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

Unaccurate Core Temps on AMD FX-4170

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
This is a two part post. Here's part one:

All temp sensors are based on electron flow, one way or another.
There are three primary flavors of sensor that I am aware of, two generally used outside of high density ICs and one used in them.

Outside option one is a thermocouple. Junctions between different metals will generate voltage, how much voltage varies depending on how hot the junction is. With very specific, pure, junctions very accurate temps can be read this way. In this sensor, the sensor itself causes electron flow.

Outside option two is a thermistor. It is a resistor that changes resistance dramatically based on its temp. All resistors change, thermistors change a lot, and by predictable and repeatable amounts. In this sensor, the sensor resists electron flow.

The internal option is a thermal diode. It is a diode with a voltage drop that changes significantly and predictably based on its temperature. This is what Intel, and I believe AMD, use for core temps. In this sensor, the sensor sucks some energy out of the electron flow.


Here's part two:
In a given medium electrons will always travel at more or less the same speed. At high load the electrons are not traveling any faster or slower than at idle.
Even if they were the difference in heat given off would be awfully small.
The reason cores run so much hotter at load is that more electrons are moving. A lot more. When a chip is actively calculating it is shuffling electrons in and out of a spectacular number of junctions, capacitors, through resistors, through transistors, all over the place.
Electron movement is, of course, measured in amps (among other measurements). More electrons moving in a given time = more amps.
Add to that the fact that power saving features reduce the clock speed (further reducing the amps) and reduce the voltage (which reduces the amount of energy involved per amp), and you have a chip that runs a lot cooler at idle than it does at full load.

The electric charge travels at about 0.7 c in a decent conductor, for whatever that is worth. The actual electrons travel far slower, a few cm/second. It's rather like those swinging metal ball things right at the moment of impact the energy is transferred from the end the swinging ball hit to the far ball far, far faster than the swinging ball was moving when it hit the other balls.

Why enabling ACC disables the core temp sensors is something I do not know, but it does. It is not involved in their accuracy though.

As for their accuracy, a large part of the idle inaccuracy comes from the traces between the sensor diode itself and the brains behind it that translate the voltage drop into a temperature. The resistance in those traces also causes a voltage drop, and worse that voltage drop varies by temperature.
Intel and AMD calibrate each individual chip's sensor brain to assume a given resistance in those traces, a resistance that corresponds to their real resistance at TJMax.
The further away you get from that temperature the further from that resistance the traces are, and the further out to lunch the sensors are.



TL: DR version:
Electrons only move at one speed.
Electrical charge only moves at one speed.
Chips move more electrons at higher loads.
More electrons dropping more energy = higher temp.
Sensors are calibrated to work when they really need to, at max temps.

nice find, I knew it was something like that, btw can i have your source ? if you don't mind I would like to further read on this subject because my understanding has now changed, as i thought that electrons could change speed
(wikipedia)
Motion and energy

According to Einstein's theory of special relativity, as an electron's speed approaches the speed of light, from an observer's point of view its relativistic mass increases, thereby making it more and more difficult to accelerate it from within the observer's frame of reference. The speed of an electron can approach, but never reach, the speed of light in a vacuum, c. However, when relativistic electrons—that is, electrons moving at a speed close to c—are injected into a dielectric medium such as water, where the local speed of light is significantly less than c, the electrons temporarily travel faster than light in the medium. As they interact with the medium, they generate a faint light called Cherenkov radiation.[124]
The plot starts at zero and curves sharply upward toward the right
Lorentz factor as a function of velocity. It starts at value 1 and goes to infinity as v approaches c.

The effects of special relativity are based on a quantity known as the Lorentz factor, defined as \scriptstyle\gamma=1/ \sqrt{ 1-{v^2}/{c^2} } where v is the speed of the particle. The kinetic energy Ke of an electron moving with velocity v is:

\displaystyle K_{\mathrm{e}} = (\gamma - 1)m_{\mathrm{e}} c^2,

where me is the mass of electron. For example, the Stanford linear accelerator can accelerate an electron to roughly 51 GeV.[125] Since an electron behaves as a wave, at a given velocity it has a characteristic de Broglie wavelength. This is given by λe = h/p where h is the Planck constant and p is the momentum.[48] For the 51 GeV electron above, the wavelength is about 2.4×10−17 m, small enough to explore structures well below the size of an atomic nucleus.[126]
-------------------------------------------------------------------------
How fast do electrons move?

As fast as you can get them going! Well not quite. One of the facts of life discovered in the 20th century is that the speed of light (300,000 kilometers per second) is the ultimate speed limit. As you add energy to the electron, it will go faster, but as you get it to go close to the speed of light, you find that you have to add even more energy just to bump it a bit faster. For example, with just over 220,000 eV (which stands for a convenient unit of energy called the "electron-volt"), you can get the electron up to 90% of the speed of light. But to get it to 99.9% (just another 9.9%), you need a total of over 11 million eV! One way of looking at this is that the electron gets "heavier" (more massive) as it goes ever faster. So it's harder to push it faster. At Jefferson Lab, a typical energy for the electrons in the beam is 4 GeV which is 4 billion eV. That means the electron is traveling at 99.9999992% of the speed of light. Close but still not 100%.

You may wonder how fast the electrons are whizzing around in the atoms around you. A good example (and the most simple to calculate) is the hydrogen atom which is in all our water. A calculation shows that the electron is traveling at about 2,200 kilometers per second. That's less than 1% of the speed of light, but it's fast enough to get it around the Earth in just over 18 seconds. Read up on what happens when nothing can go faster than the speed of light.
-------------------------------------------------------------------------

even with above debate i still agree with the MAIN point the sensor will work under load or *** you put it, 'When it needs to'

-5ghz
Author:

Carl Zorn, Detector Scientist (Other answers by Carl Zorn)
 
The wikipedia article is talking about free electrons, otherwise known as beta radiation.
They're a different beast than conducted electrons such as in copper wires or silicon dies.
The electron itself is almost the same (in conductive mode it's a quasi-electron), but its method and speed of motion are very, very different.

A few links, two wikipedia and one other:
http://en.wikipedia.org/wiki/Speed_of_electricity
http://en.wikipedia.org/wiki/Electron#Conductivity
http://www.eskimo.com/~billb/miscon/speed.html (written by some random HV/electrostatics guy)

It's a fascinating business, really.
 
The wikipedia article is talking about free electrons, otherwise known as beta radiation.
They're a different beast than conducted electrons such as in copper wires or silicon dies.
The electron itself is almost the same (in conductive mode it's a quasi-electron), but its method and speed of motion are very, very different.

A few links, two wikipedia and one other:
http://en.wikipedia.org/wiki/Speed_of_electricity
http://en.wikipedia.org/wiki/Electron#Conductivity
http://www.eskimo.com/~billb/miscon/speed.html (written by some random HV/electrostatics guy)

It's a fascinating business, really.

thank you for that.
 
dh7h2s.jpg

Here are my temps as read by speedfan and coretemp. I have a AMD FX-4170.

I have seen temp1 in speedfan go into the low 50s while playing games such as borderlands and portal 2.

From what I have read here, it sounds like the processor temps I have to worry about are in coretemp, and as long as they don't get up to the high 50s I should be fine with heat?

(I read everything pretty thoroughly, and I know they are not accurate at idle temps, but I can trust them in as much as staying under max temps right?)
 
I pay no attention to speedfan. Don't even use it. HWMonitor mostly and CoreTemp is supposed to be okay. Don't use it either though.

Using CoreTemp >> as long as they don't get up to the high 50s I should be fine with heat? = YES.

Using CoreTemp >> (I read everything pretty thoroughly, and I know they are not accurate at idle temps, but I can trust them in as much as staying under max temps right?) = YES.
 
OP, just looking at your subject line, "Unaccurate . . ."

Think maybe "Inaccurate . . ." is correct English. Picky, I know.
 
I pay no attention to speedfan. Don't even use it. HWMonitor mostly and CoreTemp is supposed to be okay. Don't use it either though.

Using CoreTemp >> as long as they don't get up to the high 50s I should be fine with heat? = YES.

Using CoreTemp >> (I read everything pretty thoroughly, and I know they are not accurate at idle temps, but I can trust them in as much as staying under max temps right?) = YES.

Thank you for the input. I just recently built this computer and I just wanna make sure the CPU is running around where it should be. One thing that bothers me is that the bios claims the CPU temp to be around 45C, that seems kinda high for just looking at the bios, I know there is some inaccuracy in a lot of these numbers but hopefully that is one of them?

If you recommend another program to watch CPU temps (and preferably all temps) I will gladly check it out.
 
using 2 different programs i show 2 different temps

core temp
549736_10151131101701644_1110652053_n.jpg


AMD overdrive
309280_10151131100921644_1894653291_n.jpg


either way this processor runs very cool at load and reads 2-5 cel at idle as well

these temps are under full load (100%) and are on an overclock of 4.6 ghz and water cooled

i have my particular case custom modded for positive air pressure as well and keep my ambient room temp at around 65-75 degrees farenheit

here is how i have my case set up for optimum air flow

252336_10151131107716644_1676439334_n.jpg

45086_10151131108571644_1416830629_n.jpg

9161_10151131109311644_457530076_n.jpg
 
Back