Project Home
Project Home
Documents
Documents
Wiki
Wiki
Discussion Forums
Discussions
Project Information
Project Info
Forum Topic - Very wierd int latency observed with P35 based motherboard: (8 Items)
   
Very wierd int latency observed with P35 based motherboard  
I am testing our real-time application on a P35 + ICH9 motherboard.  QNX boots fine.  A newer devb-eide I was given runs
 the ICH9 in DMA as well.  Everything looks great, but when I run my real-time application I see a very long interrupt 
latency occuring about every 2 seconds on the a/d converter interrupt.  I have made sure I have a clean IRQ with nothing
 else sharing (tested both IRQ 3 and 4).  This is a Gigabyte GA-P35-S3G.

The same test with the same conditions and hardware (just swapping the motherboards) works fine on a Gigabyte GA-945PL-
S3E which uses the 945 chipset + ICH7.  Both tests are using an E4500 Core 2 Duo and the non-SMP kernel.

Since it is very repeatable with this 2 second interval it seems like some kind of timer.  Any help appreciated.
Re: Very wierd int latency observed with P35 based motherboard  
> I am testing our real-time application on a P35 + ICH9 motherboard.  QNX boots
>  fine.  A newer devb-eide I was given runs the ICH9 in DMA as well.  
> Everything looks great, but when I run my real-time application I see a very 
> long interrupt latency occuring about every 2 seconds on the a/d converter 
> interrupt.  I have made sure I have a clean IRQ with nothing else sharing (
> tested both IRQ 3 and 4).  This is a Gigabyte GA-P35-S3G.
> 
> The same test with the same conditions and hardware (just swapping the 
> motherboards) works fine on a Gigabyte GA-945PL-S3E which uses the 945 chipset
>  + ICH7.  Both tests are using an E4500 Core 2 Duo and the non-SMP kernel.
> 
> Since it is very repeatable with this 2 second interval it seems like some 
> kind of timer.  Any help appreciated.


Make sure you disable all USB emulation in the BIOS.  It could the BIOS that's taking over.
Re: Very wierd int latency observed with P35 based motherboard  
The fact the BIOS could cause this after QNX is up and running is not something I would have thought.  I have ordered an
 MSI P35 motherboard to test.  It has a different BIOS.
Re: Very wierd int latency observed with P35 based motherboard  
On Tue, Apr 08, 2008 at 11:23:23PM -0400, Art Hays wrote:
> The fact the BIOS could cause this after QNX is up and running is not
something I would have thought.  I have ordered an MSI P35 motherboard to
test.  It has a different BIOS.

Google for "system management mode" and read up on the resulting evilness
:-).

-- 
Brian Stecher (bstecher@qnx.com)        QNX Software Systems
phone: +1 (613) 591-0931 (voice)        175 Terence Matthews Cr.
       +1 (613) 591-3579 (fax)          Kanata, Ontario, Canada K2M 1W8
Re: Very wierd int latency observed with P35 based motherboard  
This isn't the whole story.  I put the ClockCycles() call before and after my InterruptWait() call which is waiting on a
 1KHz square wave.  Sometimes I get 7mS latency.  However, this never happens on QNX 6.3.1 with the same hardware and 
the same executable.
Re: Very wierd int latency observed with P35 based motherboard  
> The fact the BIOS could cause this after QNX is up and running is not 
> something I would have thought.  I have ordered an MSI P35 motherboard to test
> .  It has a different BIOS.


Don't think that will help, most BIOS have the dreaded SMI.  Mostly in desktop or server class board.   Your best bet is
 try disabling any "emulation" option you can find as well as power saving feature, automatic fan control etc...



Re: Very wierd int latency observed with P35 based motherboard  
Thanks for the info on SMM.  I know that three different Gigabyte P35 motherboards exhibit this behavior. I turned off 
everything that could be turned off in the BIOS to no avail.  I will ask Gigabyte about it.
Re: Very wierd int latency observed with P35 based motherboard  
Three different Gigabyte P35 motherboards showed the same latency issue.  I tried an MSI P35 motherboard and it did not 
show the issue.  So I don't think the problem is with the P35 chipset.  SMM seems most likely.

I am going to pursue testing industrial ATX motherboards now.  These companies can guarantee product life cycles of 4-8 
years.  If I find one that works we will be able to purchase them for far longer than consumer models.