Embedded Controller accuracy when monitored by HWInfo

sotoole

New Member
I'm getting a great CPU(package) temp and decent socket temp according to HWInfo, but the ASUS EC that's sensed by HWInfo shows cores 1 & 8 temps way over limit in Prime95.  Core1 10°C higher at idle and 87°C @ load, core8 5°C higher @ idle and 75°C @ load.  Am I to be concerned about those two core temps or is the EC not accurate?  I don't know enough how these temps are sensed to make a judgement call on this.  A program called CoreTemp shows individual core temps (that never differ from core to core that I could observe) close to the HWInfo package temp and close to the CAM software CPU temp reading.

4.8 GHz at 1.44v and Ultra High LLC setting is stable in Prime95 for 20 mins minimum.  I stopped the test due to the EC's core 1 temp, but the max for the package was 53°C (avg 49°C).




build(will look for build signature):
fx8370 | ASUS Sabertooth 990fx R2.0 | G.Skill Ares (4x4Gb) 2133 MHz CL9
NZXT 500i with 2 fan exhaust | NZXT Kraken72 240mm (2push & 2pull)
soon to have 3D printed VRM exhaust shroud in pull config
 
Those CPUs (actually most AMD series before Ryzen) cannot provide accurate temperature, which has been described several times here, e.g. https://www.hwinfo.com/forum/Thread-Low-CPU-Tctl-temp-value
One can only rely on external socket temperatures, which however don't reflect the internal (core) temperatures.
Note that AMD CPUs report only a single temperature for the entire chip. If you see per-core values in other tools, then it's just a duplicate of the single value.
 
Martin said:
Those CPUs (actually most AMD series before Ryzen) cannot provide accurate temperature, which has been described several times here, e.g. https://www.hwinfo.com/forum/Thread-Low-CPU-Tctl-temp-value
One can only rely on external socket temperatures, which however don't reflect the internal (core) temperatures.
Note that AMD CPUs report only a single temperature for the entire chip. If you see per-core values in other tools, then it's just a duplicate of the single value.

Hi Martin!
Thanks for the link and quick reply.  I've read it multiple times at this point, but cannot get past the annecdotal evidence that ASUS's EC shows distinct, non-duplicated values (in HWInfo).  Core1 & core8 are HOT.  Any insight you can provide on the interworkings/legitimacy of the EC would be appreciated.  All the information from the interwebs read the same... it's a microchip on the mobo that grabs the temperature and uses it to throttle, etc.

Thanks.
 
You cannot generalize EC - it's a proprietary design and no one except ASUS knows what exactly it reads and how. It's fully up to its firmware and capabilities.
But I believe it has only 2 options - either read the unreliable internal temperature or the socket temperature from a diode on the mainboard in the socket.
 
This is the answer that I have searched for all along. It's been alluded to elsewhere on the web, but nothing as firm as this. Thanks, Martin, for providing insight on this topic. I don't have the background for EC's and may have become paralyzed with my OC. FYI, oc'ing has resumed and my 24/7 settings are 4.785 GHz @ 1.392 V. Under load with Prime95 it is 41°C and I'm happy. Trusting the info in the above replies, I will eventually try my previous 5.0 GHz setting that can run for a couple hours of gaming at a time. Again, Thanks!
 
Back
Top