Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 1 | uGuru datasheet |
| 2 | =============== |
| 3 | |
| 4 | First of all, what I know about uGuru is no fact based on any help, hints or |
| 5 | datasheet from Abit. The data I have got on uGuru have I assembled through |
| 6 | my weak knowledge in "backwards engineering". |
| 7 | And just for the record, you may have noticed uGuru isn't a chip developed by |
| 8 | Abit, as they claim it to be. It's realy just an microprocessor (uC) created by |
| 9 | Winbond (W83L950D). And no, reading the manual for this specific uC or |
| 10 | mailing Windbond for help won't give any usefull data about uGuru, as it is |
| 11 | the program inside the uC that is responding to calls. |
| 12 | |
| 13 | Olle Sandberg <ollebull@gmail.com>, 2005-05-25 |
| 14 | |
| 15 | |
| 16 | Original version by Olle Sandberg who did the heavy lifting of the initial |
| 17 | reverse engineering. This version has been almost fully rewritten for clarity |
| 18 | and extended with write support and info on more databanks, the write support |
| 19 | is once again reverse engineered by Olle the additional databanks have been |
| 20 | reverse engineered by me. I would like to express my thanks to Olle, this |
| 21 | document and the Linux driver could not have been written without his efforts. |
| 22 | |
| 23 | Note: because of the lack of specs only the sensors part of the uGuru is |
| 24 | described here and not the CPU / RAM / etc voltage & frequency control. |
| 25 | |
| 26 | Hans de Goede <j.w.r.degoede@hhs.nl>, 28-01-2006 |
| 27 | |
| 28 | |
| 29 | Detection |
| 30 | ========= |
| 31 | |
| 32 | As far as known the uGuru is always placed at and using the (ISA) I/O-ports |
| 33 | 0xE0 and 0xE4, so we don't have to scan any port-range, just check what the two |
| 34 | ports are holding for detection. We will refer to 0xE0 as CMD (command-port) |
| 35 | and 0xE4 as DATA because Abit refers to them with these names. |
| 36 | |
| 37 | If DATA holds 0x00 or 0x08 and CMD holds 0x00 or 0xAC an uGuru could be |
| 38 | present. We have to check for two different values at data-port, because |
| 39 | after a reboot uGuru will hold 0x00 here, but if the driver is removed and |
| 40 | later on attached again data-port will hold 0x08, more about this later. |
| 41 | |
| 42 | After wider testing of the Linux kernel driver some variants of the uGuru have |
| 43 | turned up which will hold 0x00 instead of 0xAC at the CMD port, thus we also |
| 44 | have to test CMD for two different values. On these uGuru's DATA will initally |
| 45 | hold 0x09 and will only hold 0x08 after reading CMD first, so CMD must be read |
| 46 | first! |
| 47 | |
| 48 | To be really sure an uGuru is present a test read of one or more register |
| 49 | sets should be done. |
| 50 | |
| 51 | |
| 52 | Reading / Writing |
| 53 | ================= |
| 54 | |
| 55 | Addressing |
| 56 | ---------- |
| 57 | |
| 58 | The uGuru has a number of different addressing levels. The first addressing |
| 59 | level we will call banks. A bank holds data for one or more sensors. The data |
| 60 | in a bank for a sensor is one or more bytes large. |
| 61 | |
| 62 | The number of bytes is fixed for a given bank, you should always read or write |
| 63 | that many bytes, reading / writing more will fail, the results when writing |
| 64 | less then the number of bytes for a given bank are undetermined. |
| 65 | |
| 66 | See below for all known bank addresses, numbers of sensors in that bank, |
| 67 | number of bytes data per sensor and contents/meaning of those bytes. |
| 68 | |
| 69 | Although both this document and the kernel driver have kept the sensor |
| 70 | terminoligy for the addressing within a bank this is not 100% correct, in |
| 71 | bank 0x24 for example the addressing within the bank selects a PWM output not |
| 72 | a sensor. |
| 73 | |
| 74 | Notice that some banks have both a read and a write address this is how the |
| 75 | uGuru determines if a read from or a write to the bank is taking place, thus |
| 76 | when reading you should always use the read address and when writing the |
Frederik Schwarzer | 025dfda | 2008-10-16 19:02:37 +0200 | [diff] [blame] | 77 | write address. The write address is always one (1) more than the read address. |
Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 78 | |
| 79 | |
| 80 | uGuru ready |
| 81 | ----------- |
| 82 | |
| 83 | Before you can read from or write to the uGuru you must first put the uGuru |
| 84 | in "ready" mode. |
| 85 | |
| 86 | To put the uGuru in ready mode first write 0x00 to DATA and then wait for DATA |
| 87 | to hold 0x09, DATA should read 0x09 within 250 read cycles. |
| 88 | |
| 89 | Next CMD _must_ be read and should hold 0xAC, usually CMD will hold 0xAC the |
| 90 | first read but sometimes it takes a while before CMD holds 0xAC and thus it |
| 91 | has to be read a number of times (max 50). |
| 92 | |
| 93 | After reading CMD, DATA should hold 0x08 which means that the uGuru is ready |
| 94 | for input. As above DATA will usually hold 0x08 the first read but not always. |
| 95 | This step can be skipped, but it is undetermined what happens if the uGuru has |
| 96 | not yet reported 0x08 at DATA and you proceed with writing a bank address. |
| 97 | |
| 98 | |
| 99 | Sending bank and sensor addresses to the uGuru |
| 100 | ---------------------------------------------- |
| 101 | |
| 102 | First the uGuru must be in "ready" mode as described above, DATA should hold |
| 103 | 0x08 indicating that the uGuru wants input, in this case the bank address. |
| 104 | |
| 105 | Next write the bank address to DATA. After the bank address has been written |
| 106 | wait for to DATA to hold 0x08 again indicating that it wants / is ready for |
| 107 | more input (max 250 reads). |
| 108 | |
| 109 | Once DATA holds 0x08 again write the sensor address to CMD. |
| 110 | |
| 111 | |
| 112 | Reading |
| 113 | ------- |
| 114 | |
| 115 | First send the bank and sensor addresses as described above. |
| 116 | Then for each byte of data you want to read wait for DATA to hold 0x01 |
| 117 | which indicates that the uGuru is ready to be read (max 250 reads) and once |
| 118 | DATA holds 0x01 read the byte from CMD. |
| 119 | |
| 120 | Once all bytes have been read data will hold 0x09, but there is no reason to |
| 121 | test for this. Notice that the number of bytes is bank address dependent see |
| 122 | above and below. |
| 123 | |
Coly Li | 73ac36e | 2009-01-07 18:09:16 -0800 | [diff] [blame] | 124 | After completing a successful read it is advised to put the uGuru back in |
Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 125 | ready mode, so that it is ready for the next read / write cycle. This way |
| 126 | if your program / driver is unloaded and later loaded again the detection |
| 127 | algorithm described above will still work. |
| 128 | |
| 129 | |
| 130 | |
| 131 | Writing |
| 132 | ------- |
| 133 | |
| 134 | First send the bank and sensor addresses as described above. |
| 135 | Then for each byte of data you want to write wait for DATA to hold 0x00 |
| 136 | which indicates that the uGuru is ready to be written (max 250 reads) and |
| 137 | once DATA holds 0x00 write the byte to CMD. |
| 138 | |
| 139 | Once all bytes have been written wait for DATA to hold 0x01 (max 250 reads) |
| 140 | don't ask why this is the way it is. |
| 141 | |
| 142 | Once DATA holds 0x01 read CMD it should hold 0xAC now. |
| 143 | |
Coly Li | 73ac36e | 2009-01-07 18:09:16 -0800 | [diff] [blame] | 144 | After completing a successful write it is advised to put the uGuru back in |
Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 145 | ready mode, so that it is ready for the next read / write cycle. This way |
| 146 | if your program / driver is unloaded and later loaded again the detection |
| 147 | algorithm described above will still work. |
| 148 | |
| 149 | |
| 150 | Gotchas |
| 151 | ------- |
| 152 | |
| 153 | After wider testing of the Linux kernel driver some variants of the uGuru have |
| 154 | turned up which do not hold 0x08 at DATA within 250 reads after writing the |
| 155 | bank address. With these versions this happens quite frequent, using larger |
| 156 | timeouts doesn't help, they just go offline for a second or 2, doing some |
| 157 | internal callibration or whatever. Your code should be prepared to handle |
| 158 | this and in case of no response in this specific case just goto sleep for a |
| 159 | while and then retry. |
| 160 | |
| 161 | |
| 162 | Address Map |
| 163 | =========== |
| 164 | |
| 165 | Bank 0x20 Alarms (R) |
| 166 | -------------------- |
| 167 | This bank contains 0 sensors, iow the sensor address is ignored (but must be |
| 168 | written) just use 0. Bank 0x20 contains 3 bytes: |
| 169 | |
| 170 | Byte 0: |
| 171 | This byte holds the alarm flags for sensor 0-7 of Sensor Bank1, with bit 0 |
| 172 | corresponding to sensor 0, 1 to 1, etc. |
| 173 | |
| 174 | Byte 1: |
| 175 | This byte holds the alarm flags for sensor 8-15 of Sensor Bank1, with bit 0 |
| 176 | corresponding to sensor 8, 1 to 9, etc. |
| 177 | |
| 178 | Byte 2: |
| 179 | This byte holds the alarm flags for sensor 0-5 of Sensor Bank2, with bit 0 |
| 180 | corresponding to sensor 0, 1 to 1, etc. |
| 181 | |
| 182 | |
| 183 | Bank 0x21 Sensor Bank1 Values / Readings (R) |
| 184 | -------------------------------------------- |
| 185 | This bank contains 16 sensors, for each sensor it contains 1 byte. |
| 186 | So far the following sensors are known to be available on all motherboards: |
| 187 | Sensor 0 CPU temp |
| 188 | Sensor 1 SYS temp |
| 189 | Sensor 3 CPU core volt |
| 190 | Sensor 4 DDR volt |
| 191 | Sensor 10 DDR Vtt volt |
| 192 | Sensor 15 PWM temp |
| 193 | |
| 194 | Byte 0: |
| 195 | This byte holds the reading from the sensor. Sensors in Bank1 can be both |
| 196 | volt and temp sensors, this is motherboard specific. The uGuru however does |
| 197 | seem to know (be programmed with) what kindoff sensor is attached see Sensor |
| 198 | Bank1 Settings description. |
| 199 | |
| 200 | Volt sensors use a linear scale, a reading 0 corresponds with 0 volt and a |
| 201 | reading of 255 with 3494 mV. The sensors for higher voltages however are |
| 202 | connected through a division circuit. The currently known division circuits |
| 203 | in use result in ranges of: 0-4361mV, 0-6248mV or 0-14510mV. 3.3 volt sources |
| 204 | use the 0-4361mV range, 5 volt the 0-6248mV and 12 volt the 0-14510mV . |
| 205 | |
| 206 | Temp sensors also use a linear scale, a reading of 0 corresponds with 0 degree |
| 207 | Celsius and a reading of 255 with a reading of 255 degrees Celsius. |
| 208 | |
| 209 | |
| 210 | Bank 0x22 Sensor Bank1 Settings (R) |
| 211 | Bank 0x23 Sensor Bank1 Settings (W) |
| 212 | ----------------------------------- |
| 213 | |
| 214 | This bank contains 16 sensors, for each sensor it contains 3 bytes. Each |
| 215 | set of 3 bytes contains the settings for the sensor with the same sensor |
| 216 | address in Bank 0x21 . |
| 217 | |
| 218 | Byte 0: |
| 219 | Alarm behaviour for the selected sensor. A 1 enables the described behaviour. |
| 220 | Bit 0: Give an alarm if measured temp is over the warning threshold (RW) * |
| 221 | Bit 1: Give an alarm if measured volt is over the max threshold (RW) ** |
| 222 | Bit 2: Give an alarm if measured volt is under the min threshold (RW) ** |
| 223 | Bit 3: Beep if alarm (RW) |
| 224 | Bit 4: 1 if alarm cause measured temp is over the warning threshold (R) |
| 225 | Bit 5: 1 if alarm cause measured volt is over the max threshold (R) |
| 226 | Bit 6: 1 if alarm cause measured volt is under the min threshold (R) |
Frederik Schwarzer | 025dfda | 2008-10-16 19:02:37 +0200 | [diff] [blame] | 227 | Bit 7: Volt sensor: Shutdown if alarm persist for more than 4 seconds (RW) |
Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 228 | Temp sensor: Shutdown if temp is over the shutdown threshold (RW) |
| 229 | |
| 230 | * This bit is only honored/used by the uGuru if a temp sensor is connected |
| 231 | ** This bit is only honored/used by the uGuru if a volt sensor is connected |
| 232 | Note with some trickery this can be used to find out what kinda sensor is |
| 233 | detected see the Linux kernel driver for an example with many comments on |
| 234 | how todo this. |
| 235 | |
| 236 | Byte 1: |
| 237 | Temp sensor: warning threshold (scale as bank 0x21) |
| 238 | Volt sensor: min threshold (scale as bank 0x21) |
| 239 | |
| 240 | Byte 2: |
| 241 | Temp sensor: shutdown threshold (scale as bank 0x21) |
| 242 | Volt sensor: max threshold (scale as bank 0x21) |
| 243 | |
| 244 | |
| 245 | Bank 0x24 PWM outputs for FAN's (R) |
| 246 | Bank 0x25 PWM outputs for FAN's (W) |
| 247 | ----------------------------------- |
| 248 | |
| 249 | This bank contains 3 "sensors", for each sensor it contains 5 bytes. |
| 250 | Sensor 0 usually controls the CPU fan |
| 251 | Sensor 1 usually controls the NB (or chipset for single chip) fan |
| 252 | Sensor 2 usually controls the System fan |
| 253 | |
| 254 | Byte 0: |
| 255 | Flag 0x80 to enable control, Fan runs at 100% when disabled. |
| 256 | low nibble (temp)sensor address at bank 0x21 used for control. |
| 257 | |
| 258 | Byte 1: |
| 259 | 0-255 = 0-12v (linear), specify voltage at which fan will rotate when under |
| 260 | low threshold temp (specified in byte 3) |
| 261 | |
| 262 | Byte 2: |
| 263 | 0-255 = 0-12v (linear), specify voltage at which fan will rotate when above |
| 264 | high threshold temp (specified in byte 4) |
| 265 | |
| 266 | Byte 3: |
| 267 | Low threshold temp (scale as bank 0x21) |
| 268 | |
| 269 | byte 4: |
| 270 | High threshold temp (scale as bank 0x21) |
| 271 | |
| 272 | |
| 273 | Bank 0x26 Sensors Bank2 Values / Readings (R) |
| 274 | --------------------------------------------- |
| 275 | |
| 276 | This bank contains 6 sensors (AFAIK), for each sensor it contains 1 byte. |
| 277 | So far the following sensors are known to be available on all motherboards: |
| 278 | Sensor 0: CPU fan speed |
| 279 | Sensor 1: NB (or chipset for single chip) fan speed |
| 280 | Sensor 2: SYS fan speed |
| 281 | |
| 282 | Byte 0: |
| 283 | This byte holds the reading from the sensor. 0-255 = 0-15300 (linear) |
| 284 | |
| 285 | |
| 286 | Bank 0x27 Sensors Bank2 Settings (R) |
| 287 | Bank 0x28 Sensors Bank2 Settings (W) |
| 288 | ------------------------------------ |
| 289 | |
| 290 | This bank contains 6 sensors (AFAIK), for each sensor it contains 2 bytes. |
| 291 | |
| 292 | Byte 0: |
| 293 | Alarm behaviour for the selected sensor. A 1 enables the described behaviour. |
| 294 | Bit 0: Give an alarm if measured rpm is under the min threshold (RW) |
| 295 | Bit 3: Beep if alarm (RW) |
Frederik Schwarzer | 025dfda | 2008-10-16 19:02:37 +0200 | [diff] [blame] | 296 | Bit 7: Shutdown if alarm persist for more than 4 seconds (RW) |
Hans de Goede | f2b84bb | 2006-06-04 20:22:24 +0200 | [diff] [blame] | 297 | |
| 298 | Byte 1: |
| 299 | min threshold (scale as bank 0x26) |
| 300 | |
| 301 | |
| 302 | Warning for the adventerous |
| 303 | =========================== |
| 304 | |
| 305 | A word of caution to those who want to experiment and see if they can figure |
| 306 | the voltage / clock programming out, I tried reading and only reading banks |
| 307 | 0-0x30 with the reading code used for the sensor banks (0x20-0x28) and this |
| 308 | resulted in a _permanent_ reprogramming of the voltages, luckily I had the |
| 309 | sensors part configured so that it would shutdown my system on any out of spec |
| 310 | voltages which proprably safed my computer (after a reboot I managed to |
| 311 | immediatly enter the bios and reload the defaults). This probably means that |
| 312 | the read/write cycle for the non sensor part is different from the sensor part. |