NF7-S temperature monitoring

Thought some of you might find this interesting. My new Abit NF7-S motherboard arrived about an hour ago. Having read that it does not read the CPU on-die therma, diode, I set about looking into whether that diode could be read using an external reader chip like the MAX6657 or AD1023 via the SMBus. The Winbond W83627HF chip on the board has three inputs that can be configured for either a 10k thermistor or thermal diode junction. The motherboard uses the first two along with the on-board thermistors to read the bottom of the CPU temperature and the system temperature. That left the third one unused. In previous AMD motherboards, all you needed to do was run a shielded two conductor cable from CPU socket pins S7 and U7 to a reader chip, assuming those two pins did not already connect to some other circuitry. Upon investigating those pins, my ohmeter said they went to some circuitry. Closer investigstion revealed that there is a small IC under the thermistor in the CPU socket well. It is an ATTANSIC ATTP1. That is a device with two inputs, one being from an on-die thermal diode and the other from the thermistor inside the CPU socket well. Its purpose is to monitor the temperature represented by either input and when a resistor programmed threshold is crossed, it shuts down the power supply. The response time is 1us. That's good news for owners worried about frying their CPU if the heatsink fails, but not such good news if you wanted to use the on-die thermal diode pins for driving an external reader.

If I get the urge, some day I will figure out how to isolate the conductor from the on-die Thermal diode going to the ATTP1 and route it out to either an external reader or the unused input on the Winbond chip. PDF files are available from the manufactures for both chips and in the case of the Winbond, a recommended circuit shows how to implement an on-die Thermal diode.

Hoot

hmm.....thats some good info there. What version are you using? Im on a 2.0 now and using a External Monitor and find the CPU temp the board reads VERY close(like 1C) to the DIE temp my probe is reading.

It's a 2.0 rev motherboard. Glad you are satisfied with your temperature reading. I know what my CPU thermal diode reads on the current board I'm using and once I migrate to the NF7, I will take some readings at the same speed and core voltage to see how the NF7 readings compare.

Hoot

I have to say, that this is BEYOND retarded.

They have dye temp monitoring on the AT7-MAX2, why the hell not on a NEWER board?

Jesus, they have some morons baking crap up out there.

A brief follow-up. I spent the morning with the board under the stereo microscope, tracing routes. It would appear that you have an either/or judgement call here. If you remove the on-die thermal diode from the ATTP1 shutdown circuit in order to reroute it to the unused 3rd input on the Winbond IC, you will lose the fast-responding hardware implemented overtemp shutdown protection.

Forget using an external reader like the MAX6657 or AD1023 due to how the accompanying circuitry is implemented with the ATTP1. No easy way to isolate both sides of the on-die thermal diode pins S7 and U7 on the socket.

Hoot

well at least i feel safe now about if something happens with my water cooling setup that she will shut down if temps rise. Thanks for the info.

Has anyone either intentionaly or accidently 'tested' the thermal diode based shutdown? The slowest AthlonXP with a thermal diode that I have is a palamino 1700+, but it may be worth heating it up to test before adding my watercooling setup.

they did:-

http://www.au-ja.org/eng/reviews-asusa7v333-cop-e.phtml

And I believe the abit system, since it is entirly indpendant, will even protect against the one that the asus failed on, but don't quote me on that if it doesn't.

Think THG need to do another test