NF7-S CPU Thermal Diode Mod, Version 2.0

[jkesa]

Yeah it would be interesting to get to the bottom of whats going on. I havent played around with the program for about 5 months, but I recall getting different offsets on different reboots, especially on the system temp offset. What have you found on your system ? Ill run some more tests on it tommorrow again and post back. Dont worry about the disclaimer, its just to scare off noobs. Unfortunately, Im using a modded bios, so that may be yet another variable.

And right now my NF-7 is on the fritz

Hope you get it up and running again soon so you can see what were talking about. That little program not only displays the w83627hf offsets, but also allows you to change them. So what Im thinking is encoded in the bios is some kind of an algorithm.

[IronStone]

Hi,

someone's out there who can confirm this mod working on a nf7-s v2 ? Ive own the nf7-s v2 and build an SMBus extension with a max6657 and tried the socket pins THDC and THDA and get as results 29 or 30°C. Then ive found this mod from gizmo and tried the pin4 of the attp, but get same results on 2 different max6657. Hardware bugs are not there with other sensors the 6657 works correct. Before im going the way to lift the winbond pins and get same results (stable between 29and 30°C) i love to get a confimation for the nf7-s v2 revision. If done the mod successful, please told with which bios.

Regards,

IronStone

[Gizmo]

Hey IronStone! Glad to see you could drop in!

I developed this mod on the NF7-S v2 board, so I know it works.

What exactly are you doing? Did you lift pin 4 of the ATTP-1?

[IronStone]

@gizmo

no, in your mod V2 stands nothing about lifting pin4. As far as i understood is pin 4 the pin where the ATTP-1 gehts the anode-signal of the thermal diode ? When i lift this pin the ATTP gets no more signal from the diode or misunderstood i something ?

Ive build a 0815' SMBus extension with the MAX1668MEE first, tried to connect THDA and THDC on a sensor input of the 1668 (used cd-Audio-cable with huge groundwires connected to ground on the Vcc Socket and on my circuit). Got temps at 29 or 30°C, no fast reactions like all others told like the behavior of the int-diode should be. So i implemented a MAX6657 in the same circuit only for the cpu-diode, coz it seems that the whole world uses 6657 for the int-diode ;-).
Same problems as before, 29-30°C and no reactions under prime95 or idle.
In the meantime got some problems with the SMBus, posting at the MBM board, got some response from someone who done this mod while disconnecting the THDA and THDC pins from the Mobo for measuring, but i thinked that the thermal guard function of the abit makes this solution not so easy for me so he found your post here and told me about it.
As far as i know, (no learned electrician) is there not big difference between the winbond and the max6657 ic in the handling of the 'hardware-input' (both specified too for cpu-diode measuring) leaving offsets out.
So after my mail to you i connected the cd-rom audio cable (only one wire)
to the pin4 and to the DXP on the 6657. Same results, 29-30°C.
Checked my circuit again, tried different sensors , all worked perfect on the 6657.
changed the 6657 to another one .. same problems...
Cutted the cable (interferences?) as short as possible (shorter then the lenght you need to reach the winbind) same problem...29-30°C....
Today i got my new cpu, an 2600mobile IQYHA, had a rest of hope that maybe it was my old Thunderbird who made the problems, rechecked the soldering of circuit and of pin4 ... all ok, but ...
yes .. same problems .. 29-30°C .... no fast reactions .. nothing like it should behave ...
so im out of ideas what to check or try with the 6657 solution...

So a new roman, but thats my whole story about my journey exploring the SMbus and the nf7-s v2 ....;-)

Regards,

IronStone

[Gizmo]

Originally Posted by IronStone

@gizmo

no, in your mod V2 stands nothing about lifting pin4. As far as i understood is pin 4 the pin where the ATTP-1 gehts the anode-signal of the thermal diode ? When i lift this pin the ATTP gets no more signal from the diode or misunderstood i something ?

Just checking.

I'm just looking over the datasheet for the MAX6657 now. In the block diagram, it indicates that DXP is fed from an internal constant current source. This is your problem, as the diode already has a bias current applied from the 3.3v reference voltage (VREF) developed by the ATTP-1. In order to make your circuit work you will need to disconnect this bias. Probably the easiest thing to do would be to lift pin 7 (the VREF source) of the Attansic chip. You will need to power your MAX6657 from the +5VSB if you do this, though, otherwise the ATTP-1 won't let you boot the computer. This should allow you to still have the ATTP-1 protecting your CPU while you read the temp from the MAX6657.

[IronStone]

ok, try this when ive got enough time to disassemble my pc. Thats the disadvantage of water cooling..;-( spend much time on disassemble and assemble to do some small mod....
Thanks for your hints, if ive test it i reply if thats the solution ...

Regards,

IronStone

[Gizmo]

Originally Posted by IronStone

ok, try this when ive got enough time to disassemble my pc. Thats the disadvantage of water cooling..;-( spend much time on disassemble and assemble to do some small mod....
Thanks for your hints, if ive test it i reply if thats the solution ...

Regards,

IronStone

Heh. You think you've got it bad, take a look at the link in my siggy. I've got to DRAIN my system every time I want to do something to it.

Please do let me know how it works out for you. I'm always happy to help a fellow hardware hacker!

[Hood]

Originally Posted by gizmo

Bridging some of the legs with solder could cause all kinds of weird things. Whether or not it would cause what you are seeing, I can't say.

If, for example, you got pins 102 and 103 shorted together on the winbond chip, that would certainly cause BOTH the thermistor and the diode input to read high.

Sorry for being late in replying but fried my mobile and not sure if the mobos good or not. Will test it out again when I have time and let you know. BTW looked at the soldering through a magnifying glass and all looks OK, so erronous reading may have been the mobile on its way out.
Hood

[jkesa]

jkesa - have you noticed whether the values used are the same each reboot?

Sorry for the late reply, my GWIOPM.SYS file was corrupted. Using D24 modbios3. First value is ts1 system offset, second value is ts2 cpu offset, and then ts3. On a cold boot the value I got was -10,-4,0. Reboot -10,-5,0. Second reboot -10,-5,0. Approx ninth reboot after a period of time -10,-2,0. What do you get on which bios ? The programs author, Sweetlow, from overclockers.com.ru also believes there is a recalibration going on.

[IronStone]

@gizmo

now im world champion in disassebling and assembling my mobo ;-( ive tried the hint you gave me and powered the MAX6657 with the 5VSBfrom the attp1 got GND from the attp1 also and lifted pin7. SMBCLK and SMBDATA to the SMBus, but on power on i got the siren. so take GND from the SMBus (clearing all possibilities) same result, checked the MAX6657 circuit after this with another sensor, works fine, so theres no problem with the circuit itself. After 4 or 5 times dis/reassembling i decided to do your mod, coz i didnt have anymore fun to dis/reassemble. So i done your mod. Pressing power on ... nothing happens ... urgh ... no siren nothing no fan is running only the red led is burning. removed the mod checked again, mobo runs fine. Today ive bought another nf7-s (new) and tried your mod there. Same procedure... on Power on nothing happens. No siren, nothing but the red led. Ive checked the soldering 10 times, measured up between pins 102,101; 97,98,99; and same with 61,60,62 .. checked with a magnifying glas, no shorts , nothing ... dont know whats going on here ....
staying shorly in front of a mental hangover ... think ive must drink a lot of beer if i dont find out why this dont work at my 2 nf7-s ....

If u have some suggestions , pls told me ...

Regards,

IronStone,

[Gizmo]

Originally Posted by IronStone

ive tried the hint you gave me and powered the MAX6657 with the 5VSBfrom the attp1 got GND from the attp1 also and lifted pin7. SMBCLK and SMBDATA to the SMBus, but on power on i got the siren.

Do you have #STBY pulled up to 5VSB also? #STBY low disables the internal bias generator. This might be the cause of your problem. If this is not the case, then I am as mystified as you.

Originally Posted by IronStone

So i done your mod. Pressing power on ... nothing happens ... urgh ... no siren nothing no fan is running only the red led is burning. removed the mod checked again, mobo runs fine. Today ive bought another nf7-s (new) and tried your mod there. Same procedure... on Power on nothing happens. No siren, nothing but the red led.

This also is exceedingly curious. About all I can offer here is some basic trouble-shooting:

Pin 68 of the Winbond chip is PSIN. It is internally pulled low. You should see it go active high when you press the power switch.

Pin 69 is PSOUT#, and is the inverted conditioned output of PSIN. It is normally high, and pulses active low when PSIN goes active.

Pin 3 of the ATTP-1 chip is PSON_IN#. When the power switch is activated, you should see this pin go low and stay low.

Pin 6 of the ATTP-1 chip is PSON_OUT#. When pin 3 goes low, pin 6 should go low also, unless the ATTP-1 thinks there is an over temperature condition, in which case pin 6 will remain high and pin 1 (SIREN#) will go low.

I'm pretty sure there is some logic between pin 69 of the Winbond chip and pin 3 of the ATTP-1, but I don't know where or what.

Hope this helps you.

[Bung]

jkesa, using D18 cold boot always -10,-10,0. Reboot typically -10,-1,0. Immediate reboot on cold morning -10,0,0. After 1hr Prime -10,-5,0. After last reboot my idle temps are 42 CPU 25 MB or what was showing while Prime was running before. So if you are doing any comparitive tests you have to reboot before starting to record temps.

[Gizmo]

Hmmm...........I haven't checked this myself yet. Is there any evidence of correlation between ambient and the offset that you see?

[Bung]

The only offset that gets changed is the CPU temp. On the cold morning reboot that changed the CPU offset from -10 initial to 0 the board didn't get a chance to heat up so I think it's only the CPU reading that is checked and adjusted.

Just had a play. The offsets are added directly to the displayed temps. Perhaps an assumption is made of what the cpu temp should be shortly after bootup before the heatsink is a factor.

[Gizmo]

Then they must be basing the offset off the initial reading.

I've never worked with precision thermistors. As near as I understand it, the manufacturer of the thermistor provides a formula that can be used to calculate the proper temp when the resistance is known, as the relationship between temperature and resistance is pretty well understood in these devices. So the BIOS comes up, takes about 4 samples at known intervals, computes the curve, and that gives the temp. However, because the thermistor is insulated from the heat source (the CPU) by the casing of the device AND the case of the CPU, an offset has to be applied. As well, local heating effects have to be taken into account. My guess is that is what is happening here, although without actually disassembling the code I have no way of knowing for sure.

[jkesa]

Yes, it seems ambient temps play a big role. Bung has a larger swing than I do. Im getting -10 offset on the system too right now, but do seem to have other numbers previously on a different bios version (which I didnt write down) Yes, by changing the offsets yourself, the result is displayed by MBM or other monitoring programs. If you dont like the standard displayed temps, you can make a batch file to start the w83627hf program with the new offsets. Ill have to find my old modded digital indoor/outdoor thermometer to check my system temps by RT2. I suspect it to be a little high, and dont really like the Abit placement of RT2.