AMD 64 CPU Diode temperature?

  • Thread starter Thread starter str
  • Start date Start date
Well from my limited experience of it the on-die thermal diode gives a good indication if too high a CPU temp is what's causing Prime95 to fail or the system to freeze or crash while stress testing.

You mentioned previously your current setup failed Prime95 after a few hours and if you check Next Sensor you might find the on-die thermal diode is actually reporting a temp that highlights why it's failing ie the actual temp of the CPU is too hot but you can't really know that by relying on the motherboard reported CPU temp as 49c isn't really too hot.

So yeah check the on-die thermal diode via Next Sensor please mate. :D
 
Last edited:
OK! :D

I'm gonna leave mine running overnight. This DFI is way better than my MSI Neo Platinum so hopefully I'll wake up to a still running PC. :)
 
Last edited:
EDIT: Found the Thermal Offset in CBID, it's in the Mobility -> P-States tab. For my setup it's reporting a thermal offset of 15c and a junction offset of 20c. I don't really trust the results from CBID though and I suspect it's because the author is guessing which sensor chip is on the motherboard and getting it wrong and there doesn't appear to be a way to force it to use the correct sensor chip like with Next Sensor.

I've also noticed in CBID the Vcore is often wrong due to incorrect sensor chip detection. I'll fire of an email to the author asking about the problems with CBID and more information on Next Sensor's on-die thermal diode reading.

Might also ask later for some info to help me start writing my own sensor program. :)

I remember reading about a protocol the sensor reading programmers adopted so that running multiple sensor reading programs didn't cause conflicting results or errors due to two or more programs reading the sensor data at the same time. I also remember reading CBID doesn't use it instead it just reads the sensors as and when it wants to which could be why problems are occuring although so far I've seen zero problems with Next Sensor so maybe he adopted the protocol for it.

Also my system is still going with the Prime95s and the on-die thermal diode temp reporting has me confident it'll be stable at 2.6Ghz thanks to this DFI and a good working Freezer 64 Pro. :D
 
Last edited:
Pfffft... very long thread at DFI Street regarding the DFI Expert low temp issue! :D

I've read most of it and it looks like different CPU steppings produce different results due to whatever calculation is being done on the on-die thermal diode temp (in otherwords the motherboard reported CPU temp is based on the on-die thermal diode!). There isn't enough information about the calculation or the difference in results with different CPU steppings though but in general it doesn't look like it's reliable and in no way is it an exact science.

However, it seems obvious for me that the calculation on the DFI UT NF3 250Gb doesn't even come close to reality for my E6 stepping Sempron 2800+ (it's reporting the CPU temp way below what it should) and no doubt relying on it is actually dangerous seeing as the calculated temperature reported by the DFI was only 46c yet the on-die thermal diode temp was 80c in Next Sensor and the absolute max temp I should allow for my CPU is 69c according to AMD. :eek:
 
str said:
Juno, CPU-Z should report the sensor used on your motherboard and hopefully it's accurate.
Click to expand...
On my dfi expert it reckons the sensor is an ITE IT8712. This isn't an option in MBM or NxSensor. I wonder if there are new MBM configuration files for the expert? :confused:
 
next sensor.JPG
 
Last edited:
joeyjojo said:
On my dfi expert it reckons the sensor is an ITE IT8712. This isn't an option in MBM or NxSensor. I wonder if there are new MBM configuration files for the expert? :confused:
Click to expand...
In Next Sensor you should trying choosing ITE IT87xxF. You might also need to use the settings tab to rearrange the sensor lines so they match up with the labels on the main status page or if you're lucky they'll match up without needing to be manually configured.

Matty: looks like your motherboard is reporting the temp reasonably accurately compared to other Experts I read about on DFI Street. Also it seems likely a new cooler will give you what you need to get a stable overclock. :)
 
New version of CBID is out today and it now reports reliably the Vcore and CPU temperature on my system.

It has the following in the info.txt file:

Processor Tab: double click on the Voltage field changes Voltage Sensor
Mobility Tab: double click on the CPU Temp. field changes CPU Temp. Sensor
 
On a DFI NF3 Ultra-D, NXsensor's K8Thermdiode thingy shows the same as the sensor I was monitoring in MBM5. Liking the interface tho'. Very clean.
 
Jimbo Mahoney said:
Good point.

:o

I hadn't checked the box for the K8diodeythingy.

Man, the interface on this prog sux0rs.....talk about low contrast and hard to read.

@ 3.1 Ghz, 1.7v:

I'm getting 41'C fully loaded, and 46'C from the diode.
Click to expand...

You got DFI, enough said! :D
We ALL know the dfi-sensors are wrong. Lol, I dual prime95 at 25c! ;)
 
Changing the CPU sensor, I can get either:
0
13
43
35
Of which only 43 and 35 can be right, but both are hotter then the diode....although this is a K8N Neo2, which aren't exactly reknown for accurate temperature readings....

EDIT: Just doing a search (for superpi), and diode temp has gone upto 40c, and CPU temp remains at 36c....so this could be right
 
Last edited:
nicRob said:
Changing the CPU sensor, I can get either:
0
13
43
35
Of which only 43 and 35 can be right, but both are hotter then the diode....although this is a K8N Neo2, which aren't exactly reknown for accurate temperature readings....

EDIT: Just doing a search (for superpi), and diode temp has gone upto 40c, and CPU temp remains at 36c....so this could be right
Click to expand...
It auto detected the right one for me (same mobo) as the W83627THF.
42C from mobo sensor
38C from diode
246C from NB :D
 
You must log in or register to reply here.
Back
Top Bottom