7 Addresses scanned: I2C 0x28 - 0x2f, ISA 0x290 (8 I/O ports)
8 Datasheet: http://www.winbond-usa.com/products/winbond_products/pdfs/PCIC/w83781d.pdf
11 Addresses scanned: I2C 0x28 - 0x2f, ISA 0x290 (8 I/O ports)
12 Datasheet: http://www.winbond.com/PDF/sheet/w83782d.pdf
15 Addresses scanned: I2C 0x2d
16 Datasheet: http://www.winbond-usa.com/products/winbond_products/pdfs/PCIC/w83783s.pdf
19 Addresses scanned: I2C 0x28 - 0x2f, ISA 0x290 (8 I/O ports)
20 Datasheet: http://www.winbond.com/PDF/sheet/w83627hf.pdf
23 Addresses scanned: I2C 0x28 - 0x2f
24 Datasheet: Unavailable from Asus
27 Frodo Looijaard <frodol@dds.nl>,
28 Philip Edelbrock <phil@netroedge.com>,
29 Mark Studebaker <mdsxyz123@yahoo.com>
36 Use 'init=0' to bypass initializing the chip.
37 Try this if your computer crashes when you load the module.
41 The driver used to reset the chip on load, but does no more. Use
42 'reset=1' to restore the old behavior. Report if you need to do this.
44 force_subclients=bus,caddr,saddr,saddr
45 This is used to force the i2c addresses for subclients of
46 a certain chip. Typical usage is `force_subclients=0,0x2d,0x4a,0x4b'
47 to force the subclients of chip 0x2d on bus 0 to i2c addresses
48 0x4a and 0x4b. This parameter is useful for certain Tyan boards.
53 This driver implements support for the Winbond W83781D, W83782D, W83783S,
54 W83627HF chips, and the Asus AS99127F chips. We will refer to them
55 collectively as W8378* chips.
57 There is quite some difference between these chips, but they are similar
58 enough that it was sensible to put them together in one driver.
59 The W83627HF chip is assumed to be identical to the ISA W83782D.
60 The Asus chips are similar to an I2C-only W83782D.
62 Chip #vin #fanin #pwm #temp wchipid vendid i2c ISA
63 as99127f 7 3 0 3 0x31 0x12c3 yes no
64 as99127f rev.2 (type_name = as99127f) 0x31 0x5ca3 yes no
65 w83781d 7 3 0 3 0x10-1 0x5ca3 yes yes
66 w83627hf 9 3 2 3 0x21 0x5ca3 yes yes(LPC)
67 w83782d 9 3 2-4 3 0x30 0x5ca3 yes yes
68 w83783s 5-6 3 2 1-2 0x40 0x5ca3 yes no
70 Detection of these chips can sometimes be foiled because they can be in
71 an internal state that allows no clean access. If you know the address
72 of the chip, use a 'force' parameter; this will put them into a more
73 well-behaved state first.
75 The W8378* implements temperature sensors (three on the W83781D and W83782D,
76 two on the W83783S), three fan rotation speed sensors, voltage sensors
77 (seven on the W83781D, nine on the W83782D and six on the W83783S), VID
78 lines, alarms with beep warnings, and some miscellaneous stuff.
80 Temperatures are measured in degrees Celsius. There is always one main
81 temperature sensor, and one (W83783S) or two (W83781D and W83782D) other
82 sensors. An alarm is triggered for the main sensor once when the
83 Overtemperature Shutdown limit is crossed; it is triggered again as soon as
84 it drops below the Hysteresis value. A more useful behavior
85 can be found by setting the Hysteresis value to +127 degrees Celsius; in
86 this case, alarms are issued during all the time when the actual temperature
87 is above the Overtemperature Shutdown value. The driver sets the
88 hysteresis value for temp1 to 127 at initialization.
90 For the other temperature sensor(s), an alarm is triggered when the
91 temperature gets higher then the Overtemperature Shutdown value; it stays
92 on until the temperature falls below the Hysteresis value. But on the
93 W83781D, there is only one alarm that functions for both other sensors!
94 Temperatures are guaranteed within a range of -55 to +125 degrees. The
95 main temperature sensors has a resolution of 1 degree; the other sensor(s)
98 Fan rotation speeds are reported in RPM (rotations per minute). An alarm is
99 triggered if the rotation speed has dropped below a programmable limit. Fan
100 readings can be divided by a programmable divider (1, 2, 4 or 8 for the
101 W83781D; 1, 2, 4, 8, 16, 32, 64 or 128 for the others) to give
102 the readings more range or accuracy. Not all RPM values can accurately
103 be represented, so some rounding is done. With a divider of 2, the lowest
104 representable value is around 2600 RPM.
106 Voltage sensors (also known as IN sensors) report their values in volts.
107 An alarm is triggered if the voltage has crossed a programmable minimum
108 or maximum limit. Note that minimum in this case always means 'closest to
109 zero'; this is important for negative voltage measurements. All voltage
110 inputs can measure voltages between 0 and 4.08 volts, with a resolution
113 The VID lines encode the core voltage value: the voltage level your processor
114 should work with. This is hardcoded by the mainboard and/or processor itself.
115 It is a value in volts. When it is unconnected, you will often find the
118 The W83782D and W83783S temperature conversion machine understands about
119 several kinds of temperature probes. You can program the so-called
120 beta value in the sensor files. '1' is the PII/Celeron diode, '2' is the
121 TN3904 transistor, and 3435 the default thermistor value. Other values
122 are (not yet) supported.
124 In addition to the alarms described above, there is a CHAS alarm on the
125 chips which triggers if your computer case is open.
127 When an alarm goes off, you can be warned by a beeping signal through
128 your computer speaker. It is possible to enable all beeping globally,
129 or only the beeping for some alarms.
131 Individual alarm and beep bits:
138 0x000020: temp2 (+temp3 on W83781D)
146 0x002000: temp3 (W83782D and W83627HF only)
147 0x010000: in7 (W83782D and W83627HF only)
148 0x020000: in8 (W83782D and W83627HF only)
150 If an alarm triggers, it will remain triggered until the hardware register
151 is read at least once. This means that the cause for the alarm may
152 already have disappeared! Note that in the current implementation, all
153 hardware registers are read whenever any data is read (unless it is less
154 than 1.5 seconds since the last update). This means that you can easily
155 miss once-only alarms.
157 The chips only update values each 1.5 seconds; reading them more often
158 will do no harm, but will return 'old' values.
162 The as99127f support was developed without the benefit of a datasheet.
163 In most cases it is treated as a w83781d (although revision 2 of the
164 AS99127F looks more like a w83782d).
165 This support will be BETA until a datasheet is released.
166 One user has reported problems with fans stopping
169 Note that the individual beep bits are inverted from the other chips.
170 The driver now takes care of this so that user-space applications
171 don't have to know about it.
174 - Problems with diode/thermistor settings (supported?)
175 - One user reports fans stopping under high server load.
176 - Revision 2 seems to have 2 PWM registers but we don't know
177 how to handle them. More details below.
179 These will not be fixed unless we get a datasheet.
180 If you have problems, please lobby Asus to release a datasheet.
181 Unfortunately several others have without success.
182 Please do not send mail to us asking for better as99127f support.
183 We have done the best we can without a datasheet.
184 Please do not send mail to the author or the sensors group asking for
185 a datasheet or ideas on how to convince Asus. We can't help.
190 783s has no in1 so that in[2-6] are compatible with the 781d/782d.
192 783s pin is programmable for -5V or temp1; defaults to -5V,
193 no control in driver so temp1 doesn't work.
195 782d and 783s datasheets differ on which is pwm1 and which is pwm2.
196 We chose to follow 782d.
198 782d and 783s pin is programmable for fan3 input or pwm2 output;
199 defaults to fan3 input.
200 If pwm2 is enabled (with echo 255 1 > pwm2), then
203 782d has pwm1-2 for ISA, pwm1-4 for i2c. (pwm3-4 share pins with
208 - PWM clock registers:
217 Answers from Winbond tech support
218 ---------------------------------
220 > 1) In the W83781D data sheet section 7.2 last paragraph, it talks about
221 > reprogramming the R-T table if the Beta of the thermistor is not
222 > 3435K. The R-T table is described briefly in section 8.20.
223 > What formulas do I use to program a new R-T table for a given Beta?
225 We are sorry that the calculation for R-T table value is
226 confidential. If you have another Beta value of thermistor, we can help
227 to calculate the R-T table for you. But you should give us real R-T
228 Table which can be gotten by thermistor vendor. Therefore we will calculate
229 them and obtain 32-byte data, and you can fill the 32-byte data to the
230 register in Bank0.CR51 of W83781D.
233 > 2) In the W83782D data sheet, it mentions that pins 38, 39, and 40 are
234 > programmable to be either thermistor or Pentium II diode inputs.
235 > How do I program them for diode inputs? I can't find any register
236 > to program these to be diode inputs.
237 --> You may program Bank0 CR[5Dh] and CR[59h] registers.
239 CR[5Dh] bit 1(VTIN1) bit 2(VTIN2) bit 3(VTIN3)
245 (error) CR[59h] bit 4(VTIN1) bit 2(VTIN2) bit 3(VTIN3)
246 (right) CR[59h] bit 4(VTIN1) bit 5(VTIN2) bit 6(VTIN3)
248 PII thermal diode 1 1 1
255 We have no datasheets for the Asus clones (AS99127F and ASB100 Bach).
256 Here are some very useful information that were given to us by Alex Van
257 Kaam about how to detect these chips, and how to read their values. He
258 also gives advice for another Asus chipset, the Mozart-2 (which we
259 don't support yet). Thanks Alex!
260 I reworded some parts and added personal comments.
264 AS99127F rev.1, AS99127F rev.2 and ASB100:
265 - I2C address range: 0x29 - 0x2F
266 - If register 0x58 holds 0x31 then we have an Asus (either ASB100 or
268 - Which one depends on register 0x4F (manufacturer ID):
270 0x12 or 0xC3: AS99127F rev.1
271 0x5C or 0xA3: AS99127F rev.2
272 Note that 0x5CA3 is Winbond's ID (WEC), which let us think Asus get their
273 AS99127F rev.2 direct from Winbond. The other codes mean ATT and DVC,
274 respectively. ATT could stand for Asustek something (although it would be
275 very badly chosen IMHO), I don't know what DVC could stand for. Maybe
276 these codes simply aren't meant to be decoded that way.
280 - If register 0x58 holds 0x56 or 0x10 then we have a Mozart-2
281 - Of the Mozart there are 3 types:
282 0x58=0x56, 0x4E=0x94, 0x4F=0x36: Asus ASM58 Mozart-2
283 0x58=0x56, 0x4E=0x94, 0x4F=0x06: Asus AS2K129R Mozart-2
284 0x58=0x10, 0x4E=0x5C, 0x4F=0xA3: Asus ??? Mozart-2
285 You can handle all 3 the exact same way :)
287 # Temperature sensors:
290 - sensor 1: register 0x27
291 - sensor 2 & 3 are the 2 LM75's on the SMBus
292 - sensor 4: register 0x17
293 Remark: I noticed that on Intel boards sensor 2 is used for the CPU
294 and 4 is ignored/stuck, on AMD boards sensor 4 is the CPU and sensor 2 is
295 either ignored or a socket temperature.
297 AS99127F (rev.1 and 2 alike):
298 - sensor 1: register 0x27
299 - sensor 2 & 3 are the 2 LM75's on the SMBus
300 Remark: Register 0x5b is suspected to be temperature type selector. Bit 1
301 would control temp1, bit 3 temp2 and bit 5 temp3.
304 - sensor 1: register 0x27
305 - sensor 2: register 0x13
309 ASB100, AS99127F (rev.1 and 2 alike):
310 - 3 fans, identical to the W83781D
313 - 2 fans only, 1350000/RPM/div
314 - fan 1: register 0x28, divisor on register 0xA1 (bits 4-5)
315 - fan 2: register 0x29, divisor on register 0xA1 (bits 6-7)
319 This is where there is a difference between AS99127F rev.1 and 2.
320 Remark: The difference is similar to the difference between
327 in3=r(0x23)*0.016*1.68
328 in4=r(0x24)*0.016*3.8
329 in5=r(0x25)*(-0.016)*3.97
330 in6=r(0x26)*(-0.016)*1.666
336 in3=r(0x23)*0.016*1.68
337 in4=r(0x24)*0.016*3.8
338 in5=r(0x25)*(-0.016)*3.97
339 in6=r(0x26)*(-0.016)*1.503
345 in3=r(0x23)*0.016*1.68
346 in4=r(0x24)*0.016*3.8
347 in5=(r(0x25)*0.016-3.6)*5.14+3.6
348 in6=(r(0x26)*0.016-3.6)*3.14+3.6
354 in3=r(0x23)*0.016*1.68
362 Additional info about PWM on the AS99127F (may apply to other Asus
363 chips as well) by Jean Delvare as of 2004-04-09:
365 AS99127F revision 2 seems to have two PWM registers at 0x59 and 0x5A,
366 and a temperature sensor type selector at 0x5B (which basically means
367 that they swapped registers 0x59 and 0x5B when you compare with Winbond
369 Revision 1 of the chip also has the temperature sensor type selector at
370 0x5B, but PWM registers have no effect.
372 We don't know exactly how the temperature sensor type selection works.
373 Looks like bits 1-0 are for temp1, bits 3-2 for temp2 and bits 5-4 for
374 temp3, although it is possible that only the most significant bit matters
375 each time. So far, values other than 0 always broke the readings.
377 PWM registers seem to be split in two parts: bit 7 is a mode selector,
378 while the other bits seem to define a value or threshold.
380 When bit 7 is clear, bits 6-0 seem to hold a threshold value. If the value
381 is below a given limit, the fan runs at low speed. If the value is above
382 the limit, the fan runs at full speed. We have no clue as to what the limit
383 represents. Note that there seem to be some inertia in this mode, speed
384 changes may need some time to trigger. Also, an hysteresis mechanism is
385 suspected since walking through all the values increasingly and then
386 decreasingly led to slightly different limits.
388 When bit 7 is set, bits 3-0 seem to hold a threshold value, while bits 6-4
389 would not be significant. If the value is below a given limit, the fan runs
390 at full speed, while if it is above the limit it runs at low speed (so this
391 is the contrary of the other mode, in a way). Here again, we don't know
392 what the limit is supposed to represent.
394 One remarkable thing is that the fans would only have two or three
395 different speeds (transitional states left apart), not a whole range as
396 you usually get with PWM.
398 As a conclusion, you can write 0x00 or 0x8F to the PWM registers to make
399 fans run at low speed, and 0x7F or 0x80 to make them run at full speed.
401 Please contact us if you can figure out how it is supposed to work. As
402 long as we don't know more, the w83781d driver doesn't handle PWM on
403 AS99127F chips at all.
405 Additional info about PWM on the AS99127F rev.1 by Hector Martin:
407 I've been fiddling around with the (in)famous 0x59 register and
408 found out the following values do work as a form of coarse pwm:
410 0x80 - seems to turn fans off after some time(1-2 minutes)... might be
411 some form of auto-fan-control based on temp? hmm (Qfan? this mobo is an
412 old ASUS, it isn't marketed as Qfan. Maybe some beta pre-attemp at Qfan
413 that was dropped at the BIOS)
415 0x82 - slightly "on-ner" than off, but my fans do not get to move. I can
416 hear the high-pitched PWM sound that motors give off at too-low-pwm.
417 0x83 - now they do move. Estimate about 70% speed or so.
420 Changing the high nibble doesn't seem to do much except the high bit
421 (0x80) must be set for PWM to work, else the current pwm doesn't seem to
424 My mobo is an ASUS A7V266-E. This behavior is similar to what I got
425 with speedfan under Windows, where 0-15% would be off, 15-2x% (can't
426 remember the exact value) would be 70% and higher would be full on.
projects / linux-2.6 / blob