• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

K10 and 11 core temps

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Wrong, but your statement considering the accuracy under idle/load is correct.

Actually I am right (personally speaking) Phenom/Opteron/Bulldozer/K10 + only have 1 DTS sensor per package. not per core..:thup:
 
So 1 sensor is able to report temperatures for all of the cores? :-/

Quote taken from The coolest core temp site alcpu.com

AMD processors report the temperature via a special register in the CPU's northbridge. Core Temp reads the value from the register and uses a formula provided by AMD to calculate the current temperature.
The formula for the Athlon 64 series, early Opterons and Semprons (K8 architecture) is: 'Core Temp = Value - 49'.
For the newer generation of AMD processors like Phenom, Phenom II, newer Athlons, Semprons and Opterons (K10 architecture and up), and their derivatives, there is a different formula: 'CPU Temp* = Value / 8'*CPU Temp is because the Phenom\Opteron (K10) have only one sensor per package, meaning there is only one reading per processor.

so Yes in a nutshell. :) by calculating the above formula you get the temp for your whole processor regardless of how many cores it has.
 
Last edited:
Thread got a bit derailed. The below 2 screenshots were just taken, the first is idle and the second is load. Easytune shows the motherboard sensor below the CPU socket, Coretemp shows the CPU sensor embeded in the CPU die, and HWMonitor shows both. The CPU is an FX8120 @ 4.0Ghz using a Silver Arrow (22C is current ambient).

At idle (first screenshot) Core temp reads 14C as does HWMonitor, this is the supposed core temp, the motherboard sensor reads 26C in Easytune and HWMonitor (TMPIN2). A delta of 12C between the CPU DTS and the Motherboard DTS.

At load (second screenshot) Core temp and HWMonitor read 35C for the CPU DTS, the motherboard DTS reads 47C. Again a delta of 12C. So the scales between the 2 DTS's are the same (if plotted the slope of the lines would be the same), however the CPU DTS must be wrong since the ambient is 22C and at idle it reads 8C below ambient which is impossible using air cooling (or water). The motherboard sensor at idle reads 26C, again with an ambient of 22C is much more believable.

The 1090T I had (recently sold) read the same way using an older model motherboard (890FX instead of 990FX), however it's DTS was further off, around 15C too low.

The point I'm making is the CPU DTS is an arbitrarily chosen value using the degrees scale, the value itself is meaningless without knowing how far off it is. Also each processor maybe different, again because the value is arbitrary (see white paper above).
 

Attachments

  • idle.png
    idle.png
    107.2 KB · Views: 57
  • load.png
    load.png
    129.7 KB · Views: 56
Last edited:
Thread got a bit derailed. The below 2 screenshots were just taken, the first is idle and the second is load. Easytune shows the motherboard sensor below the CPU socket, Coretemp shows the CPU sensor embeded in the CPU die, and HWMonitor shows both. The CPU is an FX8120 @ 4.0Ghz using a Silver Arrow (22C is current ambient).

At idle (first screenshot) Core temp reads 14C as does HWMonitor, this is the supposed core temp, the motherboard sensor reads 26C in Easytune and HWMonitor (TMPIN2). A delta of 12C between the CPU DTS and the Motherboard DTS.

At load (second screenshot) Core temp and HWMonitor read 35C for the CPU DTS, the motherboard DTS reads 47C. Again a delta of 12C. So the scales between the 2 DTS's are the same (if plotted the slope of the lines would be the same), however the CPU DTS must be wrong since the ambient is 22C and at idle it reads 8C below ambient which is impossible using air cooling (or water). The motherboard sensor at idle reads 26C, again with an ambient of 22C is much more believable.

The 1090T I had (recently sold) read the same way using an older model motherboard (890FX instead of 990FX), however it's DTS was further off, around 15C too low.

The point I'm making is the CPU DTS is an arbitrarily chosen value using the degrees scale, the value itself is meaningless without knowing how far off it is. Also each processor maybe different, again because the value is arbitrary (see white paper above).


Sorry for taking the thread off topic OP.

I have the 890FX chipset and FX-8150 as a comparison I took these screenies @idle and under full load of p95. I use core temp 1.0 rc2 & speedfan 4.46 beta 2.

As I said earlier the DTS get more accurate the hotter it gets. I guess im just used to it as my 1090T & 1055 was eactly the same...

heres temps @ idle and under full load

@ idle the core temp is out by 10 degrees according to the MB sensor
@ load the core temp is pretty bang on reading 54 from speedfan and core temp. no diffrence. I remebeer reading that the DTS sensor has problems at low temps. I will try and find the source
 

Attachments

Here is what I see. In the first screeshot speenfan's core temp is at 37C and coretemp's core temp is at 27C. In the second screenshot they read the same at 54C. This tells me one of the apps is wrong, one of them is not using the same scale as the other.

My guess is it's speedfan since I've worked with core temp and it follows the same scale as hwmonitor and hwinfo. However the delta between what speedfan reads between core temp and socket temp is 5C in both cases. I would check with hwmonitor to see if it correlates with what speenfan reads.
 
I just installed speedfan to check myself but it doesn't read my core temp, only socket temp (on 990FX).
 
I just installed speedfan to check myself but it doesn't read my core temp, only socket temp (on 990FX).

I have asked alfredo(speedfan programmer) to update/adress this in an email I sent 4 weeks ago when I first got my BD, the values that you see in my speedfan (name wise- I have changed myself) I have assumed that the top temp is core temp and 2nd temp is socket temp..which does match up under full load as you seen in the picture above..

This has been doing the same since all K10 cpus came out Phen/PhenII/BD so if anything its the DTS sensor as you mentioned earlier with the arbitrary values thats throwing the software off key.

If its using the cpu usage to calculate the formula for temp then 0% usage is hard to calculate. As the cpu usage % rises so does the accuracy of the DTS sensor as it has acutal values to calculate.. ( I am only guessing here no proof at all so dont take it as verbatim).

that is how I understand it.

It has also been mentioned that socket temp is not important its the core temp that is important.. which does ring true as the highest temp I have seen with BD was when I was @ 5.128Ghz @1.525vcore. The socket temp was 75c ( not possible - thermal protection would have kicked in as T-Jmax was achieved) and 62c on the core. both speedfan & core temp showed the same values. no shutdown was initiated.

which leads me to think that either the Asus sensor that we both have on our MB's is reading 10-13c higher than it actually is.

Having said all that I know from the forums on alcpu.com that the coolest is working on a few fixes pertaining to BD. so I would wait to pass judgement until that is released. either way it is slightly perplexing that there is no software to gaurantee what temp your running at you have to have several software suite running to get an average and then you base your desicion on that. well at least I do.

The more I research this the more interesting it gets :)

regards

Sal.
 
Last edited:
I have not found the thermal and power white papaer yet so TJmax is unknown. The max temp is listed as 61C (not necessarily TJmax) however if the offset is 12C then the socket temp could go as high 73C....I would go with AMD's engineer's on their DTS before I would on an app, I doubt that the DTS has any problem reading the temp with no load as they have nothing to do with each other, just my opinion. It could even be a bios issue, who knows. What I have seen is the same for a 965BE C3 (one of them did exhibit an offset and one did not, I had 2 of them), a 1090T, and now an FX8120 all yielding the same sort of offset on the core temp on 3 different motherboards (an Asus 890X, and 2 Gigabyte boards 890FXA and 990FX).
 
Back