Tech Support banner

Status
Not open for further replies.
1 - 20 of 39 Posts

·
Registered
Joined
·
30 Posts
Discussion Starter #1
I need advice from someone familiar with internal technical details of the Pentium 4 CPU.

The P4 has a circuit which "shuts the CPU off" if a very high temperature is detected. It is automatic and does not involve any software. I suspect that the bus signal THERMTRIP# causes the CPU to drop into a SLEEP or DEEP SLEEP state. My question #1: If the CPU goes into a SLEEP/DEEP-SLEEP state in the midst of normal operation, would I see the following?

(1) Complete absence of any sign of activity other than continued spinning of the hard disks and fans;
(2) Complete nonresponsiveness to any user action (mouse, key presses, even pressing the power-off button uinless held for 5 seconds or longer);
(3) All visible indications (screen display, indicator LEDs) frozen in the exact state occupied at the instant that the SLEEP/DEEP-SLEEP state was entered - in other words, the screen continues to display desktop and application displays (frozen), not a blue, dark, or corrupted screen.

Finally, what circumstances other than an extreme over-temp condition can cause such a switch to SLEEP/DEEP-SLEEP?

Thanks.
 

·
Registered
Joined
·
906 Posts
The THERMTRIP# line going active should shut the motherboard down as this is the catastrophic over temp mode and power needs to be removed from the processor.

The PROCHOT# will throttle the processor by reducing the clock duty cycle. However according to Intel the CPU may hang if the throttle in Bios is set to 12.5% or 25%. This could cause the freezing in place of the system.
 

·
Registered
Joined
·
30 Posts
Discussion Starter #4
To Barry_R

Thanks again Barry. I feel that we may be closing in on the cause of my problem. If, because of a cluttered air-flow pathway and poorly applied thermal compound, my CPU is often operating close to the throttling point, then I would expect it to cross that threshhold occasionally. And, because the IBM tech replaced my 1.5 GHz P4 with a 1.7 GHz P4 (and, again, was careless in applying the goop), I would expect the freezes to occure a bit more frequently than before because of the greater heat dissipation of the 1.7 - and that is happening.

As for the BIOS, however, here's what I find in Intel's P4 data sheet concerning the Thermal Monitor feature:

"For automatic mode, the 50% duty cycle is factory configured and cannot be modified. Also, automatic mode does not require any additional hardware, software drivers or interrupt handling routines."

Other duty-cycle percentages can only be selected in what is called "On Demand" mode. They also say:

"Automatic mode is required for the processor to operate within specifications and must first be enabled via BIOS."

I have checked in my AMIBIOS setup screens and cannot find any reference to an option related to either mode. I am assuming - cautiously - that automatic mode is being enabled unconditionally by the BIOS, but I will research this assumption.

Any further comments you may have are most welcome.
 

·
Registered
Joined
·
25,116 Posts
Welcome to TSF:

Also be mindful a weak PSU will also cause processor over heating ???? what make and model PSU do you have ????? a 1.7 ghz CPU is not a big power hog / however manufacturerd PC's are known for not having any more capacity for upgrading than what was originally designed, with that in mind I would try running the PC with the side case off and with a house fan blowing on it / if that trick doesnt drop the temps by atleast 10 degrees then you have other issues / heat sink and fan maybe > psu ???? there are several reasons why a P4 can heat up / I would consider isolating your problem as a priority.


regards

joe
 

·
Registered
Joined
·
30 Posts
Discussion Starter #6
To Linderman

Thanks for the response. Pardon my ignorance, but what is a PSU? "Power Supply" something, I assume.

The problem with the house-fan approach is that the freezes, which are the problem manifestation, occur only every six to ten weeks or so. I wouldn't ordinarily be overly concerned with a problem of this kind and frequency, except that after three years of use with this problem, the problem abruptly became much worse (MUCH more frequent) and IBM eventually had to replace both the CPU and the motherboard (in February). This new mobo & CPU are now doing what the original did over the previous three years and I'd like to fix the problem before it again escalates beyond reason. But I'd also like to be reasonably sure that it is a thermal problem before treating it as such. After all, it doesn't seem exactly to fit the normal thermal scenario of clock throttling or catastrophic temperature rise. For the CPU that is. So, what's this PSU?

Thanks again.
 

·
TSF Team Emeritus
Joined
·
5,580 Posts
yes, PSU is your power supply, and like linderman is saying, a weakening power supply can cause overheating.

keep in mind, that an electronic device will draw a constant number of watts, regardless of actual voltage.

ohms law states that you can get any amount of wattage out of any voltage, given enough amprage.

you can also get any amount of wattage out of any amount of amprage, given high enough voltage.

so, if it's meant to be 12, and it starts running down into the low 11's, the components will draw a higher amprage, to compensate and keep the wattage constant.

higher amprage means more heat.


so yes, you might want to check your computers' voltages, your power supply might be dying.
 

·
Registered
Joined
·
18,118 Posts
I'm with Linderman on this one. I'm not an Intel fan in any way, but I do understand thermal design. Stock PSU's (Power Supply units) are often built just to accomodate what comes with the system. My old Gateway 366c came with a 95w PSU! A good one will include at least two fans, one of which would be mounted on the bottom of the unit to suck air up and out. You'll also want to remove the heat sink and redo the thermal grease. If it's not done right you can run into SERIOUS complications. After that's doe right you'll want to install a fan on the back of the case that's oriented to exhaust air outwards.
 

·
retired
Joined
·
1,966 Posts
The P4 has a circuit which "shuts the CPU off" if a very high temperature is detected. It is automatic and does not involve any software. I suspect that the bus signal THERMTRIP# causes the CPU to drop into a SLEEP or DEEP SLEEP state. My question #1: If the CPU goes into a SLEEP/DEEP-SLEEP state in the midst of normal operation, would I see the following?
Nope. When the core temperature reaches the High-High Limit (emergency shutdown) the CPU automatically shuts down. The THERMTRIP# signal is sent to the motherboard to shut down also. The CPU emergency shutdown is a separate action from the THERMTRIP# signal, so even if the THERMTRIP# signal were somehow disabled (or ignored by mobo) the CPU still shuts down and can only be restarted by SYS_RESET# (reset whole system).

The PROCHOT# is from a different internal thermal diode thats used as a High Limit (warning) signal to the CPU and also available externally so other components can also use the signal to the CPU as a high temperature warning.

It's up to the board manufacturer to implement the on board circuitry and use microcode (bios) and software utilities for temperature management for their board.

Sleep states are unrelated to thermal management.

When there is a high temperature situation, the TCC (thermal control circuit) and if implemented ACPI (automatic, on-demand modes) is engaged (with the automatic mode's duty cycle taking precedence).

So your system would slow down. If your system is locking up then it's due to some other cause.
 

·
Registered
Joined
·
30 Posts
Discussion Starter #10
To all:

Irony of ironies - my system froze in the middle of logging into this forum to post this reply!

Here's where I am:
It almost certainly is not a catastrophe thermal shutdown - I was very skeptical, anyway, that my P4 was reaching that extreme temp.

It's not a simple clock-throttling problem because the proper response to a hot proc is to reduce the duty cycle of the CPU, not to stop it.

My system (IBM Intellistation M Pro 6849 21U) has an exhaust fan in the PSU at the top back of the case. The case has an exhaust fan about center rear. Temperatures of the exhaust air run around 80 to 85 deg F.

I did just make some BIOS changes:
Under "Power Management", APM was set to "disabled"; I changed it to "enabled". (This, however, seems to relate to periods of inactivity, not to thermal management.)
ACPI "suspend state" was set to S3; I changed it to S1 (described as "safest").

My power supply is labeled "AcBel" (made in China), IBM PN 24P6800, IBM FRU 24P6801. Output ratings:

+5 V 32 A
+3.3 V 25 A
----------- 217 W max

+12.0 VP 11 A
+12.0 VD 5.0 A
-5.0 V 0.4 A
-12.0 V 0.7 A
+5.0 V (aux) 2.0 A
-----------------------
OVERALL TOTAL 340 w MAX

I do notice sometimes that when CPU utilization (as shown in Task Manager) shoots up to or near 100%, there is a very slight decrease in the frequency of the fan hum. I have three SCSI hard disks running, each about 18 GB in size.

Although I have installed hard disks, adapter boards, and memory modules, I must say I'm a wee bit intimidated by the prospect of replacing the power supply. Obviously it must fit the space occupied by the present PSU and provide precisely the same voltages and equal or better amperages for each. I also notice a device clamped around a bundle of the power leads, presumably a device to eliminate RF radiation or crosstalk. Does the PSU usually come with this device preinstalled or, if not, must I be careful with respect to the choice of leads which I enclose in it?

I really want to thank all of you for taking the time and being so helpful. And all further comments / suggestions are welcome.
 

·
Registered
Joined
·
18,118 Posts
That PSU is INSANELY underpowered. Get a new one and that ought to make things right. Look for one with at least 20a on the 12v line to ensure stability.
 

·
Registered
Joined
·
25,116 Posts
You need not worry about getting the wrong PSU / the are standardized (ATX spec version 2.0) you need only be concerned that you buy a quality PSU big enough to handle your configuration with EASE

All good quality PSU's have noise or RF reduction as well as multiple connect variations (it will fit any motherboard)

regards

joe
 

·
Registered
Joined
·
30 Posts
Discussion Starter #13
To Stu and Barry

Stu - I just reread your post and the reference to ACPI caught my attention. Do the S1 and S3 states refer to automatic and on-demand respectively? If so, that may be the problem. It has been set to S3 and, according to Barry, there is a document from Intel warning that the P4 may lock up if on-demand duty cycle is set to certain values.

Barry - Can you give me a reference to the Intel material which describes this problem?

Thanks.
 

·
Registered
Joined
·
25,116 Posts
The motherboard has more protection features than just throttle down !!

this feature is typical of today's board

The A7M266-D incorporates special features to protect the condition of your valuable CPU. A CPU throttling feature is enabled in the BIOS, which automatically lowers the CPU frequency to a safe speed if excessive temperatures are detected during operation. Additionally, ASUS COP (CPU Overheating Protection) is a hardware protection circuit that automatically shuts down the system power before temperatures go high enough to permanently damage your CPU.



cheers

joe
 

·
retired
Joined
·
1,966 Posts
ACPI S1~S5 are sleep states.

You can get the ACPI Specifications at http://www.acpi.info/, it's an action packed, keep you on the edge of your seat, 618 page pdf file.

Intel has pdf's for Thermal Design Guidelines and Platform Design Guide and...
You can read a skinny at...
http://www.overclockers.com/articles517/
(ignore the PROCHOT# "smart fan" comment, not practical due to thermal ramping)

Check your BIOS date you might be in for an update...
IBM IntelliStation M Pro
http://www-306.ibm.com/pc/support/site.wss/document.do?lndocid=MIGR-4RLUDT

Your PSU should definitely be replaced as already suggested.

---------------------------------------------------------------
Also, after an ACPI compliant OS is installed it's not a good idea to muck about with the BIOS ACPI and PnP settings. The OS incorporates those settings while being installed and altering them afterwards can cause all kinds of problems (including freeze-ups, bsod's, crashes, etc...).
 

·
Registered
Joined
·
906 Posts
slemaker said:
Stu - I just reread your post and the reference to ACPI caught my attention. Do the S1 and S3 states refer to automatic and on-demand respectively? If so, that may be the problem. It has been set to S3 and, according to Barry, there is a document from Intel warning that the P4 may lock up if on-demand duty cycle is set to certain values.

Barry - Can you give me a reference to the Intel material which describes this problem?

Thanks.
The spec sheet is here, see page 46.
 

·
Registered
Joined
·
30 Posts
Discussion Starter #17
Stu

OK. I did download that ACPI spec. Boy, you're right, it's a real page turner!

But I will restore the BIOS power management settings to their original states because they obviously have nothing to do with my problem.

Any recommendations as to a good quality brand of PSUs?

Best regards.

Carroll (mr., not ms.)
 

·
Registered
Joined
·
30 Posts
Discussion Starter #18
Barry

Thanks for the reference to the P4 update spec. They say that the consequence of this combination of events is that the CPU goes into a sleep state from which it fails to return. Whether or not this is the exact cause of my problem, it certainly seems to be what my system is doing.

Carroll
 

·
Registered
Joined
·
30 Posts
Discussion Starter #20
To Barry_R

Thanks for the recommendations - I'll check them out.

Incidentally, one of the specs (I don't recall if it's the P4 spec or the mobo spec) states DO NOT use a standard ATX power supply. It must be an ATX12V "12-volt compliant" PS. It has two additional power leads both of which MUST be properly connected to the mobo, otherwise damage to the CPU and to the mobo will occur.
 
1 - 20 of 39 Posts
Status
Not open for further replies.
Top