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

Abit KD7 owners! Watch out...

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

@md0Cer

Senior Member
Joined
Nov 6, 2003
Location
Denver, CO
The socket diode reads very low. I always thought something was up when my northbridge temp sensor read one or 2 degrees higher than the socket. It was not until recently that I learned with my socket at 42°C load, and the system temp (right near the northbridge) at 43.5°C, that the diode inside the CPU reads 78°C!!!
:eek: :eek:


I am pretty sure this would be the accurate reading and not motherboard dependant becuase the core temp is from a diode integrated in the CPU right? So just about everyone's AMD probably is pretty acurate.

Although the socket temp, the one I think is reading VERY low, does get more accurate with lower temps. With my CPU running at 800Mhz and 1.0Vcore, it is only putting out 10.6 thermal watts compared to 115 before when it was overclocked. At full load, it reads 23 system (by the northbridge), 24 socket, and 25.5 internal core.

So anyways, I now know not to trust the socket temp on my motherboard. I hope this helps out any other KD7 owners.
 
As much as possible everybody should go by the DIODE temps instead of the sockent sensor. When setting up MBM make sure to use the LM90 REMOTE for your CPU temp sensor on the KD7 series.

Nobody can say how much thermal wattage their CPU is putting out, what's even worse is when they compute their C/W values with wrong information. The diode can only be as accurate as to how & what it was calibrated too then there is the issue of your motherboard reading the signal right.
 

Attachments

  • mbm.jpg
    mbm.jpg
    45.3 KB · Views: 908
I agree. Diode temps first. And if it was at 78°C with 42°C socket, can anyone here think what my CPU core temp was when I ran my socket at 52°C? Heh...my estimate is probably around 90 to 100°C.
 
LOL. My socket temp hovers around 56C under load but the diode temp is 64C :eek: Of course, that's an Intel chip so it's even worse... :-/
 
n17ikh said:
LOL. My socket temp hovers around 56C under load but the diode temp is 64C :eek: Of course, that's an Intel chip so it's even worse... :-/

yea. Well your socket is probably fairly accurate. For mine to be lower than the system temp, and MUCH MUCH MUCH lower than the diode, I just dont trust it anymore.
 
No one should ever trust socket readings, and considering that's all most motherboards even report, I don't bother with my temp readings or anyone elses.
 
So how do you read the diode in the chip other than on the KD7?
 
On the NF7-S the on-die diode is tied directly to an overtemp cutoff circuit (I believe), so you can't read from it inless you first cut that ciircuit, and then connect your own monitoring chip to the SMBus and the correct pions on the CPU. People were making these a while back when most boards didn't connect anything to the diode, so there are some threads around somewhere and possibly an article or two on the overclockers.com home page.
 
This fell off the first page, I think it deserves a bump. *bump*
 
I believe the case temp sensor on the KD7 reads high. It's places so it gets a lot of hot air coming off the HSF. And since the insocket one doesn't touch the back of the CPU, I often find that the case temp is higher than the CPU temp,, unless I use the on-die diode. I don't see why anyone would use the insocket whaen they could get more accurate readings from the on-die diode, unless their board doesn't read the diode..
 
Gnufsh said:
On the NF7-S the on-die diode is tied directly to an overtemp cutoff circuit (I believe), so you can't read from it inless you first cut that ciircuit, and then connect your own monitoring chip to the SMBus and the correct pions on the CPU. People were making these a while back when most boards didn't connect anything to the diode, so there are some threads around somewhere and possibly an article or two on the overclockers.com home page.

Gnufish, could you desolder the socket diode, cut the circut for the chip diode, then wire the chip diode into the socket diode monitor? Would this work? Do you understand what I'm saying?
 
I understand the idea of what you're saying, if nothing else. I think that would be awesome if it could be done.

Anyone an electrical engineer? :D
 
Cyrix_2k said:


Gnufish, could you desolder the socket diode, cut the circut for the chip diode, then wire the chip diode into the socket diode monitor? Would this work? Do you understand what I'm saying?

Maybe. If the specs for the diodes are similar it might work, but it would need to be calibrated. Also, a lot of boards use thermisistors in the socket, which definately would not work. This could use some looking into. I don't, however, have a board to test this on (just my trusty KD7, which reads the diode already, and my flaky ECS K7S5A which has a thermisistor in the socket).
 
Back