Hello.
We have a real-time application made with QNX4.25G and a National
Instruments acquisition board.
PCI bus is used only for accessing the NI card.
With Celeron slower than 733 (66 MHz FSB) system always works, with Celeron
faster (100 MHz FSB) sometime the acquisition task stops.
Motherboard is from Mitac and it is not possible to slow down any frequency.
Have you some suggestion for us?
“Davide” <utee@idrapresse.it> wrote in message
news:baluh3$9ut$1@inn.qnx.com…
Hello.
We have a real-time application made with QNX4.25G and a National
Instruments acquisition board.
PCI bus is used only for accessing the NI card.
With Celeron slower than 733 (66 MHz FSB) system always works, with
Celeron
faster (100 MHz FSB) sometime the acquisition task stops.
Could you be more specific as to what exactly stop, what happens?
Motherboard is from Mitac and it is not possible to slow down any
frequency.
Have you some suggestion for us?
Maybe some register in the NI card need some time in between access,
although that would be very strange. Have you tried this with different
motherboard/CPU to rule out it’s not a problem with the hardware.
Hello Mario,
thank you for your fast reply.
The situation is becoming “dangerous” for us, because we are sending
machines in all the world (USA, China, Europe) and they could have the
problem.
- The system HAS NOT a display, a keyboard, a mouse, and has a small (32
MB) Disk on module (few space for diagnostic, no compiler and debugger
present). - The communication is made by an UDPserver that seems to work properly
(TCP/IP licence is not installed). - We have the problem only 1-2 times every 2 days, so it is difficult to
diagnose. - If I substitute the 850 MHz Celeron with, for example, a 633 Mhz the
problem disappears. The limit seems to be the FSB of the processor.
OK, we could have a timing problem of the software, but before build a “test
system” in a customer application (VERY VERY VERY DIFFICULT!!!) we must be
sure that the operating system has not responsability in the problem.
In effect we note that, when the Operating system starts and we have a
monitor connected, the frequency of the celeron 566 (for example) is
reported correcly from the OS in the first screenshot (the monitor look
like 686/687 … 566 MHz). In the case of a Celeron 850 the frequency is
reported by the OS as 911 or 912 Mhz. The Pentium III (1 Ghz) reports 1092
Mhz. Has this frequency any influence in the PCI access?
What is your opinion?
Thank you
Davide Gardoni
Idra Casting Machines
d.gardoni@idracasting.com
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:baou0d$k1b$1@inn.qnx.com…
“Davide” <> utee@idrapresse.it> > wrote in message
news:baluh3$9ut$> 1@inn.qnx.com> …
Hello.
We have a real-time application made with QNX4.25G and a National
Instruments acquisition board.
PCI bus is used only for accessing the NI card.
With Celeron slower than 733 (66 MHz FSB) system always works, with
Celeron
faster (100 MHz FSB) sometime the acquisition task stops.Could you be more specific as to what exactly stop, what happens?
Motherboard is from Mitac and it is not possible to slow down any
frequency.
Have you some suggestion for us?Maybe some register in the NI card need some time in between access,
although that would be very strange. Have you tried this with different
motherboard/CPU to rule out it’s not a problem with the hardware.
\
“Davide” <utee@idrapresse.it> wrote in message
news:baqhaf$elh$1@inn.qnx.com…
Hello Mario,
thank you for your fast reply.
The situation is becoming “dangerous” for us, because we are sending
machines in all the world (USA, China, Europe) and they could have the
problem.
- The system HAS NOT a display, a keyboard, a mouse, and has a small (32
MB) Disk on module (few space for diagnostic, no compiler and debugger
present).- The communication is made by an UDPserver that seems to work properly
(TCP/IP licence is not installed).
Well it should -)
- We have the problem only 1-2 times every 2 days, so it is difficult to
diagnose.- If I substitute the 850 MHz Celeron with, for example, a 633 Mhz the
problem disappears. The limit seems to be the FSB of the processor.OK, we could have a timing problem of the software, but before build a
“test
system” in a customer application (VERY VERY VERY DIFFICULT!!!) we must
be
sure that the operating system has not responsability in the problem.In effect we note that, when the Operating system starts and we have a
monitor connected, the frequency of the celeron 566 (for example) is
reported correcly from the OS in the first screenshot (the monitor look
like 686/687 … 566 MHz). In the case of a Celeron 850 the frequency is
reported by the OS as 911 or 912 Mhz. The Pentium III (1 Ghz) reports 1092
Mhz.Has this frequency any influence in the PCI access?
PCI runs at 33Mzh and should stay that way whatever the processor speed,
however the number of access done over the PCI can change depending on CPU
speed (motherboard, FSB etc). Check in the BIOS if you have a configuration
for PCI latency, maybe playing with that could help, but I would try the
find the REAL source of the problem instead of just poking around in the
hope of finding a config that works.
I would not worry about the OS reporting wrong frequency.
What is your opinion?
My opinion is this won’t be easy to figure out 
I would try to create a
simple test case with the NI card installed in a PC/motherboard that gives
you full control over FSB/CPU speed (overclocker friendly motherboard) and
try to reproduce the problem.
You still haven’t reported what you meant by acquisition stops, how does it
stop, no interrupt, card dead, no more DMA?
Thank you
Davide Gardoni
Idra Casting Machines
d.gardoni@idracasting.com
“Mario Charest” postmaster@127.0.0.1 wrote in message
news:baou0d$k1b$> 1@inn.qnx.com> …“Davide” <> utee@idrapresse.it> > wrote in message
news:baluh3$9ut$> 1@inn.qnx.com> …
Hello.
We have a real-time application made with QNX4.25G and a National
Instruments acquisition board.
PCI bus is used only for accessing the NI card.
With Celeron slower than 733 (66 MHz FSB) system always works, with
Celeron
faster (100 MHz FSB) sometime the acquisition task stops.Could you be more specific as to what exactly stop, what happens?
Motherboard is from Mitac and it is not possible to slow down any
frequency.
Have you some suggestion for us?Maybe some register in the NI card need some time in between access,
although that would be very strange. Have you tried this with different
motherboard/CPU to rule out it’s not a problem with the hardware.
\