Results of latency test, list of computers tested for use with LinuxCNC

More
6 years 4 months ago #172557 by seuchato
chimeno
kindly reply to this

... and thanks for the flowers anyway :-)
greez
chris

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

More
6 years 4 months ago - 6 years 1 month ago #172681 by seuchato
Results for a hp 8300 sff, had to set interval for base thread to 50'000
Code:
LinuxCNC pc tests, version 20200624, started 25.06.2020 03:10:35 ******************************************************************************** General info: Hewlett-Packard, HP Compaq Elite 8300 SFF Bios version K01 v02.90, dated 07/16/2013 Chipset: Intel Corporation 7 Series/C216 Harddisk is non-rotational, ie. SSD The amount of swap currently used is 0 ================================================================================ CPU related data: model name : Intel(R) Core(TM) i5-3570 CPU @ 3.40GHz cpu cores : 4 stepping : 9 cache size : 6144 KB ================================================================================ RAM related data: Maximum Capacity: 32 GB Size: No Module Installed Size: 4096 MB Size: No Module Installed Size: 4096 MB ================================================================================ GPU related data: product: Xeon E3-1200 v2/3rd Gen Core processor Graphics Controller configuration: driver=i915 latency=0 ================================================================================ Os related data: PRETTY_NAME="Debian GNU/Linux 10 (buster)" Desktop: KDE, Session: ================================================================================ Kernel related data: Kernel 4.19.0-9-rt-amd64 GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_idle.max_cstate=0 processor.max_cstate=1 idle=poll isolcpus=1,2,3" Cpu idle driver: none ================================================================================ Kernel modules data: Check if pcspkr is loaded: No, good! Yes, may cause bigger latency. Check if snd modules are loaded: snd_hda_codec_hdmi 57344 1 snd_hda_codec_realtek 122880 1 snd_hda_codec_generic 86016 1 snd_hda_codec_realtek snd_hda_intel 49152 4 snd_hda_codec 151552 4 snd_hda_codec_generic,snd_hda_codec_hdmi,snd_hda_intel,snd_hda_codec_realtek snd_hda_core 98304 5 snd_hda_codec_generic,snd_hda_codec_hdmi,snd_hda_intel,snd_hda_codec,snd_hda_codec_realtek snd_hwdep 20480 1 snd_hda_codec snd_pcm 114688 4 snd_hda_codec_hdmi,snd_hda_intel,snd_hda_codec,snd_hda_core snd_timer 36864 1 snd_pcm snd 102400 16 snd_hda_codec_generic,snd_hda_codec_hdmi,snd_hwdep,snd_hda_intel,snd_hda_codec,snd_hda_codec_realtek,snd_timer,snd_pcm soundcore 16384 1 snd Yes, may cause bigger latency. ================================================================================ Keyboard & Mouse related data: Number of mice: 1 Mouse 0 : Logitech USB Receiver Mice attached to USB: Logitech, Inc. Cordless Mouse Receiver Keyboards attached to USB: MosArt Semiconductor Corp. Defender Office Keyboard (K7310) S Zodiak KM-9010 LinuxCNC related data: LinuxCNC version is: 2.9.0~pre0.2245.g09397b8d3 ================================================================================ Preparing for latency test: Started glxgears No. 1 Started glxgears No. 2 Started glxgears No. 3 Started glxgears No. 4 Started glxgears No. 5 .......... Started firefox with 'https://vimeo.com/150574260' ================================================================================ Command for latency test is: su ck -c "latency-histogram --bbinsize 1000 --base 50000 --sbinsize 1000 --nox 2>&1 | tee 1593047417.lat &" Latency testing loop started Do Jun 25 03:10:49 CEST 2020, should end after Do Jun 25 05:10:49 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 7200 secs servo min: -39.000 uS max: 30.900 uS sdev: 3.400 uS 7201 secs base min: -31.500 uS max: 34.000 uS sdev: 0.600 uS

Same pc, but with a servothread only gives
Code:
... (hw summary see above) ================================================================================ Command for latency test is: su ck -c "latency-histogram --nobase --sbinsize 1000 --nox 2>&1 | tee 1593024764.lat &" Latency testing loop started Mi Jun 24 20:53:09 CEST 2020, should end after Mi Jun 24 22:53:09 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 7202 secs servo min: -63.400 uS max: 60.500 uS sdev: 3.800 uS 7203 secs servo min: -63.400 uS max: 60.500 uS sdev: 3.800 uS

Above results repeatedly achieved several times. Why is it that latency on serothread is appr. 1.5 - 2x the value of a run with a base thread?

greez
chris
Last edit: 6 years 1 month ago by seuchato. Reason: typo

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

More
6 years 4 months ago #172695 by seuchato
Same hp 8300 sff, but with kernel 4.9.0-12-rt-amd64, basthread again at 25⁰000
Code:
================================================================================ Kernel related data: Kernel 4.9.0-12-rt-amd64 GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_idle.max_cstate=0 processor.max_cstate=1 idle=poll isolcpus=1,2,3" Cpu idle driver: none ================================================================================ ... ================================================================================ Command for latency test is: su ck -c "latency-histogram --bbinsize 1000 --base 25000 --sbinsize 1000 --nox 2>&1 | tee 1593058060.lat &" Latency testing loop started Do Jun 25 06:07:55 CEST 2020, should end after Do Jun 25 08:07:55 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 7205 secs base min: -18.000 uS max: 18.700 uS sdev: 0.500 uS 7205 secs servo min: -47.200 uS max: 44.900 uS sdev: 0.900 uS

lookee lookee: no rt delays!

greez
chris
The following user(s) said Thank You: tommylight, johnmc1

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

More
6 years 4 months ago #172735 by seuchato
Still the same hp 8300 sff, same kernel, but this time w/o isolcpus. Got a realtime delay after 15 - 30 seconds, so the run below is for basthread interval 50'000 us
Code:
Hewlett-Packard, HP Compaq Elite 8300 SFF Bios version K01 v02.90, dated 07/16/2013 Chipset: Intel Corporation 7 Series/C216 Harddisk is non-rotational, ie. SSD The amount of swap currently used is 0 ================================================================================ CPU related data: model name : Intel(R) Core(TM) i5-3570 CPU @ 3.40GHz cpu cores : 4 stepping : 9 cache size : 6144 KB ================================================================================ RAM related data: Maximum Capacity: 32 GB Size: No Module Installed Size: 4096 MB Size: No Module Installed Size: 4096 MB ================================================================================ GPU related data: product: Xeon E3-1200 v2/3rd Gen Core processor Graphics Controller configuration: driver=i915 latency=0 ================================================================================ Os related data: PRETTY_NAME="Debian GNU/Linux 10 (buster)" Desktop: KDE, Session: ================================================================================ Kernel related data: Kernel 4.19.0-9-rt-amd64 GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_idle.max_cstate=0 processor.max_cstate=1 idle=poll" Cpu idle driver: none ================================================================================ ... Command for latency test is: su ck -c "latency-histogram --bbinsize 1000 --base 50000 --sbinsize 1000 --nox 2>&1 | tee 1593077144.lat &" Latency testing loop started Do Jun 25 11:26:10 CEST 2020, should end after Do Jun 25 13:26:10 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 7201 secs base min: -28.700 uS max: 26.900 uS sdev: 0.700 uS 7201 secs servo min: -37.800 uS max: 39.900 uS sdev: 1.100 uS

next, a run without base thread
Code:
================================================================================ Command for latency test is: su ck -c "latency-histogram --nobase --sbinsize 1000 --nox 2>&1 | tee 1593088904.lat &" Latency testing loop started Do Jun 25 14:42:08 CEST 2020, should end after Do Jun 25 16:42:08 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 7200 secs servo min: -72.800 uS max: 68.900 uS sdev: 3.100 uS 7201 secs servo min: -72.800 uS max: 68.900 uS sdev: 3.100 uS

Again: servo thread w/o base thread gets 1.5 to 2x the time as with base thread. I really wonder why?

Prel. conclusions:

  • isolcpus does not do too much on this machine
  • kernel matters
  • need to check preventing of loading of sound modules. (Unfortunately bios does not allow to disable audio)
  • looks like some useable pc, given it is gen 3 core cpu based, will probably be a good starting point for quite some time

  • greez
    chris
    The following user(s) said Thank You: tommylight, johnmc1

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

    More
    6 years 4 months ago #172788 by seuchato
    Results of an old ASUS AT5MNI board: Got unexpected realtime delays even with base thread interval at 100'000! also the passive cooler on the board got very hot. So I guess, not useable at all.
    Code:
    LinuxCNC pc tests, version 20200624, started 25.06.2020 17:31:11 ******************************************************************************** General info: ASUSTeK Computer INC., System Product Name Bios version 0214, dated 12/01/2009 Chipset: Harddisk is non-rotational, ie. SSD The amount of swap currently used is 0 ================================================================================ CPU related data: model name : Intel(R) Atom(TM) CPU D510 @ 1.66GHz cpu cores : 2 stepping : 10 cache size : 512 KB ================================================================================ RAM related data: Maximum Capacity: 4 GB Size: 1024 MB Size: 1024 MB ================================================================================ GPU related data: product: Atom Processor D4xx/D5xx/N4xx/N5xx Integrated Graphics Controller product: Atom Processor D4xx/D5xx/N4xx/N5xx Integrated Graphics Controller configuration: driver=i915 latency=0 configuration: latency=0 ================================================================================ Os related data: PRETTY_NAME="Debian GNU/Linux 10 (buster)" Desktop: KDE, Session: ================================================================================ Kernel related data: Kernel 4.9.0-12-rt-amd64 GRUB_CMDLINE_LINUX_DEFAULT="quiet splash intel_idle.max_cstate=0 processor.max_cstate=1 idle=poll" Cpu idle driver: none ================================================================================ Kernel modules data: Check if pcspkr is loaded: No, good! Yes, may cause bigger latency. Check if snd modules are loaded: No, good! Yes, may cause bigger latency. ================================================================================ Keyboard & Mouse related data: Number of mice: 2 Mouse 0 : USB-compliant keyboard Mouse 1 : Logitech USB Receiver Mice attached to USB: Logitech, Inc. Cordless Mouse Receiver Keyboards attached to USB: MosArt Semiconductor Corp. Defender Office Keyboard (K7310) S Zodiak KM-9010 LinuxCNC related data: LinuxCNC version is: 2.9.0~pre0.2258.gc90c1926d ================================================================================ Preparing for latency test: Started glxgears No. 1 Started glxgears No. 2 Started glxgears No. 3 Started glxgears No. 4 Started glxgears No. 5 .......... Started firefox with 'https://vimeo.com/150574260' ================================================================================ Command for latency test is: su ck -c "latency-histogram --bbinsize 1000 --base 100000 --sbinsize 1000 --nox 2>&1 | tee 1593099059.lat &" Latency testing loop started Do Jun 25 17:31:32 CEST 2020, should end after Do Jun 25 19:31:32 CEST 2020 Unexpected realtime delay on task 0 with period 100000 should end after Do Jun 25 19:31:32 CEST 2020 Swap useage: prerun: 0, running: 0 ================================================================================ last latency data is as follows:. 15 secs base min: -93.700 uS max: 121.000 uS sdev: 3.500 uS 15 secs servo min: -57.200 uS max: 59.600 uS sdev: 4.300 uS

    Greez
    chris

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

    More
    6 years 4 months ago #172792 by tommylight
    That might still be usable for running a Mesa board, be it PCI or PCI-E or Eth.
    LinuxCNC does not stress the PC as much as latency histogram does, so there will be less heat, but at the risk of having a bit worse latency than the histogram showed. Still usable, as it seems from the results.

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

    Time to create page: 0.523 seconds
    Powered by Kunena Forum