| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 1 | Kernel driver k8temp |
| 2 | ==================== |
| 3 | |
| 4 | Supported chips: |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 5 | |
| Rudolf Marek | 4660cb3 | 2006-10-08 22:01:26 +0200 | [diff] [blame] | 6 | * AMD Athlon64/FX or Opteron CPUs |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 7 | |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 8 | Prefix: 'k8temp' |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 9 | |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 10 | Addresses scanned: PCI space |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 11 | |
| Alexander A. Klimov | f475356 | 2020-07-19 20:35:48 +0200 | [diff] [blame] | 12 | Datasheet: https://www.amd.com/system/files/TechDocs/32559.pdf |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 13 | |
| 14 | Author: Rudolf Marek |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 15 | |
| Jean Delvare | 7188cc6 | 2006-12-12 18:18:30 +0100 | [diff] [blame] | 16 | Contact: Rudolf Marek <r.marek@assembler.cz> |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 17 | |
| 18 | Description |
| 19 | ----------- |
| 20 | |
| Rudolf Marek | 4660cb3 | 2006-10-08 22:01:26 +0200 | [diff] [blame] | 21 | This driver permits reading temperature sensor(s) embedded inside AMD K8 |
| 22 | family CPUs (Athlon64/FX, Opteron). Official documentation says that it works |
| 23 | from revision F of K8 core, but in fact it seems to be implemented for all |
| 24 | revisions of K8 except the first two revisions (SH-B0 and SH-B3). |
| 25 | |
| 26 | Please note that you will need at least lm-sensors 2.10.1 for proper userspace |
| 27 | support. |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 28 | |
| 29 | There can be up to four temperature sensors inside single CPU. The driver |
| 30 | will auto-detect the sensors and will display only temperatures from |
| 31 | implemented sensors. |
| 32 | |
| 33 | Mapping of /sys files is as follows: |
| 34 | |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 35 | ============= =================================== |
| 36 | temp1_input temperature of Core 0 and "place" 0 |
| 37 | temp2_input temperature of Core 0 and "place" 1 |
| 38 | temp3_input temperature of Core 1 and "place" 0 |
| 39 | temp4_input temperature of Core 1 and "place" 1 |
| 40 | ============= =================================== |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 41 | |
| 42 | Temperatures are measured in degrees Celsius and measurement resolution is |
| 43 | 1 degree C. It is expected that future CPU will have better resolution. The |
| 44 | temperature is updated once a second. Valid temperatures are from -49 to |
| 45 | 206 degrees C. |
| 46 | |
| 47 | Temperature known as TCaseMax was specified for processors up to revision E. |
| 48 | This temperature is defined as temperature between heat-spreader and CPU |
| 49 | case, so the internal CPU temperature supplied by this driver can be higher. |
| 50 | There is no easy way how to measure the temperature which will correlate |
| 51 | with TCaseMax temperature. |
| 52 | |
| 53 | For newer revisions of CPU (rev F, socket AM2) there is a mathematically |
| 54 | computed temperature called TControl, which must be lower than TControlMax. |
| 55 | |
| 56 | The relationship is following: |
| 57 | |
| Mauro Carvalho Chehab | 0d92562 | 2019-04-17 06:46:25 -0300 | [diff] [blame] | 58 | temp1_input - TjOffset*2 < TControlMax, |
| Rudolf Marek | ef4c4fd | 2006-09-24 20:50:57 +0200 | [diff] [blame] | 59 | |
| 60 | TjOffset is not yet exported by the driver, TControlMax is usually |
| 61 | 70 degrees C. The rule of the thumb -> CPU temperature should not cross |
| 62 | 60 degrees C too much. |