[lm-sensors] FB-DIMM temperature

[lm-sensors] FB-DIMM temperature

[Rudolf Marek]

Hello,

Thanks for the improvements.

> 
> 	#include <sys/mman.h>
> 	#include <sys/stat.h>
> 	#include <fcntl.h>
> 	#include <stdio.h>
> 
> 	int main(void)
> 	{
> 	    int i;
> 	    unsigned char *addr;
> 	    int fd = open("/dev/mem", O_RDONLY);
> 
> 	    if ( fd = -1 ) {
> 		perror("failed to open /dev/mem");
> 		return 1;
> 	    }
> 
> 	    addr = mmap(0, 128*1024*1024, PROT_READ, MAP_PRIVATE, fd, 0xFE000000);
> 
> 	    if ( addr = MAP_FAILED ) {
> 		perror("mmap failed");
> 		return 1;
> 	    }
> 
> 	    /* we want function 3, 4 channels, 16AMB/channel */
> 	    for ( i = 0; i < 64; i++ ) {
> 		int idx = i*2048;
> 		if ( addr[idx] != 0xff || addr[idx+1] != 0xff ) {
> 		    printf("Intel sig at %i: %02x %02x\n",
> 	                   i, addr[idx], addr[idx+1]);

We should see the 86 80



> 		    /* offset 85h function 3 */
> 		    printf("Current temp is: %fC\n", addr[idx+(256*3)+0x85]/2.);
> 		}
> 	    }
> 
> 	    munmap(addr, 128*1024*1024);
> 
> 	    return 0;
> 	}
> 
> and here are the results:
> 
> 	# ./fbdt
> 	Intel sig at 0: 1d 11
> 	Current temp is: 71.500000C
> 	Intel sig at 16: 1d 11
> 	Current temp is: 0.000000C
> 	Intel sig at 32: 1d 11
> 	Current temp is: 61.000000C
> 	Intel sig at 48: 1d 11
> 	Current temp is: 0.000000C
> 

Hmm there is something wrong. Please provide output of lspci -vvv
and lspci -xxx

Do you have the windows software somewhere to download?

Rudolf

[lm-sensors] FB-DIMM temperature

[Vadim Zeitlin]

On Thu, 11 Jan 2007 16:34:44 +0100 Rudolf Marek <r.marek at assembler.cz> wrote:

RM> > 		    printf("Intel sig at %i: %02x %02x\n",
RM> > 	                   i, addr[idx], addr[idx+1]);
RM> 
RM> We should see the 86 80

 Ah, this is why it's called "Intel sig", I see...

RM> > and here are the results:
RM> > 
RM> > 	# ./fbdt
RM> > 	Intel sig at 0: 1d 11
RM> > 	Current temp is: 71.500000C
RM> > 	Intel sig at 16: 1d 11
RM> > 	Current temp is: 0.000000C
RM> > 	Intel sig at 32: 1d 11
RM> > 	Current temp is: 61.000000C
RM> > 	Intel sig at 48: 1d 11
RM> > 	Current temp is: 0.000000C
RM> > 
RM> 
RM> Hmm there is something wrong.

 FWIW, I've opened the box to try to cool the DIMMs a bit and now the
results are

	# ./fbdt
	Intel sig at 0: 1d 11
	Current temp is: 65.000000C
	Intel sig at 16: 1d 11
	Current temp is: 0.000000C
	Intel sig at 32: 1d 11
	Current temp is: 52.000000C
	Intel sig at 48: 1d 11
	Current temp is: 54.000000C

so not only the values did diminish but there is also 54 instead of 0 for
the last one. So I wonder: could it be that the DIMMs really overheat and
the temp of the 2nd one is just out of the range?

RM> Please provide output of lspci -vvv and lspci -xxx

 The files are relatively big so instead of posting them here I put them on

	http://www.tt-solutions.com/vz/hardware/tyan2696/pci.ls
	http://www.tt-solutions.com/vz/hardware/tyan2696/pci.dump

Please let me know if you need anything else. I'm afraid I don't really see
anything helpful in lspci output on my own.

RM> Do you have the windows software somewhere to download?

 No, not really, I use SpeedFan (http://www.almico.com/speedfan.php) but I
don't know if it's capable of reading FB-DIMMs temperatures and, as I said,
I don't really want to install Windows on this machine just to try it out.

 Thanks!
VZ

[lm-sensors] FB-DIMM temperature

[Rudolf Marek]

Vadim Zeitlin wrote:
> On Thu, 11 Jan 2007 16:34:44 +0100 Rudolf Marek <r.marek at assembler.cz> wrote:
> 
> RM> > 		    printf("Intel sig at %i: %02x %02x\n",
> RM> > 	                   i, addr[idx], addr[idx+1]);
> RM> 
> RM> We should see the 86 80
> 
>  Ah, this is why it's called "Intel sig", I see...

The 0x111d signature is company called IDT. They manufacture IDTAMB0480 which
may be used there. Question is if the register layout is same as in intel AMB

 http://www.intel.com/design/chipsets/datashts/313072.htm


> RM> > and here are the results:
> RM> > 
> RM> > 	# ./fbdt
> RM> > 	Intel sig at 0: 1d 11
> RM> > 	Current temp is: 71.500000C
> RM> > 	Intel sig at 16: 1d 11
> RM> > 	Current temp is: 0.000000C
> RM> > 	Intel sig at 32: 1d 11
> RM> > 	Current temp is: 61.000000C
> RM> > 	Intel sig at 48: 1d 11
> RM> > 	Current temp is: 0.000000C
> RM> > 
> RM> 
> RM> Hmm there is something wrong.
> 
>  FWIW, I've opened the box to try to cool the DIMMs a bit and now the
> results are
> 
> 	# ./fbdt
> 	Intel sig at 0: 1d 11
> 	Current temp is: 65.000000C
> 	Intel sig at 16: 1d 11
> 	Current temp is: 0.000000C
> 	Intel sig at 32: 1d 11
> 	Current temp is: 52.000000C
> 	Intel sig at 48: 1d 11
> 	Current temp is: 54.000000C
> 
> so not only the values did diminish but there is also 54 instead of 0 for
> the last one. So I wonder: could it be that the DIMMs really overheat and
> the temp of the 2nd one is just out of the range?

Yes it might. Please check the intel datasheet for overheat flags etc etc. It
might work for you. I asked our friend in Tyan for a help with that.




> 
> RM> Please provide output of lspci -vvv and lspci -xxx
> 
>  The files are relatively big so instead of posting them here I put them on
> 
> 	http://www.tt-solutions.com/vz/hardware/tyan2696/pci.ls
> 	http://www.tt-solutions.com/vz/hardware/tyan2696/pci.dump
> 
> Please let me know if you need anything else. I'm afraid I don't really see
> anything helpful in lspci output on my own.

I was interrested in base address and it is correct.

00:10.0 Host bridge: Intel Corporation 5000 Series Chipset Error Reporting
Registers (rev 31)
00: 86 80 f0 25 00 00 00 00 31 00 00 06 00 00 80 00
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 f1 10 96 26
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
40: 00 00 ff 00 00 08 00 00 00 00 00 fe 00 00 00 00

                               ^^^^^^^^^^^^^^^

Same as we use in our prog ;)

50: 00 00 02 00 00 00 04 04 00 10 11 01 00 00 13 11
60: 00 12 09 0d 00 e0 00 00 ff ff ff ff 00 00 00 00
70: 8c c0 e2 0f 00 00 00 00 8c c0 e2 0f 00 00 00 00
80: 01 00 80 00 00 00 00 00 00 00 00 00 00 00 00 00
90: 04 01 80 00 00 00 00 00 00 00 00 00 00 00 00 00
a0: 00 00 00 00 00 00 00 00 00 00 00 00 02 06 80 00
b0: 00 00 00 00 00 00 00 00 00 00 00 00 08 07 80 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 11 01 00 00
e0: 00 00 00 00 00 00 00 80 00 00 00 00 00 00 00 00
f0: 5c 7f 48 00 06 a1 58 01 f4 08 d8 03 80 00 00 00

> RM> Do you have the windows software somewhere to download?
> 
>  No, not really, I use SpeedFan (http://www.almico.com/speedfan.php) but I
> don't know if it's capable of reading FB-DIMMs temperatures and, as I said,
> I don't really want to install Windows on this machine just to try it out.

I think they do not have support for this - it fairly new technology.

If you have time, maybe you may investigate the overtemperature bits. I think I
may have some time for this next week...

Rudolf

[lm-sensors] FB-DIMM temperature

[Vadim Zeitlin]

On Thu, 11 Jan 2007 17:57:38 +0100 Rudolf Marek <r.marek at assembler.cz> wrote:

RM> The 0x111d signature is company called IDT. They manufacture IDTAMB0480
RM> which may be used there.

 FWIW the FB-DIMMs are 1GB PC5300 modules manufactured by Samsung but I
don't have the model number at hand (but I could look it up).

RM> Question is if the register layout is same as in intel AMB
RM> 
RM>  http://www.intel.com/design/chipsets/datashts/313072.htm

 Thanks for the link! I didn't find anything as detailed about IDTAMB0480
though, all I could find was this:

    http://www.idt.com/products/files/6186383/AMB0480_product_brief.pdf

Astonishingly, at least to me, the max temperature for these things seems
to be 124C: see page 12 of the above link or section 3.3.1 at page 25 of
http://download.intel.com/design/chipsets/designex/31306603.pdf

 Do I really understand this correctly? Or should it be divided by 2?

 I'll read Intel datasheets a bit more tonight, I'm sure they're ideal
bed-side reading ;-)

 Regards,
VZ

[lm-sensors] FB-DIMM temperature

[Vadim Zeitlin]

On Thu, 11 Jan 2007 17:57:38 +0100 Rudolf Marek <r.marek at assembler.cz> wrote:

RM> >  FWIW, I've opened the box to try to cool the DIMMs a bit and now the
RM> > results are
RM> > 
RM> > 	# ./fbdt
RM> > 	Intel sig at 0: 1d 11
RM> > 	Current temp is: 65.000000C
RM> > 	Intel sig at 16: 1d 11
RM> > 	Current temp is: 0.000000C
RM> > 	Intel sig at 32: 1d 11
RM> > 	Current temp is: 52.000000C
RM> > 	Intel sig at 48: 1d 11
RM> > 	Current temp is: 54.000000C
RM> > 
RM> > so not only the values did diminish but there is also 54 instead of 0 for
RM> > the last one.

 I think I know at least why the temperature is sometimes 0: it's a timing
issue, apparently you can't read the sensors too quickly. If I insert
usleep(1000) between access to the signature bytes and the temperature value,
I can reliably read non zero values. I also looked at the values of the other
registers described in the section 14.5.3 of Intel datasheet, here is the
full output:

	Sig at 0: 1d 11
	Device id: 80 04
	TEMPLO:         106.00C
	TEMPMID:        109.00C
	TEMPHI:         124.00C
	TEMPSTAT:       0x1
	TEMP:           68.50C
	Sig at 16: 1d 11
	Device id: 80 04
	TEMPLO:         106.00C
	TEMPMID:        109.00C
	TEMPHI:         124.00C
	TEMPSTAT:       0x1
	TEMP:           68.00C
	Sig at 32: 1d 11
	Device id: 80 04
	TEMPLO:         106.00C
	TEMPMID:        109.00C
	TEMPHI:         124.00C
	TEMPSTAT:       0x1
	TEMP:           59.00C
	Sig at 48: 1d 11
	Device id: 80 04
	TEMPLO:         106.00C
	TEMPMID:        109.00C
	TEMPHI:         124.00C
	TEMPSTAT:       0x1
	TEMP:           63.50C

So, assuming I interpret TEMPLO/MID/HI correctly (I also divided their
values by 2 to get the temperature in degrees Celsius), it really does look
that these DIMMs are supposed to run quite hot.

 Please let me know if I'm wrong (before I fry them :-),
VZ

[lm-sensors] FB-DIMM temperature

[Rudolf Marek]

Hello Vadim,

>  I think I know at least why the temperature is sometimes 0: it's a timing
> issue, apparently you can't read the sensors too quickly. If I insert
> usleep(1000) between access to the signature bytes and the temperature value,
> I can reliably read non zero values. 

Ah, that is not too convenient :/ Maybe we need to read only 4 bytes at once. I
must check the Northbridge datasheet for this. Also it might be worth to check
the 14.5.3.4 UPDATED: Update Temp Diff Bit maybe it is zero when this bit is set.

>I also looked at the values of the other
> registers described in the section 14.5.3 of Intel datasheet, here is the
> full output:

I have some news too. The registers must match by definition with the Intel
public document.

> 
> 	Sig at 0: 1d 11
> 	Device id: 80 04
> 	TEMPLO:         106.00C
> 	TEMPMID:        109.00C
> 	TEMPHI:         124.00C
> 	TEMPSTAT:       0x1
> 	TEMP:           68.50C
> 	Sig at 16: 1d 11
> 	Device id: 80 04
> 	TEMPLO:         106.00C
> 	TEMPMID:        109.00C
> 	TEMPHI:         124.00C
> 	TEMPSTAT:       0x1
> 	TEMP:           68.00C
> 	Sig at 32: 1d 11
> 	Device id: 80 04
> 	TEMPLO:         106.00C
> 	TEMPMID:        109.00C
> 	TEMPHI:         124.00C
> 	TEMPSTAT:       0x1
> 	TEMP:           59.00C
> 	Sig at 48: 1d 11
> 	Device id: 80 04
> 	TEMPLO:         106.00C
> 	TEMPMID:        109.00C
> 	TEMPHI:         124.00C
> 	TEMPSTAT:       0x1
> 	TEMP:           63.50C
> 
> So, assuming I interpret TEMPLO/MID/HI correctly (I also divided their
> values by 2 to get the temperature in degrees Celsius), it really does look
> that these DIMMs are supposed to run quite hot.

>  Please let me know if I'm wrong (before I fry them :-),

No this are thresholds for the action, when system become too hot. Your temps
are around 70C now, so your system is safe.

No need to panic ;)

Rudolf