After install I get applet removal error prompts

Sounds like a plan. I'll install 8.04/2.3, run the latency test, then upgrade to 2.5.1, and then run the latency test again. If no problems with latency numbers after much effort to break it by running lots of apps and file moves, then SMI is probably not an issue.

Can I assume that 256MB SDRAM is sufficient, or should I shelve this project until I'm rockin' at least 512MB in the system? I can get a 512MB memory card from a local computer shop tomorrow morning if it's worth the trip and the $30 (yeah, I know.. It's much cheaper on Amazon if I'm willing to wait a week).

Thanks again for your help!
Peter

Hi

256MB is not really enough, it will be painfully slow due to swap outs to HDD
Buy a second hand 512MB stick from one of the computer breakers on Ebay, makes no sense to buy new, it is expensive because it is not produced in any quantity now.
Then you will have 768MB which will be plenty.

I am rather sceptical about this SMI / video card linkage, both can cause spikes but they are unlikely to be connected.
SMI is essentially about power management, CPU monitoring and cooling etc.

Do lspci -vv > lspci.txt in a terminal and attach the generated lspci.txt file and I can tell you if you have an affected chipset.
Loading the module to fix it is not difficult, so worry about that when / if it becomes necessary.

The main thing to do is install 8.04 and do a latency test.
Don't bother running Firefox, it will lock everything up with that amount of memory, just 3 instances of glxgears, move windows around forcing repaints.
That should test the video card and openGL for now, you can do more when you have full memory.

If you get high readings, read the first FAQ on latency and run the latency test in the realtime test suite
Then you can see when the spikes are occurring and whether random or regular etc

regards

I'll go buy the memory today, even though overpriced. Gotta get this project moving... I'll see if they have an old machine I can pop a stick out of for less $$$.

ArcEye: I'll follow your plan and report back with the ispci.txt file. Thanks!

Okay, the 256MB memory limitation has been fixed. I bought two 512MB sticks and let the LiveCD's memory test application run for a few hours to beat on the entire 1GByte. No errors.

I loaded Ubuntu 8.04/LinuxCNC 2.3 from my hash-checked LiveCD (without errors) and it appears to have installed as I intended, with half the hard drive partitioned for Ubuntu and the other half for Windows XP (so I can still use Mach3 if desired).

The file attachment below is what I got when I typed "lspci -v > lspci.txt" in a terminal window. Does my machine have SMI cooties?

Here are my latency numbers (haven't upgraded to 2.5.1 or really stressed the processor much yet):
Servo (1msec) Max Jitter = 97209 (yes, ninety-seven-thousand nanoseconds)
Base (25usec) Max Jitter = 104868

According to the "Hardware That Works" table for the Optiplex GX240, the numbers given are: Servo=13698 and Base=14714. I'm 10 times higher... Uggh. Hope the SMI fix is all this beatie needs.

Regards,
Peter

Quick update (and query): I reset the latency numbers and now I can't seem to get them to go above Servo=11324 and Base=15273, even with a bunch of stuff running, opening and closing and file copies, file searching and etc...

Now I don't know whether to trust these new numbers or keep stressing it to see if the 10x figures come back. How long should I run it?

Hi

Your chipset is 82801BA (rev 12)

The 82801BA is listed in the SMI module, but only version 0 and version 10,.

The fact that you can run with low figures suggests there is no SMI problem, SMI spikes typically at 32 or 64 second intervals and does not stop.

I should run the realtime test-suite latency test I detailed and run 2 sorts of tests
Run one that is very heavily video based, 6 x glxgears, a lot of window moves across other windows to force repaints and make the graphics card work hard

Run another more like your last test but with a couple of glxgears running and a terminal running find /* > /dev/null which will tie up the disc finding every file on your system.

Leave them both for a couple of hours after initial monitoring

I can't seem to get them to go above Servo=11324 and Base=15273,

These are very usable figures and if they continue, you have a good basis for your system.

regards

LATENCY: Well.... bad news. All I did was power up into Ubuntu, start the Latency test program, then open Calculator, Terminal and three OpenOffice programs. Latency figures immediately spiked to Servo=97K and Base=127K (see attached screenshot). I'll work on disabling SMI today and see if that fixes it.


DISABLE SMI: I see in a forum entry containing the following instructions for disabling SMI for the 82801BA chipset:

Intel ® 82801BA I/O Controller Hub 2 (ICH2)
==================================
Data Sheet URL: www.intel.com/design/chipsets/datashts/29068702.pdf
Instructions: 1. The GBL_SMI_EN bit in the SMI_EN register must be cleared.
Comments: The Linux watchdog "Intel i810 TCO timer" may be used with the SMIs disabled for these two chipsets.

Is this what the FixingSMIIssues help page is trying to do? It's so cryptic that I can't decipher what to actually DO. Is there a better help page for disabling SMI?


UPDATES: I tried to load the updates (not the "Upgrade" button) available for 8.04 but the nine links Ubuntu was looking for don't work (example: one had a .10 in the middle of it but the destination wewbsite has .12 instead). I see that the version numbers from the websites the updater is trying to access are newer than the ones the Updater is trying to download. How do I resolve this or do I just NOT try to do any updating? Can I just change the APT line to match the latest updates (same as for upgrading EMC2 to LinuxCNC 2.5 shown here: wiki.linuxcnc.org/cgi-bin/wiki.pl?UpdatingTo2.5)?


EMC2 LOCKS UP: I ran EMC2 but it just locks up as the initial screens get halfway loaded. No clue what to do. I was hoping the updates might cure this, once installed.


Regards,
Peter

Hi

You seem to be taking the scatter gun approach

SMI - you need to read the FAQs I gave you links to and my previous posts and do a proper test.
Thus far you have not presented any evidence that you have a SMI problem. If you can see a regular spike every 32 - 64 seconds then it is likely to be SMI.
But as I said earlier, your chipset revision is not listed in the known offenders.

Don't try to do updates. Just concentrate upon doing the latency test and then running Linuxcnc.

EMC2 LOCKS UP: I ran EMC2 but it just locks up as the initial screens get halfway loaded.

What error messages did you get? Please post them.

Easiest way is to run emc from a terminal and cut and paste the error messages, plus include a copy of /var/log/dmsg from immediately after you try it.

Sounds like an openGL problem possibly, but without knowing exactly what you did and what else was running at the time hard to say for sure.
You can't run emc whilst doing a latency test, so hopefully that isn't what you did.

regards

Hi ArcEye,

Thanks for everyone's patience and help! I've spent many, many, many hours helping others (including dear old Mom) learn to use computers and software and it's nice to have some of that kharma come back to me when I need help myself. MUCH appreciated.

I'm really not intentionally trying everything in scatter-gun mode. I was just covering the bases in what I thought was a logical order: My thoughts were that if EMC2 locks up, then maybe it just needs the updates. And if the latency is 97K without even running glxgears, then I was seeing that, right or wrong, as the dreaded SMI problem everyone's been talking about. I see now that it might make more sense for it to be an OpenGL issue, as you mentioned, since the latency numbers don't automatically spike up every 32-64 seconds.


EMC2 LOCKS UP: There were NO error messages. It just locked up. The mouse still worked but the cursor was a black arrow and I couldn't click on anything anywhere on the screen. Right clicking didn't work either. I needed to power the system down via the power button on the front panel. I ran EMC2 from the terminal, per your request. It crashed but gave me the prompt back, so I typed "dmsg" to get the error report. I have attached the entire terminal output (pasted it into a .txt file) below.


SMI: No, EMC2 is never running while doing a latency test. That much I knew not to do. The spike in latency numbers is not at all regular every 32/64 seconds, and if I just let it sit there, the latency numbers stay below 14K/20K for hours on end. Understood that my chipset is not on the list of known offenders, but "disable SMI" is the recommended approach in the "Dell Optiplex GX240" entry in the table of various sample configurations and their measured latencies (at least if an ATI Rage card is being used, which it is). If the table is incorrect, it should be updated to reflect your information. Whatever spiked the latency to 97K is not happening very often.

I've been running eight instances of glxgears with overlapping screens, along with Firefox and Calculator for the last four hours. Every so often, I'd drag the screens across each other to cause maximum redraws. Latencies remained below 12K/14K until I opened Calculator. It rose to 12K/17K. Still great, though. No spikes to 97K as before. Attached is a screenshot of the numbers (I reset the stats at one point, so the 12K/17K numbers are currently gone).

I couldn't get your "find" command to work in the terminal window, so I tried typed "find /*" and it scrolled a list of files for several minutes. I kept entering the same command over and over. No change in latency (11K/14K) until I took the screenshot. Then, it jumped to 11K/19K.

Now, after another two hours, it seems stable at 14K/19K.

It's probably time to upgrade EMC2 to LinuxCNC 2.5.1 and repeat the stress testing. Sound like a good plan?

Hi

"disable SMI" is the recommended approach in the "Dell Optiplex GX240"...............If the table is incorrect, it should be updated to reflect your information.

Linuxcnc is an open source project and all information is collaborative.
The GX240 is a model not a set in stone MB and chipset. There have been at least 4 different processors available throughout its life, so it is likely that the person who made that entry had an earlier revision of the chipset.
I will add a rider to the entry. (EDIT: Now done)

You need to track down the problem with Axis not displaying
dmsg shows
[ 155.553277] rtapi: no version for "nano2count" found: kernel tainted.
[ 190.678962] RTAPI: ERROR: Unexpected realtime delay on task 1
[ 190.678967] This Message will only display once per session.
[ 190.678970] Run the Latency Test and resolve before continuing.
[ 5386.608782] RTAI[math]: unloaded.
[ 5386.708956] SCHED releases registered named ALIEN RTGLBH
[ 5386.748222] RTAI[malloc]: unloaded.
[ 5386.853314] RTAI[sched]: unloaded (forced hard/soft/hard transitions: traps 0, syscalls 0).
[ 5386.862315] I-pipe: Domain RTAI unregistered.
[ 5386.862339] RTAI[hal]: unmounted.

This does not help much, the realtime delay may be due to the lock up not latency per se.

You need to try running the simulator axis or axis_mm from the sim section of the configuration tree you are presented with when launching emc

Does that start?

If not find the config file axis.ini (or axis_mm.ini) change the entry in the [DISPLAY] section DISPLAY = AXIS to DISPLAY = xemc

See if that starts. If it does follow the instructions in the troubling shooting wiki for replacing hardware openGL with the software libraries, change DISPLAY back to AXIS and try again

It's probably time to upgrade EMC2 to LinuxCNC 2.5.1 and repeat the stress testing. Sound like a good plan?

There is no point upgrading until it will run, the older version works fine, just lacks some additions and refinements.

Sorry this is a bit rushed , I have to go out.

Hopefully this will get you going for now

regards