operand of "*" (variance=-2116125128)"

If he's running preempt-rt, shouldn't it say "Note: Using POSIX realtime"? Mine does:
$ latency-histogram --nobase
Note: Using POSIX realtime
$ uname -a
Linux lathe 5.10.0-11-rt-amd64 #1 SMP PREEMPT_RT Debian 5.10.92-1 (2022-01-18) x86_64 GNU/Linux
Or was this not displayed on 2.8?
The following user(s) said Thank You: Sray69

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 22:19 #248479 by Sray69
So I ran latency-histogram --nobase again to see if I get that message. My message says "Note: Using POSIX non-realtime".

Good catch! So what do I need to do to install a realtime kernel? Will this mess up the Windows OS that is loaded on it? I don't want to mess that up as that is my wife's remote computer for work.

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 22:24 #248480 by tommylight
@Arvi,
it is displayed in 2.8 also.
Op should check for it just to make sure, he is using the same kernel as in the official ISO, so it should be OK.
@Sray,
-Remove all but one memory sticks and test,
-Test all four sticks, one by one,
-Remove the BT mouse and test,
--
In general, it looks like dry capacitors on the VRM section of the motherboard.
Iwould still look for overheating processor, so remove the cooler, clean, add some new thermal paste and re-seat the cooler.

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 23:13 #248486 by Sray69
Well this is turning out to be more effort than I wanted. I was just hoping to be able to buy another one of these PCs for the LinuxCNC and move on but doesn't look like it will be that easy.

But, I have another PC that I can repurpose that has:
MB: ASUS M5A78L-M/USB3
CPU: AMD FX-6300 Six-core Processor 3.5 GHz
RAM: 16GB (2 x 8GB) DDR3-1600 (800MHz clock) PC3-12800

I just ran a latency test and a histogram. Here are the results:
   

The numbers look better to me, but I am not real sure what I am looking for/at.  What do you guys think? Would this work better?
BTW, it does say that I am using a realtime kernel.
Attachments:

Please Log in or Create an account to join the conversation.

More
27 Jul 2022 23:14 - 27 Jul 2022 23:20 #248487 by PCW
tommylight: the OP got "Note: Using POSIX non-realtime"
which means the latency app didn't run as real time, so the
latency numbers are not meaningful.

This is likely a LinuxCNC installation issue.

How was LinuxCNC installed?

Does the latency test now report "Using POSIX realtime?
Last edit: 27 Jul 2022 23:20 by PCW.

Please Log in or Create an account to join the conversation.

Time to create page: 0.152 seconds
Powered by Kunena Forum