diffrence between latency-test vs latency-histogram

Hi
I have been doing some jitter testing. I am getting consistent results with latency-test in the <9000 for base and <10000 for servo over an 8 hour run (2 of them)

I ran latency-histogram and that shows rare but consistent spikes of >30-50 uSec. I am trying to understand how these tests are used and what I should expect to see. Latency test seems clear, the histogram is more puzzling.

Does anyone know if there are any current documents describing them? I have not been able to find but may be looking in the wrong place.

Thanks in advance

Have you been running either test with graphic intensive tasks in the background?
If you just run the latency test, and no other background tasks I would expect it to show better latency than the latency-histogram, simply because the latency histogram is more graphics intensive. You really should torture the machine in either case with large file transfers, network access, video graphics intensive applications etc etc.

Thank you for the response PCW.

I have run the gears program and youtube videos.

I did get an impact when using the old graphics card. I suspect because it has very little memory and there might be some swapping. With the intel CPU graphics, the average was higher by about 9 - 10 mSec. But stayed constant. So I might just need a slightly better gpu, this one is an nVidia 210 with 1 gig. The ATI 6450 is cheap up here, but I do not know if it would be an improvement.

I also tried turning off the onboard NIC, got interesting results very bad latency spikes, I suspect there are some "blocking" processes trying to access the net. Turned it back on and things improved.

I think I also need to do a major debloating of the system, there are lots of applications I do not need. Lots of "Let me help you" apps that want the web etc.

I have also noticed some really bad spikes within the first few minutes of running, that then seem to disappear with long term use, again I suspect background processes.

I must admit I find this interesting to sort through, but it can be a frustrating maze at times.

Some good news, realVNC does not seem to have any significant impact on jitter. I use the "home license" which is free for up to 5 systems and sufficient for my needs. So I will be able to access the machines "remotely" making workshop setup a little neater and computers tucked away from the "machining" environment.