Problem with realtime on PC

Hello,

I am using AAEON motherboards, models MB-845GE and EMB-845GE. The system is hard real time where response is expected in few microseconds. I tried running different software, but no matter what I am running, there is "dead-time" of approximately 250 us. This means that every few minutes, CPU will stop executing program for 250 us and then continue. I tried two different CPUs, P4 at 2.8 and 1.7GHz and both are showing the same problems. I also tried the same CPUs in two motherboards from different manufacturers and there is no "dead-time" on CPU. The only common thing on two different motherboards is the same Intel 845GE chipset. I also found that PROCHOT# pin on CPU is always asserted (logic '0'). I am running all tests on DOS platform (no multitasking). It doesn't matter if you enable/disable interrupts or enable disable APIC. Any idea what is causing this?

Thank you,

Slavisa Zigic

Reply to
Slavisa Zigic
Loading thread data ...

Probably an SMI : System Management Interrupt => non maskable. We had to tackle the same problem for our RT application.

Check in setup fro a parameter like "thermal delay " or dont remember the name and affect it the smallest possible value like 4 minute => after that delay, the SMI does not occur anymore.

I think it is possible to really handle the problem with the Pentium MSR (model specific registers) .

Reply to
bm

Most likely, SMI's. It's just been discussed:

formatting link

-- Cyril

Reply to
Cyril Novikov

Cyril Novikov wrote: : On 05.06.2006 14:39, Slavisa Zigic wrote: :> Hello, :> :> I am using AAEON motherboards, models MB-845GE and EMB-845GE. The system is :> hard real time where response is expected in few microseconds. I tried :> running different software, but no matter what I am running, there is :> "dead-time" of approximately 250 us. This means that every few minutes, CPU :> will stop executing program for 250 us and then continue. I tried two :> different CPUs, P4 at 2.8 and 1.7GHz and both are showing the same problems. :> I also tried the same CPUs in two motherboards from different manufacturers :> and there is no "dead-time" on CPU. The only common thing on two different :> motherboards is the same Intel 845GE chipset. I also found that PROCHOT# pin :> on CPU is always asserted (logic '0'). I am running all tests on DOS :> platform (no multitasking). :> It doesn't matter if you enable/disable interrupts or enable disable APIC. :> Any idea what is causing this?

: Most likely, SMI's. It's just been discussed:

:

formatting link

: -- : Cyril

Cyril,

I downloaded 845GE chipset datasheet and wrote a small program to read control/status registers. All registers in South Bridge have default settings, which means, SMI disabled.

Now, about MSR on Pentium... Do I have to write routine in assembly (inline, if I use gcc) or there is C library available for MSR access?

Thanks,

Slavisa Zigic

Reply to
Slavisa Zigic

I disabled Thermal Management (TM1 & TM2) by writing 0x81 to IA32_MISC_ENABLE. Bit 3 is Thermal Monitor 1 (TM1) enable, bit 13 is Thermal Monitor 2 (TM2) enable.

Nothing changed regarding 250us CPU stall.

Slavisa Zigic

Slavisa Zigic wrote: : Cyril Novikov wrote: : : On 05.06.2006 14:39, Slavisa Zigic wrote: : :> Hello, : :> : :> I am using AAEON motherboards, models MB-845GE and EMB-845GE. The system is : :> hard real time where response is expected in few microseconds. I tried : :> running different software, but no matter what I am running, there is : :> "dead-time" of approximately 250 us. This means that every few minutes, CPU : :> will stop executing program for 250 us and then continue. I tried two : :> different CPUs, P4 at 2.8 and 1.7GHz and both are showing the same problems. : :> I also tried the same CPUs in two motherboards from different manufacturers : :> and there is no "dead-time" on CPU. The only common thing on two different : :> motherboards is the same Intel 845GE chipset. I also found that PROCHOT# pin : :> on CPU is always asserted (logic '0'). I am running all tests on DOS : :> platform (no multitasking). : :> It doesn't matter if you enable/disable interrupts or enable disable APIC. : :> Any idea what is causing this?

: : Most likely, SMI's. It's just been discussed:

: :

formatting link

: : -- : : Cyril

: Cyril,

: I downloaded 845GE chipset datasheet and wrote a small program to read : control/status registers. All registers in South Bridge have default : settings, which means, SMI disabled.

: Now, about MSR on Pentium... : Do I have to write routine in assembly (inline, if I use gcc) or there : is C library available for MSR access?

: Thanks,

: Slavisa Zigic

Reply to
Slavisa Zigic

After disabling TCO_EN SMI#, problem disappeared.

The only question now is: Are there any side effects?

When you disable all devices than can assert SMI# to CPU, is there any danger of overheating CPU or any other part on the motherboard?

Thanks,

Slavisa Zigic

Slavisa Zigic wrote: : I disabled Thermal Management (TM1 & TM2) by writing 0x81 to : IA32_MISC_ENABLE. Bit 3 is Thermal Monitor 1 (TM1) enable, bit 13 is Thermal : Monitor 2 (TM2) enable.

: Nothing changed regarding 250us CPU stall.

: Slavisa Zigic

: Slavisa Zigic wrote: : : Cyril Novikov wrote: : : : On 05.06.2006 14:39, Slavisa Zigic wrote: : : :> Hello, : : :> : : :> I am using AAEON motherboards, models MB-845GE and EMB-845GE. The system is : : :> hard real time where response is expected in few microseconds. I tried : : :> running different software, but no matter what I am running, there is : : :> "dead-time" of approximately 250 us. This means that every few minutes, CPU : : :> will stop executing program for 250 us and then continue. I tried two : : :> different CPUs, P4 at 2.8 and 1.7GHz and both are showing the same problems. : : :> I also tried the same CPUs in two motherboards from different manufacturers : : :> and there is no "dead-time" on CPU. The only common thing on two different : : :> motherboards is the same Intel 845GE chipset. I also found that PROCHOT# pin : : :> on CPU is always asserted (logic '0'). I am running all tests on DOS : : :> platform (no multitasking). : : :> It doesn't matter if you enable/disable interrupts or enable disable APIC. : : :> Any idea what is causing this?

: : : Most likely, SMI's. It's just been discussed:

: : :

formatting link

: : : -- : : : Cyril

: : Cyril,

: : I downloaded 845GE chipset datasheet and wrote a small program to read : : control/status registers. All registers in South Bridge have default : : settings, which means, SMI disabled.

: : Now, about MSR on Pentium... : : Do I have to write routine in assembly (inline, if I use gcc) or there : : is C library available for MSR access?

: : Thanks,

: : Slavisa Zigic

Reply to
Slavisa Zigic

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.