Standard Thermal Temps?

B

BingeBoy

Weaksauce
Joined
Jan 19, 2008
Messages
109
I'm new to creating gaming machines and just completed my first rig.
After installing PC Probe II in winxp I had a few alarms go off. The MD temp (38c) DRAM(2.4) were in the red. After some searching I noticed that other users where having the same issues and the PC Probe II wasn't very accurate. Where can I find heating threshold info for my components?

I also noticed that my Nbridge temp went into the red after playing a game for about 1 hour. Nbrige temp reached 45c which seemed a litle hot. Nothing on my system is over clocked yet except for the ram which I believe is oc by default auto value in bios. My CPU is running at 27c and I'm more concerned with SB and NB standard temps and limits.

Here are the components of my system:
Mobo:Asus Maximus Formula
Intel Q6600
ZALMAN CNPS 9700 NT
2x1024 Corsair Dominator 1066
Visiontek Radeon HD 3870
segate barracuda 250gig
Ultra X-Finity EE 600w PSU w/APFC
 
The data sheets on the X38 chipset are available at Intel.com products chipsets X38


What you will find out is that the MCH (northbridge) and ICH (southbridge) monitor there internal temps and have the ability to shutdown if they get too hot. For the MCH the relevant section is 10.4 of the datasheet.

10.4 Thermal Sensor
There are several registers that need to be configured to support the MCH thermal
sensor functionality and SMI# generation. Customers must enable the Catastrophic
Trip Point as protection for the MCH. If the Catastrophic Trip Point is crossed, then the
MCH will instantly turn off all clocks inside the device. Customers may optionally enable
the Hot Trip Point to generate SMI #. Customers will be required to then write their own
SMI# handler in BIOS that will speed up the MCH (or system) fan to cool the part.

And based on the the descriptions of the registers available to monitor the Thermal Sensor all you can tell is that if it tripped. It does not report its internal temp.


In short, there is no way to determine the internal temps of these components and you have to rely on sensors placed on the board near these devices which are notoriously inaccurate.

You problem stems from the HS/fan you are using which has eliminated the downward "blow by" of the stock HS fan and reduced airflow over the radiator sections of the heatpipe and the memory. This can easily be tested by opening the side panel and with a spare case fan, direct airflow over different parts of the board while watching your thermal sensors. Investigation and a little thought will tell you where you install additional fans to provide cooling for the board components. Intel MCH's run hot, it is almost standard procedure to put a fan on the MCH section of the heatpipe/HS on boards that are highly OCed. The ICH is not as critical. Fans exist that use the memory "ears" to hold fans above the memory sticks, OCZ and Corsair make them. It is likely the placement of a 120mm low speed fan in just the right place to wash air over the entire top 1/2 of the board would solve your issue.

Until you find exactly where those on board sensors are and exaclty what they are susposed to report, numbers are meaningless. The cpu, MCH and ICH will all shutdown automaticlly if too hot. However cooling properly for long life and reliability is another issue entirely and basically the cooler the better and what the temp actually is has almost no importance. The cpu voltage regulation components around the cpu have no ability to "save themselves", ditto the memory, Your finger is an excellent thermal probe. If you cannot hold your finger on a component for a count of 5 it is way too hot.
 
Thanks for the feedback. I have a extra 120mm fan that I'm going to test with now.
I have attached a image of my machine.

http://www.flickr.com/photos/bingeboy/2205693387/

Currently I have 120mm fan in the front and rear. However I'm thinking of mounting one between the hd slots and the gpu to get some airflow in there.

I'll post my results after messing with the 3rd fan.

Thanks
 
I have added a 3rd fan right in front my my memory.

here is a pic:
http://www.flickr.com/photos/bingeboy/2205993029/

System is running much cooler now.
MB 33c down from 38c
NB 37 down from 45c
SB 38c down from 44c

I'm a little less concerned about my sys heat now. Before it seemed like the board was getting stressed and all i was doing was playing wow maxed out at 1920x1250.

Anyone have any more cool advice?
Also do you think I can begin the oc process?
 
After running Stalker max out my mobo went up to 43c. within in minutes. Is there something I can try in the bios to get it running cooler?
 
Since my last post I have added 2 80mm fans.
One on the case door to pull more air to the gpu area and a second pulling air from the front of the case to the SB. Played games on the system all last night and everything was staying cool.

Next step will be a basic stress test on all default settings in bios before trying to OC.
 
Good show, nice to see someone sweat the details and proceeding with a plan. That extra air as you saw will reduce running temps but the real benifit will kick in when you start to OC, temps will rise, but nothing like they would have without the extra airflow. If it is getting loud you could probally 7V the fans to run them slower (red fan wire to yellow wire on 4 pin molex, black fan wire to red wire on 4 pin molex, this connects the fan to the +12 yellow and +5red and the difference is 7V and thats what the fan sees and runs slower, or a cheap fan mate 2 or other fan controller).

Use coretemp or speedfan to monitor your cpu core temps. Do not put a lot of faith in most of the readings but the cpu core temps are accurate in both programs. Speedfan will give a lot of temps, the cpu core temps are correct and most of the others but they may be labeled incorrectly by the program as it basically has to guess at what a temp is monitoring. You can edit information in speedfan to give the temps correct labels once you determine what the temp actually is monitoring. The best source for that informatin is someone with your exact board or the manuf forums if they exist or a forum specific to your brand of board. You can also ask tech support "where is the sensor located on the board that provides the temp reading labeled xxxxxx in the bios". Sometimes they even can find the answer :D

Orthos for basic stress testing while running coretemp to monitor temps

TAT for obscene thermal testing.

Several others I cannot think of right away. Oh never hurts to loop 3Dmark as well to give the video a workout.

Cpu core temps should stay under 80C at full load (100Cish is where the cpu will send out an "I am too hot signal" and about 10-15C above that, 120Cish, it will thermal shutdown. ) Absolute max Vcore from Intel datasheet is 1.55V I do not like going over 1.45V and keep my personal machine for 24x7 use under 1.4. People have been overvolting these cpus since thay came out and the higher voltages do not seem so far to affect life or reliability if you also control the temps.

You probally already found/seen it but some good info in here:
http://www.anandtech.com/mb/showdoc.aspx?i=3149&p=10
 
You must log in or register to reply here.
Back
Top