E8400 temp & IP35 fan control

[int]

Yet another thread from me, but I need more help

I've seen a fair few threads around, discussing problems with E8400 temperature detection and what software is best to get a temp as accurate as possible. However there didn't actually seem to be a "victor". I'm using speedfan, coretemp, and the abit uguru that comes with the board. All of these are giving wildly different values (at idle 31, 45, 29 respectively).

Has anyone managed to work out what program is best for these CPUs?

Secondly, speedfan and uguru are (in my opinion) terrible at controlling fan speed. Are there any decent programs lurking about that has a properly functioning auto-adjust function?

e: Figured I'd throw another problem in here as it has some relevance to overclocking.
I've disabled the speedstepping for the cpu, however cpuz still shows x6 multipliers idling and x9 when under load. Is this just cpuz being picky or do I have to do more than disable ESID in the BIOS?

 

[Fothsn]

Just like mine, mine are 40c / 52c there is a big gap between each core, no matter what tool I use, my bios seems to read the temps properly, they are reported 37c / 38c with no big temperature gap between the cores.

 

[int]

Aye, I had noticed the large gap (usually 8-10C) between both cores, the main CPU temp reading seem to be an average of the two cores.

Seems to be a huge problem with the E8400

 

[GunRunner]

I would be more inclined to accept the BIOS readings as being the more accurate.

Have you tried Real Temp version 2.5?


http://www.techpowerup.com/realtemp

You can check the temperature sensors to see whether they are working or not.

These are the main features according to the site.


Reads temperature information from all Intel Core based processors.
Ability to individually calibrate Real Temp for each core of your CPU.
Program is based on temperature data gathered using a Fluke 62 IR Thermometer.
Test Sensors feature will check your DTS sensors for any sign of problems.
Keeps track of Minimum and Maximum temperatures with full logging features.
Reporting and logging of the Intel PROCHOT# thermal throttle activity bit.
Quick, very accurate and repeatable benchmark.
Displays MHz, TjMax, VID, CPUID and Calibration settings.
No installation or registry modifications required.
Support for Windows 2000 / XP / Vista (32 and 64 bit versions)

 

[int]

Yeah I'm using real temp now as it seems to be the closest to my BIOS. However the difference in temperature between cores is getting ridiculous now. As I look at it now, Core #0 is 38c and Core #1 is 26. I've applied AS5 and reseated the heatsink a couple of times and it's still coming up with this huge difference. I'm tempted to take out the CPU and see if re-seating it will help.

 

[nightwish]

I'm pretty sure it won't. The whole problem with these CPUs is that they simply don't read the temperature correctly.

I highly doubt your cores are actually 12 degrees apart, but if that's what the sensors are saying then that's what will be read by any temp program. The point is that it's the temp sensors that are the problem, not the temps themselves.

Have you tried using testing for stuck sensors in RealTemp?

 

[GunRunner]

I removed an E6600 today and fitted a new E8400. With it running at 3.8Ghz on vcore auto(shows 1.30 volts) and prime 95, I get the following readings: (Note. I have and Arctic freezer 7pro that is set in the BIOS at 100% on an EVGA 780i).

Core temp:
core 0 = 57
core 1 = 57


CPUID Hardware Monitor

Core 0 = 57
Core 1 = 57

Real temp 2.5

Core 0 = 48
Core 1 = 48

When I look in the BIOS on boot up, the CPU temp (average of both) shows 44 degrees. With this variation, I am unsure which is the one to believe. Perhaps the highest is the best to believe to be on the safe side. Perhaps the latest batch of CPUs has the temperature difference between cores fixed?

 

[int]

Ack, wish I could replace the one I've got

e: I bought it under a fortnight ago too...

@nightwish: My sensors are both moving at 7, well within the expected range. I think the sensors are just duff.

 

[weescott]

The sensors ARE duff. Even with phase the sensors differ a large amount. I have seen countless threads on this. Take the read out from the CPU socket if your board supports it and guesstimate from there, or use a probe at the side of the CPU.

 

[int]

I'm pretty sure a couple of the programs I've used take readings from the CPU socket as opposed to directly from the cores. My BIOS only shows a CPU reading, not a core reading and the uGuru software with this board reflects that reading.