Th7-II Mainboard temp sensor?

Hi All,
Anyone know where the mainboard sensor is located on this board? Although I'm crash-free, I notice a weird thing with the temps.
I'm using the retail HS/Fan with AS2. In a 20C room, my mainboard temp is 30C, CPU idle at 39C. When the machine is under load, the CPU temp will rise 10C, but the mainboard temp does too!
With my old P3 1G/1.75 vcore, thes CPU temp would rise under load, but the mainboard temp would stay the same. My old setup 20C room, mainboard 25C, CPU idle 28C, full load 35C.
Obviously the mainboard temp sensor is in a diff location, as I am using the same InWin full tower, with a 120mm fan in the lower front, 80mm fan upper rear and the Enermax PS with the 92mm fan above the CPU sucking out the heat.
I'm not too worried as it's crash free, but wonder if I should add fans or change the retail hs?
Maybe the sensor is near the Rdram and thats heating things up?
Anyone think an 80mm fan blowing on the rdram will matter?
Any ideas welcome, TIA,
Mike G
P4 1.6A /2.1g 1.5-1.52 vcore
2-256 rdram, no volt mods, 2.48 vrimm

Mine is a 22C room temp, 28C mobo temp idle, CPU 36C idle vcore a true 1.6volts. I put a 60mm fan on the chipset heatsink and the mobo temp was unaffected, but the chipset HS surface temp went down to 31.5C idle from 35C, so I don't think it's near the northbridge.

I have the same problem, but with an ST6. I hear that the system temp sensor is not located anywhere on the board, but inside the "winbond" sensor, which is the cause of the higher system temps. It only goes down when my computer is next to a window with a breeze.

Here's the link from where I got this possible explaination:
http://groups.google.com/groups?hl=...43762%24cc9e4d1f%40news.dial.pipex.com&rnum=8

Thanks for the replys, that Google reference seems to make sense. If my mainbord is reading 7-9C higher that puts me in the ball park with my old setup.
Also, I think you're right about the sensor being "built-in" to the Winbond chip, I'm going to set up a Card Cooler blowing on that area and see if it drops.
I wonder why the Winbond chip get hotter under load, is it as simple as the whole mainboard working harder? Aslo, I notice that the vcore, +3.3, +5, and +12 voltages go up when the load goes up. I have an Enermax PS, and wonder if all PS output more on the lines as demand goes up?
Interesting, got to study this more.

Edit: oops, misinterpreted MBM voltages, they actually dropped under load, which makes more sense. So higher voltages aren't causing the heat increase on the Winbond chip, I'll keep searching for the answer.

Well, with proper cooling, system temps should go up only slightly. Mine ranges ~5c easily. Big problem. If it is true that the system temp sensor is inside the winbond chip, it makes perfect sense that temps would read higher since it is not exposed, or so it seems. You can try putting a fan on the northbridge and see if that relates to dropping the system temps. I'll be PMing tainice tomorrow (homework ) since he has the same board as mine and he has normal looking system temps. See how he's done it. Be back with replies

I was checking out the Winbond 83627HF spec sheet here
http://www.winbond-usa.com/products/winbond_products/pdfs/PCIC/627hf.pdf
It says that there are "...three thermal inputs from optionally remote thermisters..", so wouldn't that mean that the mainboard temp on this boards is being read externally to the sensor chip?
If so, one is the CPU Diode, another the external thermister you can plug in to the board (JP1) and the third the mainboard temp. So the question remains, where is this mainboard temp being read from?
I'll try and email Abit tech support again for an answer.

i think that the board just runs hot. i put about 50 bucks worth of rounded cables and a bunch of fans and my temps dropped dramatically.

I disagree. These temps are BS! I calibrate for a living and using a device thats accurate to 1% of full scale I get roughly 8C difference depending of course on the placement of the J type thermocouple. I posted a question asking if anyone knows where to locate and set the offset. My machine has 8 fans and 2 blowers. top it off with the fact it's water cooled. The 2 blowers pull 40 CFM each. Also I have set the machine in different enviroments for hours only to find a minimal change in temp when clearly theres an easy 10C diff. I honestly cannot get my CPU temp to move anywhere from 39.5C to 40.5C even after looping 3DMark 2001 for 12 hours. I hope there is an alternative to this. I use to run T-birds OC'd and even with my Alpha Pal I was able to get down to 38C idle. I was running Abit KG7 RAID which uses VIA monitor. I guess my thoughts are theres no way an AMD T-bird OC'd 200MHZ air cooled runs cooler than a P4 OC'd 1000MHZ H20 cooled. Even through all my voltage adjustments and overclocking a misleading 1-1.5C difference. If we figure out a way to offset I'll be happy to take machine into a lab enviroment and we will allow to achieve lab temp (68+/-.25F) and I will give everyone the exact offset variable.

ElroyCarbon, do you think the method on this page has any validity?

http://www.arcticsilver.com/diode_calibration.htm

LMK your thoughts, I'm thinking of offsetting MBM by 5C, although in my case I'm perfectly stable so I guess the reading doesn't matter whatever the number.

That link is dead please send it again. I'm in the process of setting up a test in the lab. I'll document my findings. Also I will take photos of equipment being used as well as the lab in case anyone is interested. I will check a number of locations inside the
enclosure to maybe locate hot & cold spots. I will also note the location of my RAM as well, again I will have photos so we will know how my results may differ from anyone elses. I could use a hand if anyone has any info on the location of these chips that measure the temp. My first test was quick this one will be much more accurate. Hopefully this will help everyone get the real temps we deserve.

ElroyCarbon said:

That link is dead please send it again.

Click to expand...

Clicking on the link doesn't work. Do some copy and paste to the address bar to get there.

Duh! Thanks! I also talked to a few Cal techs at work and "it's on" tomorrow we will do the temp comparison so I should have useful results in a day or so. Thanks for the link I will read it later hopefully it will help us in our quest to calibrate the temp sensor. That is if it needs it . I'm still thinking it's out though. I may be going overboard, but it doesn't cost me any time or money and it's accurate!

good stuff Elroy, everyone will appreciate to know how off these temperature readings are

Thank you for your efforts. I have been waiting for ever for someone to do some tests to verify if the temps are accurate or not. If they are not, I can't wait to squeeze a few hundred more mghz out of this baby.

IMHO it doesnt really matter if the temps are accurate or not (although it would be interesting to know, it is not important) as the P4 throttleing feature is going to kick in before you get high enough temps to do damage anyway.

My P4 1.6a is at 2.352Ghz 24/7 and idle temps are 37-50, av 44C.
under load it runs at 44-59 av 54C.

I have had it as high as 2.5Ghz but it took a lot of voltage to do it and the temp was averaging in the 60's with a brief peak of 72C. At this I was being throttled, I had the throttleing reduced to a minimum, but then I decided to try and find out exactly where the throttleing kicks in. To accomplish this I set throttleing to max and began heating up the processor by working it hard with bench marks and slowly upping the voltage using the bench marks to indicate where throttleing was occuring. I cannot put a hard number on it (as you know the temps fluctuate quicly within a range) but it appears that throttleing, at least with my TH7II and P4 occurs at an indicated cpu temp of approx 63C+/- 2C.
The brief transient 72C temp indicated above was only reached with extreme voltage and throttling reduced to a minimum. I would not try this again and do not reccommend it to anybody else.
However I would recommend setting throttleing to the maximum to see if it is affecting you at all, and to determine the threshold of throttling in your particular processor so you can be aware when you are getting close, and that pushing the voltage, clock or work load further will result in decreased and not increased performance.

I hear ya! I realize it doesn't matter, but a guy like myself has put forth what I would say alot of effort to control the temps only to be informed by this piece of sh... temp sensor that my efforts were worthless. It's not going to save the day for anyone just let us know perhaps what we are truly working with. I'm considering another project dealing with LN2. I suppose then according to my machine I may if I'm lucky get down to 38C. I'm doing the test as we speak. I will have class tonight, but I'll post a thread tomorrow (Thursday) with the data and I'll even include a couple of screen shots. For those who have a little extra time maybe they caould help me figure out how to tweak the sensor. Maybe this Mr. Natural fella could help us with the offset since if I'm not mistaken the readings are taken directly from the bios. I really don't know, but I'm thinking it reads the bios and just reports then with the Windbond software. If anyone has any info please post it.