RTAPI error and realtime delay at EMC launch
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
27 Dec 2010 22:17 #6274
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
Thank you, John. I will do the try before you buy when the time comes (I'm still designing the gantry for the plasma table).
I had gotten a link to that page from Andy and had tried to disable SMI but got the message:
error inserting (path to rtai_smi.ko): -1 No such device 2641
Searching for that message online (google.com) gave me few results (including a couple of links to this thread) and nothing that seemed relevant to my problem (inexact matches for parts of the search string which finally was: error inserting (path to rtai_smi.ko).
Gonna move to the next Linux fodder...I've got a Dell that gave me some trouble (going to go back and see what it was...I forget now)...then I have an HP Pavilion with a P4 and one with a Celeron. If I don't get one that will work (or I get bored playing with this stuff) I'll build a system based on the board you liked and have had good results with.
As for the screensaver, I could disable it but the instructions for the Latency Test said to "abuse the machine" so I thought that would have been a good test since I don't have GLXGears.
Thanks again.
--HC
I had gotten a link to that page from Andy and had tried to disable SMI but got the message:
error inserting (path to rtai_smi.ko): -1 No such device 2641
Searching for that message online (google.com) gave me few results (including a couple of links to this thread) and nothing that seemed relevant to my problem (inexact matches for parts of the search string which finally was: error inserting (path to rtai_smi.ko).
Gonna move to the next Linux fodder...I've got a Dell that gave me some trouble (going to go back and see what it was...I forget now)...then I have an HP Pavilion with a P4 and one with a Celeron. If I don't get one that will work (or I get bored playing with this stuff) I'll build a system based on the board you liked and have had good results with.
As for the screensaver, I could disable it but the instructions for the Latency Test said to "abuse the machine" so I thought that would have been a good test since I don't have GLXGears.
Thanks again.
--HC
Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
29 Dec 2010 03:13 #6299
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
Some more progress, some more information. I'm taking the time to post this so maybe it will help someone else some day.
I remembered why I didn't use the Dell...no onboard parallel port. The addon card I had was not recognized by Ubuntu 10.04 and I could not find drivers for it.
Next up was the HP P4 I mentioned in my last post. HP a1320y, as I recall, P4 D 3Ghz, 1.5 GB RAM. What a colossal failure. First, the system hung on bootup (POST, not Ubuntu) with my kick-butt Gigabyte video card. Hmm, okay, back to the factory onboard card. I had a SATA drive in the system. Loaded the Live CD with U10.04. Latency from HELL. Servo thread was showing numbers in the 600,000 range and up almost immediately whether I ran the test immediately after bootup or half an hour after bootup. Actually doing things like opening Firefox and moving windows on screen gave me over 700,000 for servo thread and, regardless of usage, I got over 300,000 for base thread. Attempting to disable SMI (System Management Interrupt or whatever) did not work.
F* F* F*
Hmm. Okay, I was using the same HD from the other problem-child I'd originally posted about. I grabbed a EIDE WD 40GB 7200 RPM disk I had laying around and tried that. It was magical. My servo thread jumped to over 1,000,000...with some results over 2,000,000 and higher base thread numbers, too. Not a happy camper.
I read about the SMI problem and about it being included with some Intel chipsets. So, I decided, fine, I'll give Ubuntu and old PC and see what happens (maybe an older system wouldn't have the SMI on it).
I have a system that is about 8 years old I guess. P4 1.8 Ghz, single core. I had tried to mix-n-match RAM and had 1.75 (roughly) GB installed...1GB stick, 512MB stick, and a 256MB stick but that didn't work, the system only saw the 768MB. No biggie. Latency tests showed < 35,000 for the servo thread...even though it's running on a litter-mate WD 40 GB 7,200 RPM drive and on the onboard video. I shut the system down and removed the 512 and 256MB sticks and the system then used the 1GB stick. Latency times dropped a little more (maybe because the 1GB stick is PC3200 IIRC and the 768 combined sticks were PC133). Now running the test for 10 minutes+ showing servo thread latency of 32,405. MUCH improved over the HP with the dual core P4 and 1.5 GB RAM with a SATA drive (or the EIDE drive). Go figure that one out.
The motherboard in this old system is an ASUS P4S533-MX.
Latency on this system is: Servo 99,308 and Base 32,405.
EMC2 loads with no errors about latency or RTAI and Axis runs/loads fine.
So, I'm wondering...if the SMI stuff is only prevalent on Intel chipsets (and maybe I'm being ignorant of chipset manufacturers here), why couldn't I (we) try using a system based on an AMD processor? Do they use Intel chipsets? Sorry, lonnnnng time, no hardware monkeying.
I am considering just buying one of the Intel D510MO that BigJohnT has had good luck with. This stuff with Ubuntu/EMC2 was supposed to help me with a grassroots approach to CNC but it's being a pain with the hardware problems I've had.
Thanks to those who have helped me. I hope this will give a little back.
--HC
I remembered why I didn't use the Dell...no onboard parallel port. The addon card I had was not recognized by Ubuntu 10.04 and I could not find drivers for it.
Next up was the HP P4 I mentioned in my last post. HP a1320y, as I recall, P4 D 3Ghz, 1.5 GB RAM. What a colossal failure. First, the system hung on bootup (POST, not Ubuntu) with my kick-butt Gigabyte video card. Hmm, okay, back to the factory onboard card. I had a SATA drive in the system. Loaded the Live CD with U10.04. Latency from HELL. Servo thread was showing numbers in the 600,000 range and up almost immediately whether I ran the test immediately after bootup or half an hour after bootup. Actually doing things like opening Firefox and moving windows on screen gave me over 700,000 for servo thread and, regardless of usage, I got over 300,000 for base thread. Attempting to disable SMI (System Management Interrupt or whatever) did not work.

Hmm. Okay, I was using the same HD from the other problem-child I'd originally posted about. I grabbed a EIDE WD 40GB 7200 RPM disk I had laying around and tried that. It was magical. My servo thread jumped to over 1,000,000...with some results over 2,000,000 and higher base thread numbers, too. Not a happy camper.
I read about the SMI problem and about it being included with some Intel chipsets. So, I decided, fine, I'll give Ubuntu and old PC and see what happens (maybe an older system wouldn't have the SMI on it).
I have a system that is about 8 years old I guess. P4 1.8 Ghz, single core. I had tried to mix-n-match RAM and had 1.75 (roughly) GB installed...1GB stick, 512MB stick, and a 256MB stick but that didn't work, the system only saw the 768MB. No biggie. Latency tests showed < 35,000 for the servo thread...even though it's running on a litter-mate WD 40 GB 7,200 RPM drive and on the onboard video. I shut the system down and removed the 512 and 256MB sticks and the system then used the 1GB stick. Latency times dropped a little more (maybe because the 1GB stick is PC3200 IIRC and the 768 combined sticks were PC133). Now running the test for 10 minutes+ showing servo thread latency of 32,405. MUCH improved over the HP with the dual core P4 and 1.5 GB RAM with a SATA drive (or the EIDE drive). Go figure that one out.

The motherboard in this old system is an ASUS P4S533-MX.
Latency on this system is: Servo 99,308 and Base 32,405.
EMC2 loads with no errors about latency or RTAI and Axis runs/loads fine.
So, I'm wondering...if the SMI stuff is only prevalent on Intel chipsets (and maybe I'm being ignorant of chipset manufacturers here), why couldn't I (we) try using a system based on an AMD processor? Do they use Intel chipsets? Sorry, lonnnnng time, no hardware monkeying.
I am considering just buying one of the Intel D510MO that BigJohnT has had good luck with. This stuff with Ubuntu/EMC2 was supposed to help me with a grassroots approach to CNC but it's being a pain with the hardware problems I've had.
Thanks to those who have helped me. I hope this will give a little back.
--HC
Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
29 Dec 2010 15:02 #6312
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
A little more about this system's performance.
I left the Latency Test running all night. Last night I kicked off the install for Open Office suite. This morning I let Update Manager download and install about 215 updates for a total of about 173 MB. While that downloaded and installed I searched for computer parts (for a new Linux box and a replacement hard drive for one of my laptops).
I went to amazon com, newgg com, walmart com, intel com, and some result sites I found with google com. I had as many as 6 browser (Firefox) windows open, some with multiple tabs, all while the downloading and installing was running. They system is fun-fast to use, but it's usable. However, the Servo thread reports 74,000 at this time for the max and Base thread reports just under 35,000. Up until a few minutes ago they were both under 20k. I suspect that the jump came when the updates finished and Firefox notified me that it needs to be restarted.
Much better than the P4 dual core HP a1320y I had up on the desk before with 1M+ for Servo and comparatively high Base thread.
I ordered:
Intel D510MO motherboard N82E16813121399
Crucial 4 GB memory kit N82E16820148160
Western Digital Caviar *black* N82E16822136320
Rosewill 500W power supply N82E16817182076
Rosewill case N82E16811147023
HP external DVD/CD ROM N82E16827140077
I have a few internal DVD/CD drives around but elected to buy an external with the hopes I can use it to reload the OS on my Dell Mini 9 which has no optical drive. I also already have the header connector for the parallel port so I didn't need to order one of those.
Total came to $338.64 including UPS 3 day shipping to north central Texas.
I'll build it when it arrives and post my results.
--HC
I left the Latency Test running all night. Last night I kicked off the install for Open Office suite. This morning I let Update Manager download and install about 215 updates for a total of about 173 MB. While that downloaded and installed I searched for computer parts (for a new Linux box and a replacement hard drive for one of my laptops).
I went to amazon com, newgg com, walmart com, intel com, and some result sites I found with google com. I had as many as 6 browser (Firefox) windows open, some with multiple tabs, all while the downloading and installing was running. They system is fun-fast to use, but it's usable. However, the Servo thread reports 74,000 at this time for the max and Base thread reports just under 35,000. Up until a few minutes ago they were both under 20k. I suspect that the jump came when the updates finished and Firefox notified me that it needs to be restarted.
Much better than the P4 dual core HP a1320y I had up on the desk before with 1M+ for Servo and comparatively high Base thread.
I ordered:
Intel D510MO motherboard N82E16813121399
Crucial 4 GB memory kit N82E16820148160
Western Digital Caviar *black* N82E16822136320
Rosewill 500W power supply N82E16817182076
Rosewill case N82E16811147023
HP external DVD/CD ROM N82E16827140077
I have a few internal DVD/CD drives around but elected to buy an external with the hopes I can use it to reload the OS on my Dell Mini 9 which has no optical drive. I also already have the header connector for the parallel port so I didn't need to order one of those.
Total came to $338.64 including UPS 3 day shipping to north central Texas.
I'll build it when it arrives and post my results.
--HC
Please Log in or Create an account to join the conversation.
- psha
- Offline
- Moderator
-
Less
More
- Posts: 99
- Thank you received: 1
29 Dec 2010 17:52 #6314
by psha
Replied by psha on topic Re:RTAPI error and realtime delay at EMC launch
Just FYI
Setups on D510MO boars are happy with 65W power supplies -- board is eating ~20-25W
So 500W supply is overkill
Setups on D510MO boars are happy with 65W power supplies -- board is eating ~20-25W

So 500W supply is overkill

Please Log in or Create an account to join the conversation.
- andypugh
-
- Offline
- Moderator
-
Less
More
- Posts: 23382
- Thank you received: 4967
29 Dec 2010 20:43 #6318
by andypugh
Replied by andypugh on topic Re:RTAPI error and realtime delay at EMC launch
I am running with a 20W PicoPSU and using a 4GB DOM SATA solid state drive. No moving parts at all.
Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
12 Jan 2011 04:35 #6639
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
Well, I like to go big or stay home. And the extra power will help with the flux-core capacitor I'm developing for my CNC plasma table. <GRIN>
Honestly, I didn't even consider the power consumption of the board or other components when I ordered the PS; I just wanted something that worked, having the necessary connectors for a modern motherboard (compared to my antiquated PS's from old systems). I'm now running on that system (just assembled it this evening and this is my first boot since load about 20 minutes ago). I would agree that the PS is overkill in at least the fact that it has no less than 6 power connectors I'm not using (about 5 SATA IIRC and a couple of four-pin connectors for PCI-E and maybe a couple others). But it's working...and I suppose that if the devices aren't pulling the current then the wall-draw should be less than the max the PS is rated to put out/pull.
My major concern with selecting a PS was simply that I wanted it to fit into the case I bought and I wasn't sure what different standards were at play these days (it's been 10 years now since I built a computer from scratch, having switched to just buying Dell or HP computers).
I'm off to write up my findings on the new system (trying to help others who might come along and read this).

--HC
Honestly, I didn't even consider the power consumption of the board or other components when I ordered the PS; I just wanted something that worked, having the necessary connectors for a modern motherboard (compared to my antiquated PS's from old systems). I'm now running on that system (just assembled it this evening and this is my first boot since load about 20 minutes ago). I would agree that the PS is overkill in at least the fact that it has no less than 6 power connectors I'm not using (about 5 SATA IIRC and a couple of four-pin connectors for PCI-E and maybe a couple others). But it's working...and I suppose that if the devices aren't pulling the current then the wall-draw should be less than the max the PS is rated to put out/pull.
My major concern with selecting a PS was simply that I wanted it to fit into the case I bought and I wasn't sure what different standards were at play these days (it's been 10 years now since I built a computer from scratch, having switched to just buying Dell or HP computers).
I'm off to write up my findings on the new system (trying to help others who might come along and read this).

--HC
Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
12 Jan 2011 04:43 #6640
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
What happened to when guys bragged about how big and powerful things were?

Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
12 Jan 2011 05:05 #6641
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
Hello, all, I have finally built the PC with the components I listed that I ordered. I am on that system now. I used to build PC's back in the late 90's to 2000 but not since then. It still only took me about 45 minutes to build the thing and that included a phone call and some cold beer and having fun doing it. Nothing to it.
The assembly was fairly easy, even with my 10 year hiatus from doing such projects. The Rosewill case is pretty cool, not requiring screws for the 3.5 or 5.25 inch bays. The case has two header connectors for the 4 front panel USB ports. I connected both to the MB, but one had to go to the Front Panel USB with Flash Drive Support connector and I'm not sure it will work. No big deal, at least two of the front panel USB ports will work. Everything else was pretty easy. The case I ordered has two case fans but the MB has only one connector for a case fan so I left the front one disconnected.
The system seems to only recognize 3.5 GB of RAM and 256MB are used for the video card so Ubuntu is reporting 3.2 GB of RAM. Not really surprising and no big deal.
Installation of Ubuntu was a cake walk. It seemed slower off the USB CD ROM drive versus an internal SATA or even IDE CD ROM drive but no big deal.
Once loaded, the system is fast enough to be pleasant to use for surfing the web and reading e-mail and such.
The exciting news is that the latency is really low. Immediately after booting after loading I started the Latency Test and ran it while I downloaded the ad-blocking hosts file from MVPS.org and copied that to my /etc folder (with the required addition of the hostname of my Linux box and changing the filename from upper to lower case). Ads blocked, I checked a couple of my e-mail accounts, logged on here, and began downloading the product guide from Intel for this motherboard (interesting to me is that the product guide says it is a dual-core processor (and only one) yet the system monitor shows to have four cores). I saw that one core was at 100% proc usage and stopped the Latency Test to see if that was the cause (it was). At that time the latency for the Servo Thread was 10,383 and the Base Thread was 12,576. I restarted the Latency Test and have been running it for about 10 minutes now with 5 instances of Firefox, 1 instance of GEdit, system monitor, 1 PDF open, and Update Manager downloading a bunch of updates and still have low latency (9,230 for the Servo Thread and 11,904 for the Base Thread).
I have the Visual Effects at Normal, where they were after installation. I don't know yet if turning them off will lower the Latency.
Very happy here. New stuff, low latency, no hassles with the old PC's. I'd recommend this setup and hopefully the shopping list will help, even if the PS is overkill (no ill-intent to the guy who pointed that out).
Tomorrow I'll actually run the engraving table and see how that goes but I expect no problems. I'll post back.
Thank you to all who helped here and thanks to BigJohnT for posting about the D510MO board.
--HC
The assembly was fairly easy, even with my 10 year hiatus from doing such projects. The Rosewill case is pretty cool, not requiring screws for the 3.5 or 5.25 inch bays. The case has two header connectors for the 4 front panel USB ports. I connected both to the MB, but one had to go to the Front Panel USB with Flash Drive Support connector and I'm not sure it will work. No big deal, at least two of the front panel USB ports will work. Everything else was pretty easy. The case I ordered has two case fans but the MB has only one connector for a case fan so I left the front one disconnected.
The system seems to only recognize 3.5 GB of RAM and 256MB are used for the video card so Ubuntu is reporting 3.2 GB of RAM. Not really surprising and no big deal.
Installation of Ubuntu was a cake walk. It seemed slower off the USB CD ROM drive versus an internal SATA or even IDE CD ROM drive but no big deal.
Once loaded, the system is fast enough to be pleasant to use for surfing the web and reading e-mail and such.
The exciting news is that the latency is really low. Immediately after booting after loading I started the Latency Test and ran it while I downloaded the ad-blocking hosts file from MVPS.org and copied that to my /etc folder (with the required addition of the hostname of my Linux box and changing the filename from upper to lower case). Ads blocked, I checked a couple of my e-mail accounts, logged on here, and began downloading the product guide from Intel for this motherboard (interesting to me is that the product guide says it is a dual-core processor (and only one) yet the system monitor shows to have four cores). I saw that one core was at 100% proc usage and stopped the Latency Test to see if that was the cause (it was). At that time the latency for the Servo Thread was 10,383 and the Base Thread was 12,576. I restarted the Latency Test and have been running it for about 10 minutes now with 5 instances of Firefox, 1 instance of GEdit, system monitor, 1 PDF open, and Update Manager downloading a bunch of updates and still have low latency (9,230 for the Servo Thread and 11,904 for the Base Thread).
I have the Visual Effects at Normal, where they were after installation. I don't know yet if turning them off will lower the Latency.
Very happy here. New stuff, low latency, no hassles with the old PC's. I'd recommend this setup and hopefully the shopping list will help, even if the PS is overkill (no ill-intent to the guy who pointed that out).
Tomorrow I'll actually run the engraving table and see how that goes but I expect no problems. I'll post back.
Thank you to all who helped here and thanks to BigJohnT for posting about the D510MO board.
--HC
Please Log in or Create an account to join the conversation.
- andypugh
-
- Offline
- Moderator
-
Less
More
- Posts: 23382
- Thank you received: 4967
12 Jan 2011 10:51 #6643
by andypugh
Replied by andypugh on topic Re:RTAPI error and realtime delay at EMC launch
hcb wrote:
That's interesting. I see about half that with exactly the same motherboard. (Not that there is anything wrong with 12uS latency.)
I don't know if you have isolcpus=1 as a default kernel option, but if not it might be interesting to try it.
I _think_ I only see 2 CPUs on mine, so it might be that you have Hyperthreading on in the BIOS, so you could try turning that off too.
Or, you could just get on with the machine, 12k is extremely good in absolute terms.
At that time the latency for the Servo Thread was 10,383 and the Base Thread was 12,576.
That's interesting. I see about half that with exactly the same motherboard. (Not that there is anything wrong with 12uS latency.)
I don't know if you have isolcpus=1 as a default kernel option, but if not it might be interesting to try it.
I _think_ I only see 2 CPUs on mine, so it might be that you have Hyperthreading on in the BIOS, so you could try turning that off too.
Or, you could just get on with the machine, 12k is extremely good in absolute terms.
Please Log in or Create an account to join the conversation.
- hcb
- Offline
- Junior Member
-
Less
More
- Posts: 28
- Thank you received: 1
02 Feb 2011 04:10 #6951
by hcb
Replied by hcb on topic Re:RTAPI error and realtime delay at EMC launch
Hey, Andy, thanks for the reply. I'm sorry for the delay...I've been obsessing about trying to design a CNC plasma table...when I close my eyes I see gear reduction ratios and linear bearings.
Anyway, I just got back to the Linux box here a couple of days ago and then fired up the Latency Test. I have been streaming music from DI.fm and had more browser windows open than you can shake as stick at (about 12, actually), and some PDF files and let the system run for about 36 hours and saw Servo hit 17,365 and Thread hit 15,313. Pretty high for the system so far since I built it. I had no idea about Hyperthreading being a problem with latency and realtime...I'm sorry, maybe I missed that somewhere along the way or maybe it's not been addressed in any of the postings I've read, I'm not sure. Regardless, I recorded the numbers and rebooted and turned HT off. I now see only two procs in the System Monitor. Currently I have three instances of Firefox running (DI.fm, McMaster-Carr, and here), Latency test, and Totem Movie Player (playing classic trance streamed from DI.fm). Numbers for LT are Servo 10,676 and Base 11,539.
I can spell Linux most days and sudo some commands on good days...Linux/Unix is not my area of expertise at all, I'm afraid. I will have to do some looking to learn what you mean about the isolcpus setting. I'm sorry, just not that good with 'nix. Off to Google...I'll post back.
Thanks again.
--HC
Anyway, I just got back to the Linux box here a couple of days ago and then fired up the Latency Test. I have been streaming music from DI.fm and had more browser windows open than you can shake as stick at (about 12, actually), and some PDF files and let the system run for about 36 hours and saw Servo hit 17,365 and Thread hit 15,313. Pretty high for the system so far since I built it. I had no idea about Hyperthreading being a problem with latency and realtime...I'm sorry, maybe I missed that somewhere along the way or maybe it's not been addressed in any of the postings I've read, I'm not sure. Regardless, I recorded the numbers and rebooted and turned HT off. I now see only two procs in the System Monitor. Currently I have three instances of Firefox running (DI.fm, McMaster-Carr, and here), Latency test, and Totem Movie Player (playing classic trance streamed from DI.fm). Numbers for LT are Servo 10,676 and Base 11,539.
I can spell Linux most days and sudo some commands on good days...Linux/Unix is not my area of expertise at all, I'm afraid. I will have to do some looking to learn what you mean about the isolcpus setting. I'm sorry, just not that good with 'nix. Off to Google...I'll post back.
Thanks again.
--HC
Please Log in or Create an account to join the conversation.
Time to create page: 0.078 seconds