AIDA64 2.70.2212b temperature errors with AMD A10-5800K + Gigabyte F2A85X-UP4

[LeeGordon]

The screen captures attached (CPU power ~ 100 watts AND CPU Power ~50 watts) shows several AIDA64 version 2.70.2212 beta problems with a Gigabyte Motherboard (GA-F2A85X-UP4)and a Trinity AMD A10-5800K Processor in an FM2 socket. The hardware configuration is:

APU AMD Trinity A10 5800K

APU Cooler Thermaltake WATER2.0 Performer

Motherboard Gigabyte GA-F2A85X-UP4

RAM CORSAIR Vengeance 8GB (2 x 4GB) 240-Pin DDR3, 11-11-11-27 1.5V

SSD 128G|OCZ VTX4-25SAT3-128G R. SATA 6GB Read/Write ~ 500 MBs

Power Supply ANTEC|500W EA-500D GREEN RT. Efficiency minimum 80%

The first most obvious problem is that there is no case temperature shown.

The second most obvious problem is that the embeded Graphics Processing Unit (GPU) temperature doesn't seem to be related to any reality. This is probably due to the FM2 Socket outputs.

The third problem is that all temperatures except the motherboard temperature have significant calibration errors. From what I have been reading on the internet, is there is a conspiracy fostered by AMD to keep their temperature calibration techniques classified as a trade secret. Although the vendor of Gigabyte's EasyTuen6 has been able to get the correct information. Some web sites have indicated that there is no thermal accuracy unless the CPU is dissipating enough power to raise the case temperatures.

As part of an aerospace design team, over the past 30 years, I have built and designed many complicated SOI CMOS circuits for both military and commercial satellites use. I suspect that you may need to read the device information differently and add a multiplier and offset to get near the correct value. If your company has enough clout with AMD you may be able to get the correct information.

Please let me know if you have any plans to correct the situation. I will be willing to beta test any solutions you present.

Thanks for your prompt reply - Lee Gordon - lhgordon@cox.net

AIDA bad temperatures.pdf

[Fiery]

We've asked our contact at AMD about this issue. What I can tell you right away is that since many Trinity processors seem to be affected by this anomaly, it definitely looks like it's a hardware issue rather than a software issue (or AIDA64 bug). I'll let you know in this topic once we get more info from AMD about this.

[LeeGordon]

Your statement about AMD is no surprise. I want to mention serveral issues:

1 -The Gigabit utility is the only one that got it right. (see my attachment showing the side by side comparisons). I suspect that AMD wrote the Gigabit utility.

2 - It could be that the ADIA64 developers are looking at the wrong registers since this part fits in a new socket with a different pinout (FM2).

3 - The Gigabit utility shows the CPU case temperature rather than the core temperatures and the A10 designation is probably confusing some developers.

4 - My Attached PDF file shows data taken at 100watts and 50watts. I wouldn't mind if the temperature indicated "<30C" or some other arbiturary temperature as long as the temperatures are correct at higher values.

5. It really irks me that AMD shares their "application secrets " with some vendors and not others.

I really love to keep the Aida64 gadget on the right side of my monitor just to keep track of what is changing.

Keep trying it really would be appreciated. - Lee Gordon

[Fiery]

Meanwhile we've found a bug in AIDA64 about the "CPU" temperature reading that is shown by Gigabyte EasyTune6, but not measured by AIDA64 properly. Please upgrade to the latest beta version of AIDA64 Extreme Edition available at:

http://www.aida64.co...15vk0gp7mfzyzip

After upgrading to this new version, make sure to restart Windows to finalize the upgrade.

Let me know how it works. Please note that this fix will not help about the very low core temperatures.

Thanks,

Fiery

[LeeGordon]

Hi, your work seems to be progressing nicely.

Please note This version of AIDA64 Tracks the core temperatures and the CPU case temperatures.

At low CPU power the cores and the case should be the same. At high power the cores should be approximately 10 to 15 degrees C higher than the case depending on the thermal resistance between the substrate and the case. It seems to me adding aproximately 37degrees C to the core readings would provide reasonably accurate temperatures.

I am not at all sure what you should do about the GPU since I have not included it in the stress tests above.

The GPU diode temperature seems to be different than the core temperature sensors. The GPU diode must be activated in the Preferences during stress testing and then only shows up when the test is running.

If I am reading the stress chart correctly, only core#4 shows on the thermal plot. Missing Cores 1,2,3,& GPU.

Keep up the good work – Lee Gordon

I have a PDF I would like to add to this but cannot figure out how to do it

[Fiery]

As a temporary solution, you can use the Correction feature of AIDA64 (main menu / File / Preferences / Hardware Monitoring / Correction) to apply a ratio and/or an offset on the temperature readings.

[NoiseBoy]

I stumbled across this, perhaps the software has been updated since but the CPU temp in AIDA64 is identical to the temp displayed in EasyTune6 on my system.  I cant be certain that ET6 is correct because the temps seem suspiciously lower than that displayed in the BIOS.  I notice there is a beta release BIOS version for the gigabyte A85X-UP4 motherboard which mentions 'EasyTune6 Compatibility' but as my system runs well as it is i don't intend to update just yet.   A few people have mentioned in various places that the A10 runs cooler when idling under windows than it does idling in the BIOS which could explain the odd temps.