On Fri 23 July 2004 12:41, Rudolf Marek wrote: > > Hmm. I guess I could try emailing ASRock, and see if they have > > anything to say about it. After staring at the numbers for a > > while I've noticed that they do vary slightly, so it's likely > > that something is actually connected. The raw values > > (/proc/sys/dev/sensors/w83697hf-isa-0290/in5) I've seen for in5 > > so far are 3.01 and 2.99 (187/188 on the ADC), in6 does 3.02 > > through 3.06 (189/191). I've also looked at the datasheet for > > the w83697hf now, but that didn't yield anything beyond that > > something strange is going on. My R3/R4 ratio appears to be > > about 25.4, and my R5/R6 ratio is about 14.4, that is if my > > power supply is perfect. I guess I'll have to reboot to see > > what the BIOS says. > > In the bios you will see: > > 32 Bit Transfer Mode > Size > CPU Temperature > M/B Temperature > CPU Fan Speed > Chassis FAN Speed > Vcore > + 3.30V > + 5.00V > + 12.00V > American Megatrends Inc. > P2.40 I've still got version P2.20, but the ASRock changelog doesn't say anything about sensors. > > so I guess no negative volatages > > But in bios image there is something that is prepared for > negative volatges How did you get that BIOS image? Is there a way to extract it? I'd love to have a look. I tried downloading the images from the ASRock website, but they don't even have those strings in them. Or am I missing something? > eg000:206C cmp ah, 25h ; '%' > Ieg000:206F jb loc_0_209F > > continue here if IN number is bigger eq than 25 (so in5 in6) > > Ieg000:2071 cmp ah, 50h ; 'P' > Ieg000:2074 jnb loc_0_209F > > continue here if register number is smaller than 50 (I must ask > khali what is it) According to the data sheet, register mappings between 25h and 50h are: 20h-3Fh Monitor Value Registers 49h VID<4>/Device ID 4Ah Temperature 2, 3 Serial Bus Address 4Bh-4Dh Control Register 4Eh Select bank for 50h-5Fh Reg. 4Fh Winbond vendor ID 50h-5Fh are mapped to 5 different banks which are apparently selected through register 4Eh. I guess this 50h limit is pretty arbitrary... > Ieg000:2076 mov byte ptr es:[si], 2Dh ; '-' > > Here: write '-' before printing voltage !!!! > > > Ieg000:207A inc si > Ieg000:207B mov ah, 0A0h ; '?' > Ieg000:207D mul ah > Ieg000:207F mov cx, ax > Ieg000:2081 mov ax, 24h ; '$' > Ieg000:2084 mul bx Hmm, multiply by 160. The LSB of the ADC output is 16mV. So we're working in tenths of millivolts apparently. > The BX value is somewhere hidden or even not included so I cant > tell anything about the formula. Well, I worked it out from that disassembly, and we can say something about the formula. If I didn't make any mistakes, it seems to follow the formula in the datasheet, with the mysterious BX being the beta value. The 24h (36) in there is the 3.6V reference voltage. This means that my original estimate is probably somewhere near correct, but it leaves the problem of the resolution being way too small. Theory: the hardware guys at ASRock miswired in5 and in6. The BIOS software guys programmed the BIOS according to the datasheet. When testing, they noticed the problem, and disabled the negative voltage readout. It is a low-budget product after all, so fixing the hardware might simply have been too expensive. And negative voltages are hardly used in todays PCs anyway. I've emailed ASRock, whose first reaction was to tell me to try the program at http://mbm.livewiredev.com. I took a second try at explaining them that I'm running Linux, and that I would like to know the formulas. I haven't had any response so far, but it's only been half a day. We'll see what happens. > Please dont forget to CC to list. Of course. Cheers, Lourens -- GPG public key: http://home.student.utwente.nl/l.e.veen/lourens.key
