[SOLVED] Strange Temps on water

[blazerboy]

I have a Xeon 1285l-V4, just redid the cooling system on it with a new water cooled setup. Running high 50's on max Intel burntest and prime95 with sub 70F ambient. My package Temps in hw monitor are solid within a few degrees of cores. Once I verified cooling effectiveness, I have been running it as intended (movie server / camera system / NAS). Loads never really pass 50% for more than a few moments. Leaving it on 24/7 and monitoring Temps I hit 32-40c on my cores and package. On occasion I will check into my server in the morning and see that my max on package hit 90-95c, but the cores never are beyond the normal 40c or so. How can this be? Mind you iv never witnessed the package temp being that high myself, simply what the max on hw monitor are reporting when I check into it after running for several days to weeks. Thoughts?

 

[blazerboy]

Image

 

[Karadjgne]

Package temp is the temp of the IHS where it meets the cooler in engineering samples, but generally in retail cpus it's the hottest core, bios temp. Not sure how that could possibly be in the 90's with a constant load cooler and cores that don't exceed the 50's.

Unless the cooler has air getting trapped or the pump has momentary stoppages lasting 2 digit second counts.

 

[blazerboy]

I ran with a Graphite pad so air getting trapped is highly unlikely. Pump stoppage would cause the cores to jump up in Temps as well though

 

[blazerboy]

Just pushed up again to 93c on the package. 93c package - 38 core

 

[EndEffeKt_24]

Most likely a messurement error. I would not bother unless you really see those temps live and can verify those are accurate.

 

[blazerboy]

I'm not sure how I can physically verify package Temps. What is odd is that this problem never occurred on my old setup even if it ran much hotter it never hit 93c on the package.

 

[Karadjgne]

Not air between the cpu/block, air in the tubing. All it takes is 1 decent bubble to hit the block and if it's timed right at the 256ms mark where the temps get registered, you'll see the spike.

I'd run Afterburner, or something else with a graphing capability that's set for overnight use, and check to see if it's a momentary spike, or a workload draw/pump slowdown taking minutes etc. Personally, I'd be inclined to believe it's a spike from a bad read, but could be a power fluctuation/low power state causing pump issues after long idle periods. Can maybe fix that by disabling low power states below C3, which keeps sleep, but disables ultra deep sleep.

I'd also disable hibernation, totally-permanently, it's a desktop not a laptop.

Edit: was just brought up to me that it's entirely possible that's it's a software anomoly from the use of Hwmonitor (say thank you @CompuTronix, local Intel temp-god type person), and he has suggested that you forgo the use of that software and use HWInfo (sensors only) instead, as it's proven accurate and reliable.

If that too shows the package temps skyrocketing out of the blue, then you'd have cause to dig deeper, but if HWInfo checks all as good, I'd assume all is good and there's no worries.