HUGE temperature on AMD Phenom II x4 960T

[Aly007]

Hello,
Today have just arrived my new cooler Xigmatek Loki SD963 and after that I tried to OC my CPU.
I OC it to 3.8ghz @1.4125 V [ 200 x 19 ] + CPU-NB 2400 @1.2500V with C1E, C'n'Q and Spread Spectrum disabled.
I have to mention that the temperatures are ok at 3.4ghz [ 200 x 17 ].
I attached two printscreens, first one contain the temperatures in full load at 3.4ghz and the second one contain the temperatures in full load at 3.8ghz.
CPU temperature seems ok but the CPU Core temperature, the one that really matter it's huge.
It's really strange for me and I really don't know what could be the problem.
Hope I'll get some help from you.

My system specs:

CPU:
AMD Phenom II X4 960T
Motherboard:
Placa de baza MSI 760GM-P21 (FX)
Memory:
KingMax 4GB DDR3 1333MHz
HDD:
SAMSUNG HD200HJ
Graphics card(s):
VTX3D Radeon HD6850 X-Edition 1GB DDR5 256-bit
Power Supply:
Recom Power Engine 450 Plus II
Case:
Segotep C1
Operating System:
Windows 7 x64

 

[capttripppp]

I can't quite tell what your temps are because it is too small. But from what I can tell, it looks like the program is bugging out. If you could check them in the Bios, that would give you a more accurate idea of what your temps are. It's not going to give you full load temps, but it will give you accurate idea of where you stand with your overclock. Also maybe try another monitoring program like cpuid monitor (that's what I use anyway)

 

[Aly007]

Hello capttripppp,
Thank for you response.
Right click on attachment image and click Open, that will open the image on full resolution as I upload them, I didn't make any resizing.

 

[keny]

Download HW monitor and see what that reads as those core temps are not possible so AIDA is reading the sensors wrong , post back with a screen of the HWM window

 

[Aly007]

The HWMonitor temperatures coincide with AIDA64 temperatures.
I tried a different OC setting but the result is the same:
218 * 17.5 = 3817.5 MHz
CPU-NB & HT - 2400mhz
CPU Voltage - 1.4125 V
CPU-NB Voltage - 1.25V

It looks like a sensor fault for me but I'm worried because I don't know what cause this fault...

 

[capttripppp]

I would still check them in bios to make sure. Bios temps are true temps. Cpuid monitor works for AMD very well. I have never had a problem with it, and my system is a PIIx4 965

 

[keny]

Looks like you're stuck with using the socket temp to read temps. Just out of curiosity, you don't happen to have core unlocker enabled in the bios ? as having that enabled screws the temp readings , if it's enabled just disable it and check temps again.

 

[capttripppp]

Had a brainfart and realized that hw monitor is done by cpuid. You could also try core temp or real temp. One works best for Intel, but I can't remember which one. Also as Keny said core unlocker will mess up the temps.

 

[petteyg359]

Download HW monitor and see what that reads as those core temps are not possible so AIDA is reading the sensors wrong

All programs are reading the sensors from the same place. Some have different offsets by default, and some people prefer the interfaces of some programs, but when the temperature sensors are reporting wrong values, NO program will read the correct values. The only "better" program in this case is one where you can manually set the offset and hope that the sensors are somewhat precise, if not accurate.

 

[keny]

Dont know what this temp is but it also seems super high mossfet or SB.Ive ringed it in the immage

 

[Aly007]

I would still check them in bios to make sure. Bios temps are true temps. Cpuid monitor works for AMD very well. I have never had a problem with it, and my system is a PIIx4 965

I've check BIOS temp and it reports the same as any software do.

Looks like you're stuck with using the socket temp to read temps. Just out of curiosity, you don't happen to have core unlocker enabled in the bios ? as having that enabled screws the temp readings , if it's enabled just disable it and check temps again.

No, it's disabled and had been that way all the time.

Had a brainfart and realized that hw monitor is done by cpuid. You could also try core temp or real temp. One works best for Intel, but I can't remember which one. Also as Keny said core unlocker will mess up the temps.

It's not a software fault, I'm sure. Anyway I tried Core Temp too and it's the same thing. I also tried to stress cpu with anoter software [ Intel Burn Test ], same result.

All programs are reading the sensors from the same place. Some have different offsets by default, and some people prefer the interfaces of some programs, but when the temperature sensors are reporting wrong values, NO program will read the correct values. The only "better" program in this case is one where you can manually set the offset and hope that the sensors are somewhat precise, if not accurate.

Couldn't say it better.

Dont know what this temp is but it also seems super high mossfet or SB.Ive ringed it in the immage

It's the "AUX" temperature which it's a bogus one, don't have to worry about it.

What does the "Aux" temperature shown on the Computer / Sensor page mean?

There is no standard for sensor registers layout, so the "Aux" temperature could show the temperature of the CPU or the motherboard, or might be a non-connected wire of the sensor chip, and so it could show a bogus value.

I'm still waiting for something that could point me right while I'm doing hundred of tests with different settings .

 

[keny]

Looks like the sensor on you're CPU is screwed , my 960T is also screwed but not as bad as yours, mine will read well below but never high, I think these chips are coming out with bogus readings more often than not lately, I read my core temps by a rough guess of -7c lower than my CPU temp which it looks like you're mobo is reading correctly, that's all I can suggest and as I say "that's what I'm doing"

 

[Aly007]

I succeed to stabilize it at 3.6ghz 16 x 225 @ 1.3125 V CPU Vcore and CPU-NB at 2250 @ 1.2500 V.

I've also changed the thermal compound few minutes ago to be sure that's not the problem and at 3.7ghz after 15 minutes in Prime95 I got it again. Maybe the sensor it's not able to read temperature on high frequency/voltage ? Does it make any sense what I'm saying ?

I'm still waiting for someone who might know where is the problem here...

 

[RGone]

I'm still waiting for someone who might know where is the problem here...

Internal cpu temp sensor bad. RMA the cpu seems prudent. All this based on your board and bios actually supporting the 960T processor and you have n0t tried to unlock any cores. Often trying to unlock cores and then n0t unlocking cores must have a clear CmOs in between.

 

[capttripppp]

keny pretty much told you the problem. The sensor is not working correctly. Did you check the temp in the bios? That's where you need to check it. If it is messed up there then you know for sure that the probe is not reading correctly. If it is good then it's the program not reading it right

 

[trents]

keny pretty much told you the problem. The sensor is not working correctly. Did you check the temp in the bios? That's where you need to check it. If it is messed up there then you know for sure that the probe is not reading correctly. If it is good then it's the program not reading it right

The temperature in bios would not be from the CPU core but from the motherboard socket area below the processor. Two different sensors.

 

[Frakk]

With 184c Cores i think you will have had to suck them off the back side of your GPU with a straw....

 

[Aly007]

Internal cpu temp sensor bad. RMA the cpu seems prudent. All this based on your board and bios actually supporting the 960T processor and you have n0t tried to unlock any cores. Often trying to unlock cores and then n0t unlocking cores must have a clear CmOs in between.

Ok, I understand, it's a sensor problem. BTW, I did a clear cmos after I tried to unlock any cores.

keny pretty much told you the problem. The sensor is not working correctly. Did you check the temp in the bios? That's where you need to check it. If it is messed up there then you know for sure that the probe is not reading correctly. If it is good then it's the program not reading it right

Yes, I already told you, the temperatures in BIOS are the same as any other software reports. The temperatures in Idle are always fine, the problem is in full load.
For example at 3.4ghz I have no sensor fault in full load.

The temperature in bios would not be from the CPU core but from the motherboard socket area below the processor. Two different sensors.

With 184c Cores i think you will have had to suck them off the back side of your GPU with a straw....

Hahaha, right .

So, under the circumstances, should I continue to OC or I should stop ?
I know the cores temperatures by reading the CPU temp - 10

 

[RGone]

IF the temp sensor is messed-up...well what else is wrong with that cpu? RMA it while you can and is new.

 

[Aly007]

Everything else works fine. BTW, what RMA stands for ?