Re: [Emc-users] Latency problem with debian 7 and not with ubuntu 10.04
Hi pierre the same was observed in my p4 asus 775i65G, really bad latencys on debian pae kernels, dont know if its a kernel issue or anything, switch back to 10.04 agh. regards rck 2016-05-24 15:41 GMT-03:00 Forum Deswysen : > Hello, > > i86 pentium 4 Fujitsu siemens scenic E600 : > > with ubuntu 10.04 and 2.7 linuxcnc > *latency +/- 22,000 max* > > I format the pc and i install debian 7 and 2.7 linuxcnc > *latency +/- 65,000 max* > > Why do you have an idea? > > Best regards, > > Pierre > > -- > Mobile security can be enabling, not merely restricting. Employees who > bring their own devices (BYOD) to work are irked by the imposition of MDM > restrictions. Mobile Device Manager Plus allows you to control only the > apps on BYO-devices by containerizing them, leaving personal data > untouched! > https://ad.doubleclick.net/ddm/clk/304595813;131938128;j > ___ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users > -- Mobile security can be enabling, not merely restricting. Employees who bring their own devices (BYOD) to work are irked by the imposition of MDM restrictions. Mobile Device Manager Plus allows you to control only the apps on BYO-devices by containerizing them, leaving personal data untouched! https://ad.doubleclick.net/ddm/clk/304595813;131938128;j ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
Dave wrote: > You won't have any issues with the D525MW and an SSD on that Pico power > supply. I have used several and never had a failure yet. > > Some have 3+ years on them now..running 24x7. > Yes, the D525MW is supposed to have a normal power draw of 13 W, while a worst-case maximum test was able to draw something over 20 W. The SSDs generally draw less power when running full blast than many spinning hard drives do when spun down. So, the total draw of the system should be well under 20 W. Jon -- Own the Future-Intel(R) Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2 ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
You won't have any issues with the D525MW and an SSD on that Pico power supply. I have used several and never had a failure yet. Some have 3+ years on them now..running 24x7. Dave On 3/28/2013 2:53 PM, bjørn wrote: > snip > /"I keep mulling over this "5-6 hours" observation. I know we recommend > running the latency test for a long time but I don't think we understand > why these spiking events show up rarely. It's all well and good to blame > power saving features in modern CPUs but they are functioning over much > smaller time scales. If they are involved in 5-6 hour spikes, then it > would seem we're experiencing some sort of "perfect storm" phenomenon."/ > > You are probably right, this happens after to long a time, and not the > same each time either. But i am still suspicious to the blank screen. > All screensaver functions are already set to "never" thats a bit of a > mystery. > > Now i have updated BIOS but no new options to turn of ACPI and still the > same latency. so thats a dead end. > > The Picopsu is rated 80w as you say, but the "12v brick" that came with > it is only 60w. and i see in the manual that G0 = Full power> 30 W > but G1 = 5 W< power< 52.5 W then with the SSD it could be that it's to > smal and running hot. > I will test with a "real powersupply" after the hollydays. > > To be continued.. > > Tusen takk for hjelpen. > > Regards Bjørn. > > -- Own the Future-Intel(R) Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2 ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
On Mar 28, 2013 2:58 PM, "bjørn" wrote: > > snip > /"I keep mulling over this "5-6 hours" observation. I know we recommend > running the latency test for a long time but I don't think we understand > why these spiking events show up rarely. It's all well and good to blame > power saving features in modern CPUs but they are functioning over much > smaller time scales. If they are involved in 5-6 hour spikes, then it > would seem we're experiencing some sort of "perfect storm" phenomenon."/ > > You are probably right, this happens after to long a time, and not the > same each time either. But i am still suspicious to the blank screen. > All screensaver functions are already set to "never" thats a bit of a > mystery If it's an LCD monitor, perhaps it has to do with two way handshaking between the monitor and the unix box. Is there a power save mode built into the monitor itself? Anything special about the monitor? As an aside, turn off (in BIOS) any ports you do not need. I've seen Dell desktops get good latency figures over 24 hour testing with unused ports disabled. Those hosts used Mesa cards instead of the parallel ports. If I recall correctly, the rs232 serial ports were disabled as well. I'll be putting one of them into service on my first retrofit soonish. If there was interest I could test disabling the ports separately. Best, Jason -- Own the Future-Intel(R) Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2 ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
snip /"I keep mulling over this "5-6 hours" observation. I know we recommend running the latency test for a long time but I don't think we understand why these spiking events show up rarely. It's all well and good to blame power saving features in modern CPUs but they are functioning over much smaller time scales. If they are involved in 5-6 hour spikes, then it would seem we're experiencing some sort of "perfect storm" phenomenon."/ You are probably right, this happens after to long a time, and not the same each time either. But i am still suspicious to the blank screen. All screensaver functions are already set to "never" thats a bit of a mystery. Now i have updated BIOS but no new options to turn of ACPI and still the same latency. so thats a dead end. The Picopsu is rated 80w as you say, but the "12v brick" that came with it is only 60w. and i see in the manual that G0 = Full power > 30 W but G1 = 5 W < power < 52.5 W then with the SSD it could be that it's to smal and running hot. I will test with a "real powersupply" after the hollydays. To be continued.. Tusen takk for hjelpen. Regards Bjørn. -- Med vennlig hilsen Bjørn Kristiansen. www.spesmek.no -- Own the Future-Intel(R) Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2 ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
On 3/28/2013 5:23 AM, bjørn wrote: > that I bought the power supply together with the D525MW board. here are a > link - > http://www.ebay.com/itm/160721043666?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1438.l2649 > it says 60w but i don't know the quality of it. I will test another one > when i get it in a week. Actually the "80" in its model number signifies 80w rating. It's a lightweight power supply but its specifications (http://www.logicsupply.com/products/picopsu_80?gclid=CNjTooGsn7YCFdGd4Aod6lIA-g) suggests it is adequate for the D525MW itself (see table 30, http://downloadmirror.intel.com/19123/eng/D525MW_D525MWV_TechProdSpec.pdf). I don't know how much additional load you have with your SSD drive, etc. I don't know how well these little guys filter spikes on the power mains but that can be tested by powering it from an UPS or power conditioner. (This is just good housekeeping. I haven't had problems with powerline noise in my home but it was a problem in my lab at work. It can come from unexpected sources. Years ago, a new copier machine in a nearby office generated enough EMI/RFI to drive my experimental apparatus crazy.) These PicoPSUs don't have much thermal mass. You probably should make sure it is staying cool. > > I to have seen the problem regarding USB and latency on my mill and > plasma table. I told the guys in the shop to transfer files from USB > thumbdrive to desktop and remove USB thumbdrive > before running Linuxcnc > > at first i started out with a wireless mouse and keypad with a USB > dongle, but replaced it with USB when the spikes appeared. i have now > tested using a ps2 mouse, i do not have a ps2 keyboard so that's still > USB. but i have tried to disconnect and connect it while the > latency-test where running and no spikes i left it out and no other USB > devices connected, let the test run for 5-6 hours but still spikes at > 5 - 6. I'd say you've exercised due diligence:-) I keep mulling over this "5-6 hours" observation. I know we recommend running the latency test for a long time but I don't think we understand why these spiking events show up rarely. It's all well and good to blame power saving features in modern CPUs but they are functioning over much smaller time scales. If they are involved in 5-6 hour spikes, then it would seem we're experiencing some sort of "perfect storm" phenomenon. > > just a tought, what about SATA and SSD disk could that have some > influence on the latency? This seems unlikely but device substitution is the usual quick-n-dirty test. You could try running the LinuxCNC/LiveCD distribution from a USB thumbdrive. I've never heard anyone raise issues regarding SATA itself. > > But anyway i still think i have some powersave funktion running. I > noticed that the screen was black when i come back to the office, witch > i havent seen before. Ubuntu 10.04LTS offers a screen-saver mode that can be disabled via the menu bar. Select System.../Preferences.../Power Management and select "never" for the display sleep function. If this function is already disabled then I'm out of ideas. > I think i have to go back to check my bios version to see if there are > an uppgrade or alternative. It looks like 0130 is the current version (http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=22377&lang=eng&wapkw=d525mw+bios). I have to say I'm skeptical that your BIOS is the cause of this mysterious spiking but I suppose upgrading is worth a try (some say upgrading the BIOS falls into the "if it ain't broke, don't fix it" category). > Regards Bjørn "Lykke til!" Regards, Kent -- Own the Future-Intel® Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://p.sf.net/sfu/intel_levelupd2d ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
Message: 1 Date: Mon, 25 Mar 2013 15:25:31 -0400 From: "Kent A. Reed" Subject: Re: [Emc-users] Latency problem To: "Enhanced Machine Controller (EMC)" Message-ID: <5150a4ab.6050...@gmail.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed On 3/25/2013 9:56 AM, Bj?rn wrote: >> now i have tested: boot with isolcpu(big >> improvement in general latency but still spikes) >> turn of smi, (improvement in general latency but still spikes) >> 'idle=poll' and 'nohlt'(little or no effect) >> >> acpi = off and api=off (little or no effect) >> >> For the moment i have been running the system for 1,5 hour and have - >> servothread interval 999655 and max jitter 4495 ns - base thread interval >> 32069 max jitter 7190. >> >> witch i am very happy with, but if i run for several hours i still get >> spikes up to 5 and above. >> i suspect Michael are right, the problem is related toidle/powersave states. >> >> I have no settings in bios to turn of ACPI >> my bios is: MWNT10N.86A.0083.2011.0524.1600 >> is this the same as others are running? and do you have option to turn off >> ACPI from bios? > Bj?rn: > > I don't have an Intel D525MW in my stable of motherboards but I do have > an ASUS AT5NM10-I whose numbers I posted last year to the table of > latency-test results in the Wiki. Your "pre-spike" numbers are > reasonably consistent with the numbers I got. The "best" numbers I > posted were obtained with hyperthreading turned off in the BIOS and > isolcpus=1 set in the boot parameters. Nothing heroic. I did my tests > long before list discussions brought out the importance of other > parameter settings like idle=poll. In recent shortterm tests, I also > have not seen dramatic improvements manipulating these parameters. > > Testing this ASUS board, I haven't seen the kinds of spikes you are > experiencing. (Just for the record, I was testing LinuxCNC over 10.04LTS > with the RTAI kernel). > > I'm curious to know what all is attached to your motherboard, especially > anything attached via USB. I once had a motherboard that started > misbehaving when I switched from a PS/2 mouse to a USB mouse and a > number of us have seen the simple act of plugging in a USB thumbdrive > cause bursts in latency numbers. Is it possible you have a peripheral > interface/device that is waking up on the same time scale as the > occurrence of your spikes? > > My system guy never failed to ask if the computer were plugged to a live > outlet as his first step in diagnosing problems. Obviously your computer > is plugged in but how are the power supply and the AC mains? Adequate, > stable, clean? > > Regards, > Kent > > I bought the power supply together with the D525MW board. here are a link - http://www.ebay.com/itm/160721043666?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1438.l2649 it says 60w but i don't know the quality of it. I will test another one when i get it in a week. I to have seen the problem regarding USB and latency on my mill and plasma table. I told the guys in the shop to transfer files from USB thumbdrive to desktop and remove USB thumbdrive before running Linuxcnc at first i started out with a wireless mouse and keypad with a USB dongle, but replaced it with USB when the spikes appeared. i have now tested using a ps2 mouse, i do not have a ps2 keyboard so that's still USB. but i have tried to disconnect and connect it while the latency-test where running and no spikes i left it out and no other USB devices connected, let the test run for 5-6 hours but still spikes at 5 - 6. just a tought, what about SATA and SSD disk could that have some influence on the latency? But anyway i still think i have some powersave funktion running. I noticed that the screen was black when i come back to the office, witch i havent seen before. I think i have to go back to check my bios version to see if there are an uppgrade or alternative. Regards Bjørn -- Own the Future-Intel® Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://p.sf.net/sfu/intel_levelupd2d ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
On 3/25/2013 9:56 AM, Bjørn wrote: > now i have tested: boot with isolcpu(big > improvement in general latency but still spikes) >turn of smi, (improvement in general latency but still spikes) > 'idle=poll' and 'nohlt'(little or no effect) > > acpi = off and api=off (little or no effect) > > For the moment i have been running the system for 1,5 hour and have - > servothread interval 999655 and max jitter 4495 ns - base thread interval > 32069 max jitter 7190. > > witch i am very happy with, but if i run for several hours i still get spikes > up to 5 and above. > i suspect Michael are right, the problem is related toidle/powersave states. > > I have no settings in bios to turn of ACPI > my bios is: MWNT10N.86A.0083.2011.0524.1600 > is this the same as others are running? and do you have option to turn off > ACPI from bios? Bjørn: I don't have an Intel D525MW in my stable of motherboards but I do have an ASUS AT5NM10-I whose numbers I posted last year to the table of latency-test results in the Wiki. Your "pre-spike" numbers are reasonably consistent with the numbers I got. The "best" numbers I posted were obtained with hyperthreading turned off in the BIOS and isolcpus=1 set in the boot parameters. Nothing heroic. I did my tests long before list discussions brought out the importance of other parameter settings like idle=poll. In recent shortterm tests, I also have not seen dramatic improvements manipulating these parameters. Testing this ASUS board, I haven't seen the kinds of spikes you are experiencing. (Just for the record, I was testing LinuxCNC over 10.04LTS with the RTAI kernel). I'm curious to know what all is attached to your motherboard, especially anything attached via USB. I once had a motherboard that started misbehaving when I switched from a PS/2 mouse to a USB mouse and a number of us have seen the simple act of plugging in a USB thumbdrive cause bursts in latency numbers. Is it possible you have a peripheral interface/device that is waking up on the same time scale as the occurrence of your spikes? My system guy never failed to ask if the computer were plugged to a live outlet as his first step in diagnosing problems. Obviously your computer is plugged in but how are the power supply and the AC mains? Adequate, stable, clean? Regards, Kent -- Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
Am 20.03.2013 um 08:48 schrieb Bj?rn: > I bougt this. > http://www.ebay.com/itm/200542103383?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1438.l2649 > running the latest version of linuxcnc. specifically which version? I have a D525MW and get much better results > Hypertreading off in bios, > ISOLCPU is done. > > when runnnig latencytest and 3 x glxgears + some video i can run for 30 > min and get: > servo thread (1.0ms) max interval (ns) 1006129 max jitter 11104 > Base thread (25.0ps) max interval (ns) 42342 max jitter 17463 > > i think i could do with this but if i run overnight the same way i get > servo thread (1.0ms) max interval (ns) 1020138 max jitter 24978 > Base thread (25.0ps) max interval (ns) 83039 max jitter 58160 > > does anyone have any idea of hvat this can be? is there a way to se any > debugging information? usually that points to CPU idle/powersave states; it has turned out that wakeup from idle/powersave states is rather nondeterministic and a likely cause of latency spikes to isolate whether this is the case, try to boot the kernel with the option 'idle=poll' if things improve, idle/powersave states are very likely the cause -m > And another question, is it possible to dedicate one cpu to run gui and > other non realtime funktions? > > Regards Bjorn > I am running 10.04 LTS and LinuxCNC 2.5.2 when you say, boot the kernel with the option 'idle=poll' is this done the same way as isolcpu? Regards Bjorn Kristiansen. yes try both 'idle=poll' and 'nohlt' see section 4 of http://wiki.linuxcnc.org/cgi-bin/wiki.pl?XenomaiKernelPackages this applies to RTAI just alike - Michael now i have tested: boot with isolcpu(big improvement in general latency but still spikes) turn of smi, (improvement in general latency but still spikes) 'idle=poll' and 'nohlt'(little or no effect) acpi = off and api=off (little or no effect) For the moment i have been running the system for 1,5 hour and have - servothread interval 999655 and max jitter 4495 ns - base thread interval 32069 max jitter 7190. witch i am very happy with, but if i run for several hours i still get spikes up to 5 and above. i suspect Michael are right, the problem is related toidle/powersave states. I have no settings in bios to turn of ACPI my bios is: MWNT10N.86A.0083.2011.0524.1600 is this the same as others are running? and do you have option to turn off ACPI from bios? -- Regards: Bjørn Kristiansen -- Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
>> >> Regards Bjorn >> > I am running 10.04 LTS and LinuxCNC 2.5.2 > > when you say, boot the kernel with the option 'idle=poll' is this done > the same way as isolcpu? yes try both 'idle=poll' and 'nohlt' see section 4 of http://wiki.linuxcnc.org/cgi-bin/wiki.pl?XenomaiKernelPackages this applies to RTAI just alike - Michael -- Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
Am 20.03.2013 um 08:48 schrieb Bj?rn: > I bougt this. > http://www.ebay.com/itm/200542103383?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1438.l2649 > running the latest version of linuxcnc. specifically which version? I have a D525MW and get much better results > Hypertreading off in bios, > ISOLCPU is done. > > when runnnig latencytest and 3 x glxgears + some video i can run for 30 > min and get: > servo thread (1.0ms) max interval (ns) 1006129 max jitter 11104 > Base thread (25.0ps) max interval (ns) 42342 max jitter 17463 > > i think i could do with this but if i run overnight the same way i get > servo thread (1.0ms) max interval (ns) 1020138 max jitter 24978 > Base thread (25.0ps) max interval (ns) 83039 max jitter 58160 > > does anyone have any idea of hvat this can be? is there a way to se any > debugging information? usually that points to CPU idle/powersave states; it has turned out that wakeup from idle/powersave states is rather nondeterministic and a likely cause of latency spikes to isolate whether this is the case, try to boot the kernel with the option 'idle=poll' if things improve, idle/powersave states are very likely the cause -m > And another question, is it possible to dedicate one cpu to run gui and > other non realtime funktions? > > Regards Bjorn > I am running 10.04 LTS and LinuxCNC 2.5.2 when you say, boot the kernel with the option 'idle=poll' is this done the same way as isolcpu? Regards Bjorn Kristiansen. -- Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Latency problem
Am 20.03.2013 um 08:48 schrieb Bjørn: > I bougt this. > http://www.ebay.com/itm/200542103383?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1438.l2649 > running the latest version of linuxcnc. specifically which version? I have a D525MW and get much better results > > Hypertreading off in bios, > ISOLCPU is done. > > when runnnig latencytest and 3 x glxgears + some video i can run for 30 > min and get: > servo thread (1.0ms) max interval (ns) 1006129 max jitter 11104 > Base thread (25.0ps) max interval (ns) 42342 max jitter 17463 > > i think i could do with this but if i run overnight the same way i get > servo thread (1.0ms) max interval (ns) 1020138 max jitter 24978 > Base thread (25.0ps) max interval (ns) 83039 max jitter 58160 > > does anyone have any idea of hvat this can be? is there a way to se any > debugging information? usually that points to CPU idle/powersave states; it has turned out that wakeup from idle/powersave states is rather nondeterministic and a likely cause of latency spikes to isolate whether this is the case, try to boot the kernel with the option 'idle=poll' if things improve, idle/powersave states are very likely the cause -m > > And another question, is it possible to dedicate one cpu to run gui and > other non realtime funktions? > > Regards Bjorn > > > -- > Everyone hates slow websites. So do we. > Make your web apps faster with AppDynamics > Download AppDynamics Lite for free today: > http://p.sf.net/sfu/appdyn_d2d_mar > ___ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users -- Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users