Is disabling SMIs on Pentium 4 safe?

[zigic]

Hello,

I am talking about Intel Pentium 4 (2.8GHz, 478-pin, 533MHz FSB, 512KB L2, 0.13um)and 82801DB ICH4 (this is part of 845GE chipset). We are using DOS to test real-time performance of Pentium 4 motherboard. Expected response time is few microseconds. System is in PIC mode (APIC disabled), all interrupts on master and slave PIC disabled, all interrupts on CPU disabled (asm("cli");), digital scope connected to D7 pin on the parallel port, we are using gcc compiler to do this:

...

while(1)

{

outportb(0x378, 0x80);

outportb(0x378, 0x00);

}

...

Digital scope is set to trigger when pulse width is >20us.

When we start program, after few minutes scope will trigger, showing about 250us stall on CPU. This is happening every few minutes, sometimes in less than one minute and it is not periodic.

We found the following:

1) On all motherboards using above mentioned Intel CPU and chipset, this problem is present.

2) On all motherboards with this problem, CPU pin PROCHOT# is asserted (logic '0'). The state of PROCHOT# is not changing, it always stays asserted.

This means that even the CPU was at room temperature for 24 hr, the moment you turn on power, PROCHOT# will become asserted and stay asserted until you turn off the power. The same CPU in different motherboard (with different chipset) never asserts PROCHOT#.

3) 250us CPU stall is caused by Intel 82801DB ICH4 chip. Once you disable GBL_SMI_EN bit in SMI Control and Enable Register on 82801DB ICH4, 250us gap will disappear.

4) Device which really asserted SMI# is TCO logic inside the 82801DB ICH4.

Once you disable bit TCO_EN, with GBL_SMI_EN enabled, problem will also disappear and we will have desired real-time performance.

Now questions:

1) Is disabling all SMIs in the system safe? In this case, after setting GBL_SMI_EN to 0, there will be no more activity on CPU SMI# pin. Am I in risk of burning something on the motherboard (including CPU)?

2) Is disabling TCO logic (by setting TCO_EN bit to 0) safe? Is it possible to specify what TCO logic exactly is doing and find actual event that caused SMI#?

I tried to find answers to these questions in datasheets and app. notes on Intel website, but without success.

Thank you,

Slavisa Zigic

[Intel_Software_Netw1]

Hello,

Thank you for posting your question on the Intel Software Network forum.

Your design inquiry will require a closer look by an Intel Product Representative.

If your company has its own Intel representative, you may wish to inquire whether they are able to assist with this inquiry. Your company's Purchasing Department will normally have your Intel representative's contact information. If you have no contact, please see http://www.intel.com/buy/networking/design.htm under "Design Components".

If your location is not listed, please see an Intel Authorized Distributor and ask for a Field Application Engineer (FAE). Our Intel Authorized Distributor list is also linked from the URL above.

Best regards,
Jim A
IntelSoftware NetworkSupport
http://www.intel.com/software
Contact us

[zigic]

Hello Jim,

Thank you for fast response. We are using AAEON motherboards models MB-845GE and EMB-845GE and also one more embedded motherboard from different manufacturer. The common thing is Intel Pentium 4, 845GE chipset (82801DB ICH4) and problem that I mentioned earlier.

Is it possible to get answers to posted questions without regard to the problem?

I will repeat questions:

1) Is disabling SMIs on Pentium 4 safe?

2) Is disabling TCO logic on 82801DB ICH4 safe?

Thank you,

Slavisa Zigic

[Intel_Software_Netw1]

We posed this question to one of our hardware contacts, who responded:

SMI must not be disabled. It is the mechanism used to signal over temperature conditions for Pentium 4 processors. There might also be some chipset fixes requiring SMI.

Even in PIC mode, the Pentium 4 processor-based platforms require the APIC be enabled via MSR 1Bh.

Hope this helps,

Lexi S.

IntelSoftware NetworkSupport

http://www.intel.com/software

Contact us