[Emc-users] Effect of isolcpus boot parameter with D510 cpu ( was Re: hal taking machine down)
On 1/2/2012 11:24 PM, Kent A. Reed wrote: On 1/2/2012 10:35 PM, Jon Elson wrote: gene heskett wrote: Is anyone running emc on this? How is the latency? I've tested the Intel D525MW motherboard, and put it in the database. The servo thread jitter was 15595, the base thread was 11921. That is fine for any system with a hardware interface (servo or stepper) but may be a little marginal for software step generation. Since I'm a servo bigot, it was quite good for me. The on-board parallel port passed all the tests with my EPP devices. EMC 2.4.x with Ubuntu 10.04 loads and runs with no quirks at all. I did the install on a 16 GB SSD hard drive. Jon Jon: This is a multi-cpu board. Did you set the isolcpus boot parameter? The reason I ask is IIRC I got similar numbers with a standard boot from the LiveCD on my ASUS board and saw them drop below 1 when I set isolcpus=1. This may be a false memory (I did the tests last winter) and I'm away from the board at the moment. I'll do the tests again in the morning and this time I'll post the results on the Wiki. Regards, Kent Jon (and Gene if he gets a motherboard outfitted with an Intel Atom cpu): To follow up on this, I just ran 15-minute latency tests on my ASUS AT5NM10-I board first with a stock boot and then with isolcpus=1 as a boot parameter. I've posted my results in the table on the Wiki. In brief, isolating one cpu shaved approximately 6000ns off the max jitters reported for both the base thread and the servo thread. Of course the latency test should be run much longer but I believe these numbers are indicative. YMMV Regards, Kent -- Write once. Port to many. Get the SDK and tools to simplify cross-platform app development. Create new or port existing apps to sell to consumers worldwide. Explore the Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join http://p.sf.net/sfu/intel-appdev ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Effect of isolcpus boot parameter with D510 cpu ( was Re: hal taking machine down)
On Tue, Jan 03, 2012 at 02:11:46PM -0500, Kent A. Reed wrote: To follow up on this, I just ran 15-minute latency tests on my ASUS AT5NM10-I board first with a stock boot and then with isolcpus=1 as a boot parameter. I've posted my results in the table on the Wiki. That info is a great candidate for: http://wiki.linuxcnc.org/emcinfo.pl?TroubleShooting -- Ben Jackson AD7GD b...@ben.com http://www.ben.com/ -- Write once. Port to many. Get the SDK and tools to simplify cross-platform app development. Create new or port existing apps to sell to consumers worldwide. Explore the Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join http://p.sf.net/sfu/intel-appdev ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Effect of isolcpus boot parameter with D510 cpu ( was Re: hal taking machine down)
On Tuesday, January 03, 2012 02:41:08 PM Kent A. Reed did opine: On 1/2/2012 11:24 PM, Kent A. Reed wrote: On 1/2/2012 10:35 PM, Jon Elson wrote: gene heskett wrote: Is anyone running emc on this? How is the latency? I've tested the Intel D525MW motherboard, and put it in the database. The servo thread jitter was 15595, the base thread was 11921. That is fine for any system with a hardware interface (servo or stepper) but may be a little marginal for software step generation. Since I'm a servo bigot, it was quite good for me. The on-board parallel port passed all the tests with my EPP devices. EMC 2.4.x with Ubuntu 10.04 loads and runs with no quirks at all. I did the install on a 16 GB SSD hard drive. Jon Jon: This is a multi-cpu board. Did you set the isolcpus boot parameter? The reason I ask is IIRC I got similar numbers with a standard boot from the LiveCD on my ASUS board and saw them drop below 1 when I set isolcpus=1. This may be a false memory (I did the tests last winter) and I'm away from the board at the moment. I'll do the tests again in the morning and this time I'll post the results on the Wiki. Regards, Kent Jon (and Gene if he gets a motherboard outfitted with an Intel Atom cpu): On order A/O 15 minutes ago. But this didn't get sent, snow needed shoveling, so further update since. I put the better PSU in, but had problems getting a clean boot, and once I did get one, uptime was in minutes, 2 more times. Going to my IR thermometer because gkrellm said the hard drive was pretty warm, I noted that all the caps in the cpu supply on the mobo had tops all swelled up. All 6 of them. Duh, perfect explanation, everything fits. They can be replaced as Dave keeps a grocery sack of those at the tv station, all I have to do is get it there, or just go get some since I have the tools to replace them myself. But with the D525 already coming, it can wait till I need another box for the lathe. I did develop some interesting data vis-a-vis the usb mouse and keyboard. Yes, they hurt latency, as much as 15 u-secs worth. Unforch, my junk box did not disgorge any ps2 mice to facilitate a test for those. So it will be interesting to repeat those tests with the new box when it gets here. To follow up on this, I just ran 15-minute latency tests on my ASUS AT5NM10-I board first with a stock boot and then with isolcpus=1 as a boot parameter. I've posted my results in the table on the Wiki. In brief, isolating one cpu shaved approximately 6000ns off the max jitters reported for both the base thread and the servo thread. Of course the latency test should be run much longer but I believe these numbers are indicative. YMMV Regards, Kent -- Write once. Port to many. Get the SDK and tools to simplify cross-platform app development. Create new or port existing apps to sell to consumers worldwide. Explore the Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join http://p.sf.net/sfu/intel-appdev ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users Cheers, Gene -- There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order. -Ed Howdershelt (Author) My web page: http://coyoteden.dyndns-free.com:85/gene Although the moon is smaller than the earth, it is farther away. -- Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex infrastructure or vast IT resources to deliver seamless, secure access to virtual desktops. With this all-in-one solution, easily deploy virtual desktops for less than the cost of PCs and save 60% on VDI infrastructure costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Effect of isolcpus boot parameter with D510 cpu ( was Re: hal taking machine down)
gene heskett wrote: I put the better PSU in, but had problems getting a clean boot, and once I did get one, uptime was in minutes, 2 more times. Going to my IR thermometer because gkrellm said the hard drive was pretty warm, I noted that all the caps in the cpu supply on the mobo had tops all swelled up. All 6 of them. Duh, perfect explanation, everything fits. Yes, of COURSE! For motherboards of a certain vintage this is a REALLY well-known problem. I've done a couple of them myself at work. Jon -- Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex infrastructure or vast IT resources to deliver seamless, secure access to virtual desktops. With this all-in-one solution, easily deploy virtual desktops for less than the cost of PCs and save 60% on VDI infrastructure costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] Effect of isolcpus boot parameter with D510 cpu ( was Re: hal taking machine down)
On Tuesday, January 03, 2012 08:27:52 PM Jon Elson did opine: gene heskett wrote: I put the better PSU in, but had problems getting a clean boot, and once I did get one, uptime was in minutes, 2 more times. Going to my IR thermometer because gkrellm said the hard drive was pretty warm, I noted that all the caps in the cpu supply on the mobo had tops all swelled up. All 6 of them. Duh, perfect explanation, everything fits. Yes, of COURSE! For motherboards of a certain vintage this is a REALLY well-known problem. I've done a couple of them myself at work. Jon I figured that would ring a bell or 6. This particular board however, is about 18 months newer that that certain very famous vintage you refer to. OTOH, for highly stressed parts, which those are, it is like asking if the cement will crack, because it is not a question of if at all, its when with the when largely timed by the quality of the OEM parts. The last I had cement poured, supposedly 5k psi stuff about 6 to 7 inches thick, took about 2 months. :) This one was a 'lifetime warranted' Mach Speed Venom V4DP board, boxed version. But according to Murphy their interpretation of lifetime is for the lifetime of the company. They still exist, but now the warranty is for 1 year to the reseller, and I guess that's me. OTOH, I have gotten about 6 years out of it, so I guess it doesn't owe me a lot. Turning it in for a warranty after all this time would just make me look like a jerk. I will give that board credit though for being a very cleanly designed and timed board. I took the Athlon XP1400 cpu off a biostar board that had failed from this famous capacitor viri, and where that XP1400 had never run under 140F and if I didn't keep the 6k rpm fan sink spotless, it could get to 180F here in an air conditioned house. That board actually died when an nvidia video card crowbared the agp bus cooked some logic. So I pulled the memory and the cpu and put then in this Mach-Speed Venom, at which point I found that it was running the XP1400 at 1600 and couldn't be set any slower. Funny thing, the cpu only ran maybe 10F over room temps, and yet today, the hottest items in that case are the HD standing vertical just behind the front panel, and some little TSOP on the ATI X-1650 video card that is in it now. IMO you cannot reduce temps on that sort of a scale if that biostar board wasn't fighting with itself and eating 20 amps just from two drivers trying to drive the bus in opposite directions for 3ns out of 10. When everything is running like its all solid cmos logic, no heating, you have a truly excellent logic design on a board that is cycling its memory at 333mhz. For that reason alone, this board _will_ eventually be fixed. Just not in the next 3 or 4 days. Cheers, Gene -- There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order. -Ed Howdershelt (Author) My web page: http://coyoteden.dyndns-free.com:85/gene Practice is the best of all instructors. -- Publilius -- Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex infrastructure or vast IT resources to deliver seamless, secure access to virtual desktops. With this all-in-one solution, easily deploy virtual desktops for less than the cost of PCs and save 60% on VDI infrastructure costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users