Re: [lm-sensors] sensors-detect killed my CPU

[Achim Gottinger <achim@ag-web.biz>]

Jean Delvare schrieb:
> On Sun, 04 May 2008 17:27:22 +0200, achim wrote:
>   
>> Now with the 9600BE in the sapphire board there are additional chips 
>>
>> X2 5000BE
>> ---------------------------------------------------
>>     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
>> 00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
>> 10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
>> 20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- 2e -- 
>> 30: -- -- -- -- -- -- -- -- 38 -- -- -- -- -- -- -- 
>> 40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- 4e -- 
>> 50: -- -- 52 53 -- -- -- 57 -- -- -- -- -- -- -- -- 
>> 60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- 6e -- 
>> 70: 70 -- -- -- -- -- -- --                         
>> ---------------------------------------------------
>>
>> 9600BE
>> ---------------------------------------------------
>>      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
>> 00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
>> 10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
>> 20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- 2e -- 
>> 30: -- -- -- -- -- -- -- -- 38 -- -- -- -- -- -- -- 
>> 40: -- -- -- -- -- -- -- 47 -- -- -- -- 4c -- 4e -- 
>> 50: -- -- 52 53 -- -- -- 57 -- -- -- -- -- -- -- -- 
>> 60: -- -- -- -- -- -- -- -- -- 69 -- -- -- -- 6e -- 
>> 70: 70 -- -- -- -- -- -- --                      
>> ---------------------------------------------------
>>
>> Now the clock chips shows up at 0x69 maybe a second run with the X2
>> would also have shown it.
>>     
>
> Strange... This could be related to the chip at 0x70 if it's an I2C mux
> as I suspect. Another possibility would be that the extra addresses are
> in the CPU itself, but at least for 0x69 we know it's not the case.
>   
I searched the "Bios and Kernel Developer Guide" for smbus and found 
those results

http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/31116.PDF

------------------------------------------------------------------------------------------------
• SMBus. System management bus. Refers to the protocol on which the 
serial VID interface (SVI) commands
and SBI are based. See section 2.4.1 [Processor Power Planes And Voltage 
Control] on page 28, 2.13.3
[Sideband Interface (SBI)] on page 123, and section 1.2 [Reference 
Documents] on page 13.
------------------------------------------------------------------------------------------------
2.4.1 Processor Power Planes And Voltage Control

The processor includes the following power planes:
• VDDIO: used for the DRAM and miscellaneous pins on DDR products only. 
Voltage level is nominally 1.8V
or 1.9V in support of DDR2; 1.5V in support of DDR3. This plane is 
powered during S3 (suspend to RAM).
• VTT: used for the DDR DRAM interface. Voltage level is specified to be 
half of the VDDIO level. This
plane is powered during S3 (suspend to RAM). See section 2.4.4 [ACPI 
Suspend to RAM State (S3)] on
page 52.
• VLDT: used for each of the links. Voltage level is nominally 1.2V.
• VDDA: filtered PLL supply. Voltage level is nominally 2.5V.
• VDD or VDD[1:0]: main supply for core logic. “VDD” refers generically 
to the core voltage plane(s). Voltage
level is specified by the VID interface.
• VDDNB: main supply for NB logic. Voltage level specified by the VID 
interface.
The voltage level of VDD and VDDNB may be altered in various states to 
control power consumption. All the
other supplies are fixed. Refer to the EDS for power plane sequencing 
requirements.
The processor includes two interfaces, intended to control external 
voltage regulators, called the parallel VID
(voltage level identifier) interface (PVI) and the serial VID interface 
(SVI). The PVI is a simple 6-bit VID code
provided on 6 pins. The SVI encodes voltage regulator control commands, 
including the VID code, using
SMBus protocol over two pins, SVD and SVC, to generate write commands to 
external voltage regulators. The
processor is the master and the voltage regulator(s) are the slave(s). 
Both pins are outputs of the master; SVD is
driven by the slave as well. SVC is a clock that strobes the data pin, 
SVD, on the rising edge. Refer to the AMD
------------------------------------------------------------------------------------------------
2.13.3 Sideband Interface (SBI)

The sideband interface (SBI) is an SMBus v2.0 compatible 2-wire 
processor slave interface. SBI is also
referred as the Advanced Platform Management Link. All I2C v2.1 speeds 
are supported.
SBI is used to communicate withthe Temperature Sensor Interface (SB-TSI) 
(see the SBI Temperature Sensor
Interface (SB-TSI) Specification, #40821).
------------------------------------------------------------------------------------------------
Information about the SBI SMBus register address can be read from the 
pci register F3x1E8. lspci does not output values beyond 0xFF. However 
baredit
under windows does. I could not find the SB-TSI document at the amd website.
Will look at this thing again tomorrow.
>   
>> If I run sensors-detect and pass 0x2e,0x6e,0x47 as excludes, will it
>> pass?
>>     
>
> Yes, it should pass.
>   
Good, i'll try that.

I also skimmed over the SB600 doc.
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/42119_sb600_ds_nda_3.02.pdf
Maybe the chip at 0x38 is the south bridge. On the M3A and the Sapphire 
this is SB600 and on the GBT780G it's SB700. The M2A-VM might simply not 
use the smbus.



_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors